💾 Archived View for blitter.com › apl-books › APLSE › www.sigapl.org › Archives › waterloo_archive ›… captured on 2024-08-18 at 19:33:50.
View Raw
More Information
⬅️ Previous capture (2022-07-17)
-=-=-=-=-=-=-
From owner-jsoftware@lists.interlog.com Thu Apr 1 15:27:42 1999
Date: Thu, 01 Apr 1999 14:23:58 -0500
From: David Ness <DNess@Home.Com>
Subject: Re: Jforum: Hint needed: Quad
References: <199903301947.VAA20632@isis.wu-wien.ac.at>
Dr. Faller's recent question reminded me that I have never known
how to avoid the carriage return/linefeed that occurs after the
prompt when you do something like:
x =: 1!:1]1['Prompt Here' 1!:2]2
peter.faller@wu-wien.ac.at wrote:
>
...
--------------------------------------------------------------------------------
J Forum: for information about this list, see http://www.jsoftware.com/forum.htm
From owner-jsoftware@lists.interlog.com Thu Apr 1 17:16:11 1999
Date: Thu, 01 Apr 1999 16:56:42 -0500
From: "Joseph G. Kunkel" <joe@bio.umass.edu>
Organization: University of Massachusetts
Subject: Jforum: Distribution of one's scripts.
References: <199903301947.VAA20632@isis.wu-wien.ac.at> <3703C7CE.E9F5E882@Home.Com>
I have just recently started converting my APL (STSC ver. 8) programs to
J and am just starting on my third 14 day trial of J 4.00.5000. As I
have lapsed for a few days between trials I have seen that my '.ijr'
scripts will still work as runtime programs using the lapsed j.exe
despite the fact that I can not modify them while lapsed other than by
programmed input.
I guess I can conclude that a colleague of mine does not have to
purchase the J program at all to run my runtime scripts. He or she just
need download a trial version of J that works on their machine and run
my runtime script or compile my '.ijs' form into a runtime on their
machine. Then they can continue to run the program using the runtime
capability even when their free trial runs out. I find that pretty nice
if I am correct. I had never been able to get my hands on the
compilable version of APL.
Are runtime scripts (which look like binary when you try to load them
into an editor) specific to an operating system or does the j.exe of
each operating system interpret them and they will run on any J-capable
operating system?
Joe
--
______________________________
Joe Kunkel, Professor
Biology Department, University of Massachusetts at Amherst
joe@bio.umass.edu http://www.bio.umass.edu/biology/kunkel/
--------------------------------------------------------------------------------
J Forum: for information about this list, see http://www.jsoftware.com/forum.htm
From owner-jsoftware@lists.interlog.com Thu Apr 1 18:26:56 1999
Date: Thu, 01 Apr 1999 18:09:21 -0500
From: Eric Iverson <eiverson@interlog.com>
Organization: Iverson Software Inc.
Subject: Re: Jforum: Distribution of one's scripts.
References: <199903301947.VAA20632@isis.wu-wien.ac.at> <3703C7CE.E9F5E882@Home.Com> <3703EB9A.6EDEF9A8@bio.umass.edu>
Joseph G. Kunkel wrote:
> I guess I can conclude that a colleague of mine does not have to
> purchase the J program at all to run my runtime scripts.
They do not even have to download and install a J version. You can
freely distribute the required j.exe and j.dll files along with your ijr
files to anyone.
You might want to look at the help documentation on building a proper
setup for distributing runtime applications.
You do not need to encode your ijs files as ijr files to use the
runtime. You can distribute ijs files as a runtime application.
We hope your colleagues are sufficiently impressed with your
applications that they want to get their own licensed copy and join in
the fun.
> Are runtime scripts (which look like binary when you try to load them
> into an editor) specific to an operating system or does the j.exe of
> each operating system interpret them and they will run on any J-capable
> operating system?
J runtimes and the associated ijr files are only supported in J for
Windows, Mac, and Wince. There is no J Runtime or ijr support in the
Unix versions.
--------------------------------------------------------------------------------
J Forum: for information about this list, see http://www.jsoftware.com/forum.htm
From owner-jsoftware@lists.interlog.com Thu Apr 1 22:53:18 1999
From: "Roger Hui" <RHui@interlog.com>
Subject: Re: Jforum: Hint needed: Quad
Date: Thu, 1 Apr 1999 22:35:31 -0500
charset="iso-8859-1"
David Ness writes on Thursday, April 1:
>Dr. Faller's recent question reminded me that I have never known
>how to avoid the carriage return/linefeed that occurs after the
>prompt when you do something like:
> x =: 1!:1]1['Prompt Here' 1!:2]2
There is no way, but the 11!:0 WD facility is a superior substitute.
--------------------------------------------------------------------------------
J Forum: for information about this list, see http://www.jsoftware.com/forum.htm
From owner-jsoftware@lists.interlog.com Fri Apr 2 03:21:55 1999
From: "Leo V�handu" <leov@cc.ttu.ee>
Subject: Jforum: Divisors quickly
Date: Thu, 3 Dec 1998 10:54:20 +0200
charset="iso-8859-4"
Content-Transfer-Encoding: 8bit
Dear Forum! I wish to find all divisors of integer numbers. It can be done
easily using Phrases book.
m4=: >: @ i. @(>./) |/] m5=: 0: = m4 m23=: <: @(+/\) i. i. @ (+/)
F.e. >: m23 m5 18 and you get 1 2 3 6 9 18. On my computer it works
nicely up to 6000000. After that the calculations get verry slow (about 20
and more times slower). Prying into the Phrases gives to me an impression
that this elementary calculation can be done much more quickly. (Either by
diminishing somehow the amount of slashes or using the quicker q: verb). Can
any of old J Hands help me out. I would be glad.
Leo V�handu.
--------------------------------------------------------------------------------
J Forum: for information about this list, see http://www.jsoftware.com/forum.htm
From owner-jsoftware@lists.interlog.com Fri Apr 2 08:11:08 1999
From: "Chris Burke" <cdburke@interlog.com>
References: <000e01be1e9a$97260240$33fe28c1@user>
Subject: Re: Jforum: Divisors quickly
Date: Fri, 2 Apr 1999 07:50:48 -0500
charset="iso-8859-4"
Content-Transfer-Encoding: 8bit
sort , */ &> { <@(1&,)@(*/\) /.~ q: 120
1 2 3 4 5 6 8 10 12 15 20 24 30 40 60 120
----- Original Message -----
From: Leo V�handu <leov@cc.ttu.ee>
Sent: Thursday, December 03, 1998 3:54 AM
Subject: Jforum: Divisors quickly
> Dear Forum! I wish to find all divisors of integer numbers. It can be done
> easily using Phrases book.
> m4=: >: @ i. @(>./) |/] m5=: 0: = m4 m23=: <: @(+/\) i. i. @ (+/)
> F.e. >: m23 m5 18 and you get 1 2 3 6 9 18. On my computer it works
> nicely up to 6000000. After that the calculations get verry slow (about 20
> and more times slower). Prying into the Phrases gives to me an impression
> that this elementary calculation can be done much more quickly. (Either by
> diminishing somehow the amount of slashes or using the quicker q: verb).
Can
> any of old J Hands help me out. I would be glad.
> Leo V�handu.
--------------------------------------------------------------------------------
J Forum: for information about this list, see http://www.jsoftware.com/forum.htm
From owner-jsoftware@lists.interlog.com Fri Apr 2 09:02:17 1999
From: "Roger Hui" <RHui@interlog.com>
Subject: Jforum: Re: Divisors quickly
Date: Fri, 2 Apr 1999 08:46:11 -0500
charset="iso-8859-4"
Content-Transfer-Encoding: 8bit
Leo V�handu writes on Friday, April 2:
>Dear Forum! I wish to find all divisors of integer numbers. It can be done
>easily using Phrases book.
>m4=: >: @ i. @(>./) |/] m5=: 0: = m4 m23=: <: @(+/\) i. i. @ (+/)
>F.e. >: m23 m5 18 and you get 1 2 3 6 9 18. On my computer it works
>nicely up to 6000000. After that the calculations get verry slow (about 20
>and more times slower). Prying into the Phrases gives to me an impression
>that this elementary calculation can be done much more quickly. (Either by
>diminishing somehow the amount of slashes or using the quicker q: verb). Can
>any of old J Hands help me out. I would be glad.
In addition to the solution that Chris Burke gave, there is also:
odometer=: #: i.@(*/) NB. phrase book, chapter 7, m33
divisors=: /:~ @: ({. */ . ^"1 odometer@:>:@:{:) @: (__&q:)
divisors 120
1 2 3 4 5 6 8 10 12 15 20 24 30 40 60 120
__ q: 120
2 3 5
3 1 1
odometer >: 3 1 1
0 0 0
0 0 1
0 1 0
0 1 1
1 0 0
1 0 1
1 1 0
1 1 1
2 0 0
2 0 1
2 1 0
2 1 1
3 0 0
3 0 1
3 1 0
3 1 1
See also the index entry for "Divisors of" in the phrasebook, from which
you find chapter 7, m66.
--------------------------------------------------------------------------------
J Forum: for information about this list, see http://www.jsoftware.com/forum.htm
From owner-jsoftware@lists.interlog.com Fri Apr 2 09:11:04 1999
Date: Fri, 02 Apr 1999 08:57:49 -0500
From: "Kenneth E. Iverson" <kei@interlog.com>
Organization: Iverson Software Inc.
Subject: Re: Jforum: Divisors quickly
References: <000e01be1e9a$97260240$33fe28c1@user>
Content-Transfer-Encoding: 8bit
Leo V�handu wrote:
>
> Dear Forum! I wish to find all divisors of integer numbers. It can be done
> easily using Phrases book.
> m4=: >: @ i. @(>./) |/] m5=: 0: = m4 m23=: <: @(+/\) i. i. @ (+/)
You might try the following:
t=: #:@i.@(2:^#)
u=: t */ . (^~) ]
v=: /:~@~.@u@q:
v 24
1 2 3 4 6 8 12 24
q:24
2 2 2 3
t q: 24
0 0 0 0
0 0 0 1
0 0 1 0
0 0 1 1
0 1 0 0
0 1 0 1
0 1 1 0
0 1 1 1
1 0 0 0
1 0 0 1
1 0 1 0
1 0 1 1
1 1 0 0
1 1 0 1
1 1 1 0
1 1 1 1
u q: 24
1 3 2 6 2 6 4 12 2 6 4 12 4 12 8 24
v 24
1 2 3 4 6 8 12 24
$ a=: v 1234567890x
48
12 4 $ a
1 2 3 5
6 9 10 15
18 30 45 90
3607 3803 7214 7606
10821 11409 18035 19015
21642 22818 32463 34227
36070 38030 54105 57045
64926 68454 108210 114090
162315 171135 324630 342270
13717421 27434842 41152263 68587105
82304526 123456789 137174210 205761315
246913578 411522630 617283945 1234567890
12 4 $ 1234567890 % a
1234567890 617283945 411522630 246913578
205761315 137174210 123456789 82304526
68587105 41152263 27434842 13717421
342270 324630 171135 162315
114090 108210 68454 64926
57045 54105 38030 36070
34227 32463 22818 21642
19015 18035 11409 10821
7606 7214 3803 3607
90 45 30 18
15 10 9 6
5 3 2 1
--------------------------------------------------------------------------------
J Forum: for information about this list, see http://www.jsoftware.com/forum.htm
From owner-jsoftware@lists.interlog.com Fri Apr 2 09:19:24 1999
From: "Roger Hui" <RHui@interlog.com>
Subject: Jforum: Re: Divisors quickly
Date: Fri, 2 Apr 1999 08:59:53 -0500
charset="iso-8859-4"
Roger Hui writes on Friday, April 2:
> odometer=: #: i.@(*/) NB. phrase book, chapter 7, m33
> divisors=: /:~ @: ({. */ . ^"1 odometer@:>:@:{:) @: (__&q:)
div1=: /:~ @: ((*/ . ^"1 odometer@:>:)/) @: (__&q:)
--------------------------------------------------------------------------------
J Forum: for information about this list, see http://www.jsoftware.com/forum.htm
From owner-jsoftware@lists.interlog.com Fri Apr 2 10:19:44 1999
From: "Leo V�handu" <leov@cc.ttu.ee>
Subject: Re: Jforum: Divisors quickly
Date: Thu, 3 Dec 1998 17:54:58 +0200
charset="iso-8859-4"
Content-Transfer-Encoding: 8bit
Thanks Chris ! This program works really quickly. Even with lon numbers. LV
-----Original Message-----
From: Chris Burke <cdburke@interlog.com>
Date: 02. aprill 1999. a. 15:12
Subject: Re: Jforum: Divisors quickly
> sort , */ &> { <@(1&,)@(*/\) /.~ q: 120
>1 2 3 4 5 6 8 10 12 15 20 24 30 40 60 120
>
>----- Original Message -----
>From: Leo V�handu <leov@cc.ttu.ee>
>To: <forum@jsoftware.com>
>Sent: Thursday, December 03, 1998 3:54 AM
>Subject: Jforum: Divisors quickly
>
>
>> Dear Forum! I wish to find all divisors of integer numbers. It can be
done
>> easily using Phrases book.
>> m4=: >: @ i. @(>./) |/] m5=: 0: = m4 m23=: <: @(+/\) i. i. @ (+/)
>> F.e. >: m23 m5 18 and you get 1 2 3 6 9 18. On my computer it works
>> nicely up to 6000000. After that the calculations get verry slow (about
20
>> and more times slower). Prying into the Phrases gives to me an impression
>> that this elementary calculation can be done much more quickly. (Either
by
>> diminishing somehow the amount of slashes or using the quicker q: verb).
>Can
>> any of old J Hands help me out. I would be glad.
>> Leo V�handu.
>
>
>
>---------------------------------------------------------------------------
-----
>J Forum: for information about this list, see
http://www.jsoftware.com/forum.htm
>
--------------------------------------------------------------------------------
J Forum: for information about this list, see http://www.jsoftware.com/forum.htm
From owner-jsoftware@lists.interlog.com Fri Apr 2 12:47:57 1999
Date: Fri, 2 Apr 1999 19:30:35 +0200 (MET DST)
From: Martin Neitzel <neitzel@gaertner.de>
Subject: Jforum: Prompting on a line
Hello jugglers,
somebody on the forum list asked how to create a decent prompt
without a newline. (That is, the equivalent of an "echo -n" in
Unix.) The problem was that
'Prompt here ' (1!:2) 2
involves a sometimes unwanted linefeed.
On J &. Unix, it is easy to prompt without the additional newline.
Just use the stdout channel (4) or the "diagnostic" channel (5).
That is:
answer =: 1!:1 1: 'Prompt here ' (1!:2) 5
The character arguments and result respectively are handled
faithfully when you do 1!:1 and 1!:2 with the 3 4 5 channels.
1!:2&2 and 1!:&1 will add/strip line ends.
The stdin/stdout/stderr channels (3 4 5) were introduced with J 3.01
for Unix. I had asked for them because a clean distinction between
programmer IO (entering J phrases and showing the results & errors)
on the one hand and data IO (raw data, result data) is so important
in Unix. I wanted it to be possible to redirect data to/from files
using the usual "| < >" mechanisms.
It was always possible to redirect the "programmer input" by the use of
script arguments. The system will currently jump into interactive mode
on error, though.
The J system will also try to interpret any stdin which hasn't been
read in and digested by scripts. Therefore:
neitzel 5 > echo "+/i.10" | \j --
+/i.10
45
neitzel 6 >
Depending on yor setting and perspective, this can be useful or
annoying. I wanted to let you know that ISI is open to your
suggestions in this respect -- this area isn't set in stone yet.
I registered a J & Unix presentation/tutorial/workshop with the APL99
board where we will deal with all these and related issues. Come there
to get familiar with things like this and voice your wishes and
preferences.
Martin
--------------------------------------------------------------------------------
J Forum: for information about this list, see http://www.jsoftware.com/forum.htm
From owner-jsoftware@lists.interlog.com Fri Apr 2 13:33:03 1999
From: "Chris Burke" <cdburke@interlog.com>
Subject: Jforum: Eric Iverson at the NYPC C++ & Java SIG
Date: Fri, 2 Apr 1999 13:15:18 -0500
charset="Windows-1252"
Eric will be speaking on J at the New York PC C++ & Java SIG on Thursday
April 8th. Anyone can attend.
For more information: http://www.jsoftware.com/nypres.htm
--------------------------------------------------------------------------------
J Forum: for information about this list, see http://www.jsoftware.com/forum.htm
From owner-jsoftware@lists.interlog.com Fri Apr 2 21:11:07 1999
From: "Roger Hui" <RHui@interlog.com>
Subject: Jforum: Re: Divisors quickly
Date: Fri, 2 Apr 1999 20:40:49 -0500
charset="iso-8859-4"
odometer=: #: i.@(*/) NB. phrase book, chapter 7, m33
divisors=: /:~ @: ({. */ . ^"1 odometer@:>:@:{:) @: (__&q:)
My original solution
div1=: /:~ @: ((*/ . ^"1 odometer@:>:) / )@: (__&q:)
Using f/ to replace ({. g h@{:)
div2=: /:~ @: (*/ . ^"1 odometer@:>:) / @: (__&q:)
Removing redundant parentheses
div3=: /:~ @: , @: (*/&>) @: { @: ((^ i.@>:)&.>/) @: (__&q:)
A combination of Chris' use of the monad {
and of using __&q:
--------------------------------------------------------------------------------
J Forum: for information about this list, see http://www.jsoftware.com/forum.htm
From owner-jsoftware@lists.interlog.com Sat Apr 3 01:14:34 1999
Date: Fri, 02 Apr 1999 21:45:38 -0800
From: David Steinbrook <davidst@pacbell.net>
Subject: Jforum: wd 'ptop'
Content-type: text/plain; charset="iso-8859-1"
Once this command is invoked, is there a way to remove its effect other than
by closing the form itself?
Thanks, David
--------------------------------------------------------------------------------
J Forum: for information about this list, see http://www.jsoftware.com/forum.htm
From owner-jsoftware@lists.interlog.com Sat Apr 3 06:47:00 1999
From: "Chris Burke" <cdburke@interlog.com>
References: <000701be7d95$550c6f70$561baace@davids>
Subject: Re: Jforum: wd 'ptop'
Date: Sat, 3 Apr 1999 06:27:18 -0500
charset="iso-8859-1"
wd 'ptop 0' will switch it off.
> Once this command is invoked, is there a way to remove its effect other
than
> by closing the form itself?
>
> Thanks, David
--------------------------------------------------------------------------------
J Forum: for information about this list, see http://www.jsoftware.com/forum.htm
From owner-jsoftware@lists.interlog.com Sat Apr 3 10:51:22 1999
Date: Sat, 03 Apr 1999 10:30:41 -0500
From: David Ness <DNess@home.com>
Subject: Re: Jforum: Hint needed: Quad
References: <002e01be7cb9$e2940ce0$8e0014d1@f3nbp>
Sorry to be so dense, but I'm not sure what `11!:0 wd' commands
have to do with getting input. Don't they (1) just apply to
windows/mac versions of J; and (2) require that I set up some window(s)?
and; (3) isn't it a pretty elaborate solution to a pretty simple
problem?
I guess the real question is that since it is so easy to append a
`13 10{a.' to a `prompt' string if you want it, how come 1!:2&2
forces them to be there?
Roger Hui wrote:
>
> David Ness writes on Thursday, April 1:
>
> >Dr. Faller's recent question reminded me that I have never known
> >how to avoid the carriage return/linefeed that occurs after the
> >prompt when you do something like:
> > x =: 1!:1]1['Prompt Here' 1!:2]2
>
> There is no way, but the 11!:0 WD facility is a superior substitute.
>
>
--------------------------------------------------------------------------------
J Forum: for information about this list, see http://www.jsoftware.com/forum.htm
From owner-jsoftware@lists.interlog.com Sat Apr 3 15:12:27 1999
From: "Roger Hui" <RHui@interlog.com>
Subject: Jforum: Re: Hint needed: Quad
Date: Sat, 3 Apr 1999 14:44:43 -0500
charset="iso-8859-1"
David Ness writes on Saturday, April 3:
>Sorry to be so dense, but I'm not sure what `11!:0 wd' commands
>have to do with getting input. Don't they (1) just apply to
>windows/mac versions of J; and (2) require that I set up some window(s)?
>and; (3) isn't it a pretty elaborate solution to a pretty simple
>problem?
>
>I guess the real question is that since it is so easy to append a
>`13 10{a.' to a `prompt' string if you want it, how come 1!:2&2
>forces them to be there?
And why is the "pretty simple problem" worth solving? Why is it so
important that the prompt appear on the same line as the user input?
11!:0 wd is the facility offered for user interactions, in a way that
fits the Windows environment.
--------------------------------------------------------------------------------
J Forum: for information about this list, see http://www.jsoftware.com/forum.htm
From owner-jsoftware@lists.interlog.com Sun Apr 4 15:32:54 1999
Date: Sun, 04 Apr 1999 15:15:58 -0400
From: Eric Iverson <eiverson@interlog.com>
Organization: Iverson Software Inc.
Subject: Re: Jforum: Prompting on a line
References: <199904021730.TAA00858@ohura.gaertner.de>
You can get an APL quadprime style prompt and input on the same line
with the following trick. The smappend command is complicated with the
"s delimiting the text so that it can get the 0 parameter that supresses
deletion of trailing blanks. If you didn't need trailing blanks you
could use use 'smappend *',y.
NB. answer=. prompt 'enter something here: '
prompt=: 3 : 0
NB. append line to ijx window (keep trailing blanks)
wd'smappend "',y.,'" 0'
1!:1[1 NB. read from keyboard (ijx window)
)
--------------------------------------------------------------------------------
J Forum: for information about this list, see http://www.jsoftware.com/forum.htm
From owner-jsoftware@lists.interlog.com Tue Apr 6 05:13:54 1999
From: "Ondrus, Milan" <milan.ondrus@csfb.com>
Subject: Jforum: bug
Date: Tue, 6 Apr 1999 10:59:49 +0200
Content-Type: multipart/mixed;
boundary="----_=_NextPart_000_01BE800B.8A86ADDB"
This message is in MIME format. Since your mail reader does not understand
this format, some or all of this message may not be legible.
------_=_NextPart_000_01BE800B.8A86ADDB
<<error.txt>>
../
------_=_NextPart_000_01BE800B.8A86ADDB
name="error.txt"
Content-Transfer-Encoding: quoted-printable
Content-Disposition: attachment;
filename="error.txt"
Content-Location: ATT-0-29B6F403F8EBD211AFC80060B0688B0D-e
rror.txt
I ran into a very annoying problem that I somehow can't reproduce=20
easily.
The situation was the following.
I had a J session open that reads data from DB2.The DB2 origin
was quite big (about 30 mb).The mecano simply took 5 records out of =
DB 2
and appended them in a Jfile.
After a while,somewhere in the middle of the process -=20
I found an error in my J-session saying:
interface error j_write
(bl=F6k {. dat) 1!:3 h
First I thought it might have sth to do with the DB2 records but
now I'm no longer sure.
This J file has about 4600 components ,when I copied (jappend ..) =
parts of it
into an other J-file I ran into the same following problem:
a) copying cn 2000..2600 ended again in an interface error.
b) BUT copying 2000..2300 and then 2301..2600 worked.
The places where I got interface errors seems to be inpredictable.
One additional remark : sometimes if I get this interface error =
things do not
work correct: like jcreate 'filexy' gives me a "1" but jsize 'filexy' =
gives _1
although the file exists and the file 'filexy' has nothing to do with =
the DB2
file.
Once a jfile-access ended with an interface error - the file is =
corrupted
(a zip of the file terminates with an error).
I mean is it possible that during that process reading DB2 (via ODBC)
sth went wrong? using the sequence :
...
for_scanner. lp do.
r =3D. ddfet c =3D. sh,blkSize NB..which is 5
while. _1 -: r do. delay t
r =3D. ddfet c
end.
tempfile jappend~ <r
delay t
end.
end.
......=20
Ran anyone into a similar problem ?=20
I'm sorry that I can't be more precise about this but it looks like
I'm fighting invisible mills in that particular case.
/Milan Ondrus
=20
Ps. I'm using NT 4.0 and J 4.02a=20
=20
=20
------_=_NextPart_000_01BE800B.8A86ADDB--
--------------------------------------------------------------------------------
J Forum: for information about this list, see http://www.jsoftware.com/forum.htm
From owner-jsoftware@lists.interlog.com Tue Apr 6 10:49:37 1999
From: "John D. Baker" <bakerjd@kos.net>
Subject: Re: Jforum: bug
Date: Tue, 6 Apr 1999 09:27:11 -0400
charset="iso-8859-1"
Milan Wrote:
>I mean is it possible that during that process reading DB2 (via ODBC)
>sth went wrong? using the sequence :
>
>...
>for_scanner. lp do.
>r =. ddfet c =. sh,blkSize NB..which is 5
>while. _1 -: r do. delay t
> r =. ddfet c
>end.
>tempfile jappend~ <r
>delay t
>end.
>end.
Without knowing more about what you are trying to do
it's hard to help but here are some things to consider.
1) What ODBC interface are you using. The native 14!:x verbs
have some problems that crop up under NT. If you haven't
got it already try the beta ODBC script that calls the
Windows ODBC API with the J dll interface. This is available
on the beta scripts page of www.jsoftware.com.
Future releases of J will
remove the 14!:x interface and replace it with this system
so it's not a waste of time to study it now. The comments
in the code are fairly comphrehensive and give
lots of hints about buffer sizes et cetera.
2) Unless your DB2 records are VERY large using a blocksize
of 5 is not a good idea. You said the db was 30mb. What is
the the average table record size? If they are reasonable
try and put as many as you can in one block. I typically
use blocksizes of 500 to 5000 when I'm doing things like
this. This cuts down on the number of components in
the jfiles and will help out.
3) Do you even need jfiles? If your data is all fixed
length using memory mapped files will be easiar and faster.
Memory mapped files scale to millions of records without
problems. The J demo of this facility uses a million record table to
get started.
3) It's better to use the ddfch verb to get large amounts of
SQL data. The ddfch in the new interface uses column
binding which I have used to fetch 350,000 record tables
without blocking at all. (My machine has 196mb of memory)
I haven't experienced your jfile problems. Yesterday I was
testing with a 20,000 component jfile with no problems.
I don't doubt that once a jfile is corrupted it's toast. There is
NO protection from components overwriting each other
once the internal component boundries are broken. Yet
another reason to backup.
Hope this helps.
John D. Baker
bakerjd@kos.net
......
-----Original Message-----
From: Ondrus, Milan <milan.ondrus@csfb.com>
Date: Tuesday, April 06, 1999 5:03 AM
Subject: Jforum: bug
--------------------------------------------------------------------------------
J Forum: for information about this list, see http://www.jsoftware.com/forum.htm
From owner-jsoftware@lists.interlog.com Tue Apr 6 11:25:58 1999
From: "Chris Burke" <cdburke@interlog.com>
References: <000101be8037$c0c3b300$01ffde6f@papabear>
Subject: Re: Jforum: bug
Date: Tue, 6 Apr 1999 10:48:35 -0400
charset="iso-8859-1"
I think Milan's problem is related to little hard disk space remaining.
"interface error" means that some file foreign conjunction failed.
----- Original Message -----
From: John D. Baker <bakerjd@kos.net>
Sent: Tuesday, April 06, 1999 9:27 AM
Subject: Re: Jforum: bug
> Milan Wrote:
>
> >I mean is it possible that during that process reading DB2 (via ODBC)
> >sth went wrong? using the sequence :
--------------------------------------------------------------------------------
J Forum: for information about this list, see http://www.jsoftware.com/forum.htm
From owner-jsoftware@lists.interlog.com Tue Apr 6 23:01:17 1999
Date: 6 Apr 99 18:47:06 -0800
Subject: Jforum: local vs global (=. vs =:)
From: "Piet de Jong" <piet.dejong@commerce.ubc.ca>
X-Fontfamily: Geneva
X-Fontsize: 12
Content-Transfer-Encoding: quoted-printable
I just helped someone start J up on their Wintel machine, (I use a Mac
myself) and noticed that under J4.02 all definitions defined with =3D.
in the script window are not actually defined. (That is running the
script window after putting in the
=3D. definitions leads to them not accessible in the execution window.
This is a change from at least my Mac version 3.x. Why the change?
And what happens to these =3D. definitions after running the script
window?
--------------------------------------------------------------------------------
J Forum: for information about this list, see http://www.jsoftware.com/forum.htm
From owner-jsoftware@lists.interlog.com Tue Apr 6 23:37:09 1999
From: "John D. Baker" <bakerjd@kos.net>
Subject: Re: Jforum: local vs global (=. vs =:)
Date: Tue, 6 Apr 1999 23:19:02 -0400
charset="iso-8859-1"
Ahhh....
The window script is now processed with a "load"
verb. All local definitions =. run inside this verb and
are treated as local variables. This is actually
very handy because it let's scripts have
temps that don't clutter things up when you're
done with them.
Just change =. to =: if you want a global to persist.
John
-----Original Message-----
From: Piet de Jong <piet.dejong@commerce.ubc.ca>
Date: Tuesday, April 06, 1999 10:50 PM
Subject: Jforum: local vs global (=. vs =:)
>I just helped someone start J up on their Wintel machine, (I use a Mac myself)
and noticed that under J4.02 all definitions defined with =. in the script
window are not actually defined. (That is running the script window after
putting in the
>=. definitions leads to them not accessible in the execution window.
>
>This is a change from at least my Mac version 3.x. Why the change? And what
happens to these =. definitions after running the script window?
>-------------------------------------------------------------------------------
- J Forum: for information about this list, see
http://www.jsoftware.com/forum.htm
>
--------------------------------------------------------------------------------
J Forum: for information about this list, see http://www.jsoftware.com/forum.htm
From ljdickey Wed Apr 7 06:13:13 1999
Subject: Re: Jforum: unpacking waterloo archived files
Have you been able to get the files you wanted?
> From: mellemf@nimo.com
> To: forum@jsoftware.com
> Date: Tue, 30 Mar 1999 09:39:07 -0500
> Subject: Jforum: unpacking waterloo archived files
>
> Could the executables to unpack the files in the waterloo archives be placed in
> the archives, or references(conections) to the sites where they are; posted
> there? I am in the PC MS Windows & NT world.
>
>
>
> --------------------------------------------------------------------------------
> J Forum: for information about this list, see http://www.jsoftware.com/forum.htm
>
From mellemf@nimo.com Wed Apr 7 10:12:22 1999
From: mellemf@nimo.com
X-Lotus-FromDomain: NMPC
Date: Wed, 7 Apr 1999 10:09:15 -0400
Subject: Re: Jforum: unpacking waterloo archived files
Content-type: text/plain; charset=us-ascii
Content-Disposition: inline
My email went bust due to NOTES. My email should be repaired soon , I using a
work-arround right now.
I should be able to use PKUNZIP ( when I last tried it; it did not support as
many modes) . My thanks to all for their patient replies. As usual when I am
puzzled it helps to query our supportive forum.
From ljdickey Wed Apr 7 10:43:58 1999
Subject: Re: Jforum: unpacking waterloo archived files
Yes, PKUNZIP should do it. Good luck. Sorry to hear
the bad news about NOTES.
Lee
From owner-jsoftware@lists.interlog.com Thu Apr 8 03:50:42 1999
From: "Ondrus, Milan" <milan.ondrus@csfb.com>
Subject: Jforum: FW: bug
Date: Thu, 8 Apr 1999 09:16:09 +0200
Content-Type: multipart/mixed;
boundary="----_=_NextPart_000_01BE818F.63AFEB1C"
This message is in MIME format. Since your mail reader does not understand
this format, some or all of this message may not be legible.
------_=_NextPart_000_01BE818F.63AFEB1C
The clouds are gone.
I could prove that this behaviour was caused by a vanishing space on the
harddisk.
This was not evident as before I started to run the program there was
space enough - and it was not
the Jfile who filled up the harddisk (the Db2 original was hardly 30meg)
but it was a log file that has been created
during the download process (did not know about this) in a totally
different place on the harddisk and this logfile was much bigger
than the Db2 original ..( > 680meg)..
/Milan Ondrus
sorry for stirring up the things - but its at least a "nice to
know-feature".
> -----Original Message-----
> From: Ondrus, Milan
> Sent: Dienstag, 6. April 1999 11:00
> To: 'Jforum'
> Subject: bug
>
>
>
>
> <<error.txt>>
>
> ../
------_=_NextPart_000_01BE818F.63AFEB1C
name="error.txt"
Content-Transfer-Encoding: quoted-printable
Content-Disposition: attachment;
filename="error.txt"
Content-Location: ATT-0-0881ADBC43EDD211AFC80060B0688B0D-e
rror.txt
I ran into a very annoying problem that I somehow can't reproduce=20
easily.
The situation was the following.
I had a J session open that reads data from DB2.The DB2 origin
was quite big (about 30 mb).The mecano simply took 5 records out of =
DB 2
and appended them in a Jfile.
After a while,somewhere in the middle of the process -=20
I found an error in my J-session saying:
interface error j_write
(bl=F6k {. dat) 1!:3 h
First I thought it might have sth to do with the DB2 records but
now I'm no longer sure.
This J file has about 4600 components ,when I copied (jappend ..) =
parts of it
into an other J-file I ran into the same following problem:
a) copying cn 2000..2600 ended again in an interface error.
b) BUT copying 2000..2300 and then 2301..2600 worked.
The places where I got interface errors seems to be inpredictable.
One additional remark : sometimes if I get this interface error =
things do not
work correct: like jcreate 'filexy' gives me a "1" but jsize 'filexy' =
gives _1
although the file exists and the file 'filexy' has nothing to do with =
the DB2
file.
Once a jfile-access ended with an interface error - the file is =
corrupted
(a zip of the file terminates with an error).
I mean is it possible that during that process reading DB2 (via ODBC)
sth went wrong? using the sequence :
...
for_scanner. lp do.
r =3D. ddfet c =3D. sh,blkSize NB..which is 5
while. _1 -: r do. delay t
r =3D. ddfet c
end.
tempfile jappend~ <r
delay t
end.
end.
......=20
Ran anyone into a similar problem ?=20
I'm sorry that I can't be more precise about this but it looks like
I'm fighting invisible mills in that particular case.
/Milan Ondrus
=20
Ps. I'm using NT 4.0 and J 4.02a=20
=20
=20
------_=_NextPart_000_01BE818F.63AFEB1C--
--------------------------------------------------------------------------------
J Forum: for information about this list, see http://www.jsoftware.com/forum.htm
From owner-jsoftware@lists.interlog.com Thu Apr 8 03:55:38 1999
From: "qq" <gccinc@usa.net>
References: <B330072B-18CD318@137.82.66.177>
Subject: Jforum: Linux local installation
Date: Thu, 8 Apr 1999 03:38:31 -0400
charset="koi8-r"
I was unable to properly install/register J on Linux locally.
I am a Linux user (proper). And with growing popularity
of Linix there will more users like me who care little about
the intrinsics of the system as is, especially in corporate
environment.
Recently I went out for trying openAPL discussed in c.l.a.
And was stunned by a phrase "and now change to root with su".
I think it's only possible when you have Linux at home to play.
Or it's possible for Linux admins who very rarely are APL/J
programmers (except for distinguished few).
On the other hand one can try to convince the sys admin to install
the software globally, but
1. You need a site licence -- which is an issue by itself
2. Even in a very small company asking sys admin for something
is painstaking and fruitless.
But I was not able to make out how do I do a local installation
(in my home hierarchy). I tried a rule-of-thumb approach, but
was not able to register even a trial reg key, although the dynamic
program worked OK and paths were fine too, e.g.
> {. "1 ] 1!:0'*'
lib
bin
man
lib-402a
load'stats'
lsfit
{:@] %. {.@] ^/ i.@>:@[
I installed into $HOME/usr/local/, changing $PREFIX accordingly.
When running jreg it tried to write into $HOME/usr/local/lib/j-402a/...
'Unable to write to license file /home/olegyk/usr/local/lib-402a/j402.lic'
was the error. I manually created this folder where j402.lic was
placed but this did not help -- the result was
'Registration key set.'
but starting j produced the same countdown.
--------------------------------------------------------------------------------
J Forum: for information about this list, see http://www.jsoftware.com/forum.htm
From owner-jsoftware@lists.interlog.com Thu Apr 8 13:00:20 1999
From: "John D. Baker" <bakerjd@kos.net>
Subject: Re: Jforum: FW: bug
Date: Thu, 8 Apr 1999 12:37:52 -0400
charset="iso-8859-1"
Glad to hear your problems have been resolved.
Don't you think it's just plain rude of software to
deposit massive files without informing the user.
-----Original Message-----
From: Ondrus, Milan <milan.ondrus@csfb.com>
Date: Thursday, April 08, 1999 3:37 AM
Subject: Jforum: FW: bug
>
> The clouds are gone.
>
> I could prove that this behaviour was caused by a vanishing space on the
>harddisk.
> This was not evident as before I started to run the program there was
>space enough - and it was not
> the Jfile who filled up the harddisk (the Db2 original was hardly 30meg)
>but it was a log file that has been created
> during the download process (did not know about this) in a totally
>different place on the harddisk and this logfile was much bigger
> than the Db2 original ..( > 680meg)..
>
> /Milan Ondrus
> sorry for stirring up the things - but its at least a "nice to
>know-feature".
>
>
>> -----Original Message-----
>> From: Ondrus, Milan
>> Sent: Dienstag, 6. April 1999 11:00
>> To: 'Jforum'
>> Subject: bug
>>
>>
>>
>>
>> <<error.txt>>
>>
>> ../
>
--------------------------------------------------------------------------------
J Forum: for information about this list, see http://www.jsoftware.com/forum.htm
From owner-jsoftware@lists.interlog.com Thu Apr 8 15:59:25 1999
Date: Thu, 8 Apr 1999 21:36:54 +0200 (MET DST)
From: Martin Neitzel <neitzel@gaertner.de>
Subject: Re: Jforum: Linux local installation
[More or less full quote, Cc & Reply-To set to juggle@gaertner.de
and its archive because the question is specific to the Unix platforms.]
> I was unable to properly install/register J on Linux locally.
>
> I am a Linux user (proper). And with growing popularity
> of Linix there will more users like me who care little about
> the intrinsics of the system as is, especially in corporate
> environment.
>
> Recently I went out for trying openAPL discussed in c.l.a.
> And was stunned by a phrase "and now change to root with su".
> I think it's only possible when you have Linux at home to play.
>
> Or it's possible for Linux admins who very rarely are APL/J
> programmers (except for distinguished few).
>
> On the other hand one can try to convince the sys admin to install
> the software globally, but
>
> 1. You need a site licence -- which is an issue by itself
> 2. Even in a very small company asking sys admin for something
> is painstaking and fruitless.
>
> But I was not able to make out how do I do a local installation
> (in my home hierarchy). I tried a rule-of-thumb approach, but
> was not able to register even a trial reg key, although the dynamic
> program worked OK and paths were fine too, e.g.
Hello Oleg (?),
the J kits are designed to be installable without root privileges
in a location under control by an ordinary user.
BTW: Packages like RPMs (Red Hat Linux), dpkg (Debian Linux), pkg_add
(Free/NetBSD), pkgadd (SUN / SysV), installp (AIX), inst (SGI), and
netdist (HPUX) are fine for lazy admins. Unfortunately, they make it
even more difficult for non-admins to override the defaults, no matter
if lazy or not. This is one of the reasons why I chose to provide
simple tar kits as the packaging form.
> I installed into $HOME/usr/local/, changing $PREFIX accordingly.
This setting is a bit, um, strange. For a "private installation",
you'd set
PREFIX = $HOME
and things would go into $HOME/bin, $HOME/lib, and $HOME/man/man1
then. Hanging off two intermediate directories "usr/local" is
really not necessary.
> [...sample J trace omitted...]
>
> When running jreg it tried to write into $HOME/usr/local/lib/j-402a/...
That path is good.
> 'Unable to write to license file /home/olegyk/usr/local/lib-402a/j402.lic'
> was the error.
The "lib-402a" is in error. This was a known bug in jreg up to and including
j402.a. It has been fixed for j402b.
Which leads me to the following hint: Please check out
http://juggle.gaertner.de/pr/ how to report J/Unix problems
and the list of bugs and fixes already documented there:
http://juggle.gaertner.de/pr/pr-list.cgi
In your case, the Problem Report 16, titled "jreg refers to wrong
library directory" (http://juggle.gaertner.de/pr/full-report.cgi/16),
would have given you an immediate answer. (I refuse to label the PR
section at juggle as "Knowledge Base", though---it's a "Bug Base", period. ;-)
> I manually created this folder where j402.lic was
> placed but this did not help -- the result was
>
> 'Registration key set.'
>
> but starting j produced the same countdown.
No, j402.lic goes into $HOME/usr/local/lib/j-402a/
Said Problem Report 16 explains how to fix your $HOME/usr/local/bin/jreg.
Martin
PS: j402b has been made three weeks ago:
-rw-r--r-- 1 neitzel jsource 466245 Mar 20 05:50 j402b-Linux.tgz
It incorporates above fix, a small change to 3!:1, and a new license key.
These changes do not warrant an upgrade from 402a to 402b, BUT:
above package is based on glibc aka libc6. I will put it into the
usual distribution channel shortly. Again: unless you are dependent
on glibc, there's no reason to upgrade to patch level "b".
--------------------------------------------------------------------------------
J Forum: for information about this list, see http://www.jsoftware.com/forum.htm
From owner-jsoftware@lists.interlog.com Thu Apr 8 19:11:35 1999
From: Charles Fisk <charles.fisk@usaa.com>
Subject: Jforum: Capturing data from mainframe host directly int J using FTP
Date: Thu, 8 Apr 1999 17:51:23 -0500
I am building an application on the PC that captures data from the
mainframe, does calculations on the data and
displays graphs of the results.
I built a prototype in APL 2 that
gathers data from an MVS PDS. The way I am doing this is
- by using AP100 which invokes the FTP command prompt in our NT systems.
- The FTP command gets the data and puts it in a temporary directory.
- APL2 then reads the data into the workspace.
This is a bit cumbersome as it involves
filing the ftp commands,
executing the file with the ftp commands which gets the mainframe data and
files it on the PC
reading in the individual files into the APL workspace.
I was wondering if I could do a better job with J.
A sample of the FTP commands created by APL2 on the PC is shown here:
open jesl900
userID
password
cd 'PLP.ACT.AUCMI.USAA'
get FL001 c:\temp\FL001.prn // get files from the PDS and put them on
the PC
get FL010 c:\temp\FL010.prn
get FL018 c:\temp\FL018.prn
get FL020 c:\temp\FL020.prn
bye
The commands are then placed in a file and executed by sending the following
APL AP100 command:
'ftp -s:C:\temp\gather.bat>c:\temp\response.txt' // execute ftp to get
data. PIPE ftp reply to response.txt
--------------------------------------------------------------------------------
J Forum: for information about this list, see http://www.jsoftware.com/forum.htm
From owner-jsoftware@lists.interlog.com Thu Apr 8 21:11:27 1999
From: "Chris Burke" <cdburke@interlog.com>
Subject: Jforum: Martin Neitzel at the APL Club Germany
Date: Thu, 8 Apr 1999 20:54:19 -0400
charset="iso-8859-1"
Martin will be speaking on J at the APL Club Germany on Friday April
16th.
For more information: http://www.jsoftware.com/gaplpres.htm
--------------------------------------------------------------------------------
J Forum: for information about this list, see http://www.jsoftware.com/forum.htm
From owner-jsoftware@lists.interlog.com Fri Apr 9 16:13:38 1999
Date: Fri, 9 Apr 1999 15:47:03 -0400 (EDT)
From: Oleg Kobchenko <gccinc@usa.net>
Subject: Re: Jforum: Linux local installation
In-Reply-To: <199904081936.VAA18717@ohura.gaertner.de>
Thank you for the directions for local install.
It functions all right, even the man pages, at
$HOME/bin, $HOME/lib, and $HOME/man/man1
However, there are some questions:
Different from J Win32, though very (more) logical:
1!:40'' NB. Shows current path, whereas JWin32,
The directory path of the J application executable file.
/home/olegyk/
1!:42'' NB. This coinsides with Win32
/home/olegyk/lib/j-402a/
5{.>{."1]1!:0'.*' NB. shows current path, where J was started
. whereas JWin32 shows same as 1!:42
..
.bashrc
.Xauthority
.bash_profile
Is there a way to change working directory for 1!:40
and 1!:0 etc.? Look (I started in /home/olegyk/wget):
2!:0'cd ; pwd'
/home/olegyk
2!:0'pwd'
/home/olegyk/wget
I am a little confused about where the lib directory
should reside and where the .lic file should go,
here are exerpts from different places and they almost
all seem to differ.
Roadmap:
/usr/local/lib/j-3.05c <-- J script libraries
/usr/local/lib/j-4.02a
Your mail to JForum:
$HOME/usr/local/lib/j-402a/
In my folders actually created:
$HOME/lib/j-402a/
16th "Bug-Base":
As a temporary fix, one can simply change
the last line in e.g. /usr/local/bin/jreg
from: /usr/local/bin/jregist /usr/local/lib $*
to: /usr/local/bin/jregist /usr/local/lib-401d $*
The actual jreg script:
/home/olegyk/bin/jregist /home/olegyk/lib-402a $*
So what I did was I manually changed:
from /home/olegyk/bin/jregist /home/olegyk/lib-402a $*
to /home/olegyk/bin/jregist /home/olegyk/lib/j-402a $*
and it worked.
And finally,
PS: j402b has been made three weeks ago:
-rw-r--r-- 1 neitzel jsource 466245 Mar 20 05:50 j402b-Linux.tgz
what's the FTP (?) path.
--------------------------------------------------------------------------------
J Forum: for information about this list, see http://www.jsoftware.com/forum.htm
From owner-jsoftware@lists.interlog.com Sat Apr 10 23:25:49 1999
Date: Sat, 10 Apr 1999 19:44:36 -0500
From: Peter Nichols <pnichols@airmail.net>
Subject: Jforum: Sizing Child Windows
I have a main window that has a child window accessed from a menu. When
the child window opens I want it to be the same size and location as the
parent window. This is so that the parent window will not be visible
when the child window is open. How do I accomplish this?
Thanks,
Peter
--------------------------------------------------------------------------------
J Forum: for information about this list, see http://www.jsoftware.com/forum.htm
From owner-jsoftware@lists.interlog.com Sun Apr 11 05:42:59 1999
From: "Chris Burke" <cdburke@interlog.com>
References: <370FF074.646B@airmail.net>
Subject: Re: Jforum: Sizing Child Windows
Date: Sun, 11 Apr 1999 05:33:14 -0400
charset="iso-8859-1"
In the parent form, use
pos=: wd 'qformx'
to find the position and size of the form.
After creating the child form, and before showing it, use
wd 'pmovex',":pos
to move it to pos.
----- Original Message -----
From: Peter Nichols <pnichols@airmail.net>
Sent: Saturday, April 10, 1999 8:44 PM
Subject: Jforum: Sizing Child Windows
> I have a main window that has a child window accessed from a menu.
When
> the child window opens I want it to be the same size and location as
the
> parent window. This is so that the parent window will not be
visible
> when the child window is open. How do I accomplish this?
--------------------------------------------------------------------------------
J Forum: for information about this list, see http://www.jsoftware.com/forum.htm
From owner-jsoftware@lists.interlog.com Sun Apr 11 21:50:58 1999
Date: Sun, 11 Apr 1999 20:39:25 -0500
From: Peter Nichols <pnichols@airmail.net>
Subject: Jforum: JFiles
With JFiles is there a way to drop components?
Thanks,
Peter
--------------------------------------------------------------------------------
J Forum: for information about this list, see http://www.jsoftware.com/forum.htm
From owner-jsoftware@lists.interlog.com Mon Apr 12 00:16:03 1999
Date: Sun, 11 Apr 1999 23:49:17 -0400
From: Eric Iverson <eiverson@interlog.com>
Organization: Iverson Software Inc.
Subject: Re: Jforum: Sizing Child Windows
References: <370FF074.646B@airmail.net>
Peter Nichols wrote:
> I have a main window that has a child window accessed from a menu. When
> the child window opens I want it to be the same size and location as the
> parent window. This is so that the parent window will not be visible
> when the child window is open. How do I accomplish this?
Making the second form the same size simply to hide the first is
unusual. Usually forms should be whatever size they need to be. Perhaps
you should use commands to hide the initial form (pshow). If the new
form is owned by the first, then you should use the parent owner
property so that the original form is disabled (but still visible).
--------------------------------------------------------------------------------
J Forum: for information about this list, see http://www.jsoftware.com/forum.htm
From owner-jsoftware@lists.interlog.com Mon Apr 12 07:45:08 1999
From: "Chris Burke" <cdburke@interlog.com>
References: <37114ECD.5E26@airmail.net>
Subject: Re: Jforum: JFiles
Date: Mon, 12 Apr 1999 07:06:28 -0400
charset="iso-8859-1"
Not as such. However, you can jdup the file with only the components
you want to keep.
----- Original Message -----
From: Peter Nichols <pnichols@airmail.net>
Sent: Sunday, April 11, 1999 9:39 PM
Subject: Jforum: JFiles
> With JFiles is there a way to drop components?
--------------------------------------------------------------------------------
J Forum: for information about this list, see http://www.jsoftware.com/forum.htm
From owner-jsoftware@lists.interlog.com Mon Apr 12 09:35:00 1999
Date: Mon, 12 Apr 1999 08:55:27 -0400
From: Cliff Reiter <reiterc@lafvax.lafayette.edu>
Subject: Jforum: J in math classroom
Organization: Lafayette College
Content-type: text/plain; charset=us-ascii
References: <37114ECD.5E26@airmail.net>
<002d01be84d7$e97756e0$0c0114d1@Pcdburke.lafayette.edu>
Dear J forum:
I thought I'ld make you aware of a J based
workshop for this summer in case you have interest
or want to twist a friends arm. I expect this to
be a fun time of discussion and work on using
J to teach mathematics. Messiah College has a
beautiful pastorial campus.
Txt version of the announement is below.
I can provide a formatted version document upon
request.
Note that Marvin Brubaker is the contact in
order to get more information or to express
interest.
Best,
Cliff
- ***************************************************************
The Eastern Pennsylvania & Delaware Section
of the
Mathematical Association of America
1999 Summer Workshop
J in the Math Classroom:
Visualization, Number Theory and Linear Algebra
at
Messiah College
Grantham, PA 17027
June 14-18, 1999
Presented by Clifford A. Reiter
Professor at Lafayette College
This workshop is designed for mathematics faculty to explore using J in
the mathematics classroom. J is high-level computer language with a
mathematical bent. No previous experience with J is expected. Cliff
Reiter will give an introduction to the language and offer illustrations
from his classroom use of J in teaching mathematical visualization,
linear algebra and number theory. Participants will explore uses of J of
their choice. Selected topics from those fields could include such
things as image processing, fractals, searches for empirical evidence of
number theory conjectures, empirical discovery of quadratic reciprocity,
implementing error correcting codes and exploring eigenvalues.
Participants are expected to share their own experiences and brainstorm
with other participants about what makes a valuable computer based
mathematics laboratory experiment. Time for developing some of those
ideas into classroom experiments using J is planned.
The daily schedule will be 9:00 to 12:00 noon and 1:30 to 4:00 p.m.
Cliff Reiter is an Associate Professor of Mathematics at Lafayette
College where he was the site director for several years of an NSF
Research Experiences for the undergraduates program. He is the author of
the text Fractals, Visualization and J and has used APL, Mathematica and
J in his teaching for many years. In addition to teaching with J in
several of his classes, he has been supervising J based undergraduate
research projects investigating chaotic dynamics with symmetry in recent
years.
Information about registration may be obtained by contacting:
Marvin L. Brubaker
Messiah College
Grantham, PA 17027
Phone: 717-766-2511, x7283
Email: mbrubake@messiah.edu
- ******************************************************************
--
Clifford A. Reiter
Mathematics Department, Lafayette College
Easton, PA 18042 USA, 610-330-5277
http://www.lafayette.edu/~reiterc
--------------------------------------------------------------------------------
J Forum: for information about this list, see http://www.jsoftware.com/forum.htm
From owner-jsoftware@lists.interlog.com Tue Apr 13 04:07:04 1999
Date: Tue, 13 Apr 1999 10:38:11 +0300
From: Anssi Seppala <anssi.seppala@enease.fi>
Subject: Jforum: Time series formats in J?
In-Reply-To: <3711ED3F.114E@lafvax.lafayette.edu>
References: <37114ECD.5E26@airmail.net>
<002d01be84d7$e97756e0$0c0114d1@Pcdburke.lafayette.edu>
Content-Transfer-Encoding: 8bit
X-MIME-Autoconverted: from quoted-printable to 8bit by plus.interlog.com id DAA15103
Hi everybody,
My applications manage time dependent data, mostly hourly or half-hourly.
Usually I handle it in n x 24 matrix format or n x 48 format. After doing
this many times before I try again find the optimal paradigm to represent
and store such time series so that there is always a short way to summarize
on hourly, daily, weekly level and do calculations between time series.
Is there anybody who has been playing with hourly or half-hourly time
series data in J? What J tools work best here? What functions are the most
suitable?
Thanks,
Anssi
-------------------------------------------------------------------
Anssi Sepp�l�
Consultant, electricity market and distribution: Applications and software
e-mail: Anssi.Seppala@enease.fi tel: +358 9 45400550
Enease Oy fax: +358 9 45400551
Mannerheimintie 66 A mob: +358 400 760077
FIN - 00260 HELSINKI, Finland http://www.kolumbus.fi/enease/
--------------------------------------------------------------------------------
J Forum: for information about this list, see http://www.jsoftware.com/forum.htm
From owner-jsoftware@lists.interlog.com Tue Apr 13 11:50:08 1999
Date: Tue, 13 Apr 1999 16:41:43 +0100
From: Ewart Shaw <strgh@warwick.ac.uk>
Organization: Statistics Dept, University of Warwick
Subject: Jforum: numbers bug?
I'm finally able to use J4, which is great. There seems to be a problem
mixing the notations *x* and *j*. For example:
1x1 1j1 1p1 NB. OK
1x1 1p1 NB. OK
1p1 1j1 NB. OK
1x1,1j1 NB. OK
1x1 1j1 NB. gives "ill-formed number"
[This also happened in J3]
Regards, Ewart Shaw
--------------------------------------------------------------------------------
J Forum: for information about this list, see http://www.jsoftware.com/forum.htm
From owner-jsoftware@lists.interlog.com Tue Apr 13 12:11:39 1999
From: "Roger Hui" <RHui@interlog.com>
Subject: Jforum: Re: Time series formats in J?
Date: Tue, 13 Apr 1999 08:08:25 -0400
charset="iso-8859-1"
Anssi Seppala writes on Tuesday, April 13:
>My applications manage time dependent data, mostly hourly or half-hourly.
>Usually I handle it in n x 24 matrix format or n x 48 format. After doing
>this many times before I try again find the optimal paradigm to represent
>and store such time series so that there is always a short way to summarize
>on hourly, daily, weekly level and do calculations between time series.
>
>Is there anybody who has been playing with hourly or half-hourly time
>series data in J? What J tools work best here? What functions are the most
>suitable?
I would store the data in inverted form (by column), with one of the columns
being a time column. The time column would be day numbers and minutes
i.e. (day number**/24 60)+minutes (or seconds, if the resolution is finer).
If there is a lot of data I would use mapped files, one file per column.
Summarizing is applying a function f that classifies the time (those that are
the same day, same hour, same minute, etc.) and then using key (/.) to
add up the data that belong to the same class.
(f time) +//. data
The scheme readily accommodates calculations between series, even
when their resolutions are different. It also readily accommdates
non-periodic series.
--------------------------------------------------------------------------------
J Forum: for information about this list, see http://www.jsoftware.com/forum.htm
From owner-jsoftware@lists.interlog.com Tue Apr 13 21:11:22 1999
Date: Wed, 14 Apr 1999 02:52:17 +0200
From: "d.alis" <dalis@balcab.ch>
Subject: Jforum: Mapped files (was: Re: Time series formats in J?)
References: <003c01be85a6$59729760$e8b0ba89@f3nbp>
Roger Hui wrote:
> .....
> If there is a lot of data I would use mapped files, one file per column.
Could someone be kind enough to give a couple of examples how this is done?
Thankyou
David Alis
--------------------------------------------------------------------------------
J Forum: for information about this list, see http://www.jsoftware.com/forum.htm
From owner-jsoftware@lists.interlog.com Tue Apr 13 23:23:27 1999
Date: Tue, 13 Apr 1999 23:09:09 -0400
From: Eric Iverson <eiverson@interlog.com>
Organization: Iverson Software Inc.
Subject: Re: Jforum: Mapped files (was: Re: Time series formats in J?)
References: <003c01be85a6$59729760$e8b0ba89@f3nbp> <3713E6C0.69A94A7F@balcab.ch>
David Alis wrote:
> Could someone be kind enough to give a couple of examples how this is done?
If you haven't done so already, you could take a look at the two labs on
mapped files. Or is your question more about the use of classification?
--------------------------------------------------------------------------------
J Forum: for information about this list, see http://www.jsoftware.com/forum.htm
From owner-jsoftware@lists.interlog.com Tue Apr 13 23:50:00 1999
From: "Roger Hui" <RHui@interlog.com>
Subject: Jforum: Re: numbers bug?
Date: Tue, 13 Apr 1999 20:36:23 -0400
charset="iso-8859-1"
Ewart Shaw writes on Tueday, April 13:
>I'm finally able to use J4, which is great. There seems to be a problem
>
>mixing the notations *x* and *j*. For example:
> 1x1 1j1 1p1 NB. OK
> 1x1 1p1 NB. OK
> 1p1 1j1 NB. OK
> 1x1,1j1 NB. OK
> 1x1 1j1 NB. gives "ill-formed number"
>[This also happened in J3]
I have now found and fixed the source of this problem.
Thanks for reporting it.
--------------------------------------------------------------------------------
J Forum: for information about this list, see http://www.jsoftware.com/forum.htm
From owner-jsoftware@lists.interlog.com Wed Apr 14 07:58:29 1999
Date: Wed, 14 Apr 1999 13:42:30 +0200
From: "d.alis" <dalis@balcab.ch>
Subject: Re: Jforum: Mapped files (was: Re: Time series formats in J?)
References: <003c01be85a6$59729760$e8b0ba89@f3nbp> <3713E6C0.69A94A7F@balcab.ch> <371406D5.3E54@interlog.com>
Eric Iverson wrote:
> David Alis wrote:
> > Could someone be kind enough to give a couple of examples how this is done?
>
> If you haven't done so already, you could take a look at the two labs on
> mapped files. Or is your question more about the use of classification?
(Roger wrote....)
> I would store the data in inverted form (by column), with one of the columns
> being a time column. The time column would be day numbers and minutes
> i.e. (day number**/24 60)+minutes (or seconds, if the resolution is finer).
> If there is a lot of data I would use mapped files, one file per column.
>
> Summarizing is applying a function f that classifies the time (those that are
> the same day, same hour, same minute, etc.) and then using key (/.) to
> add up the data that belong to the same class.
>
> (f time) +//. data
I interpreted what Roger wrote to mean that the noun 'data', could be mapped onto a
number of different files. I dont see where this is done in either of the two labs.
The mapped file database lab treats table columns as. mapped-files that are
independent of each other.
In the mapped file database lab the nouns 'lic' and 'make' are two mapped files ,
but the expression
data=. lic,.make
doesn't seem to generate 'data' as a mapped file.
What should I do to create data so that it maps to the two files 'lic.jmf' and
'make.jmf' ?
Is this actually possible?
--------------------------------------------------------------------------------
J Forum: for information about this list, see http://www.jsoftware.com/forum.htm
From owner-jsoftware@lists.interlog.com Wed Apr 14 08:05:55 1999
From: "Bjorn G. Helgason" <gosi@centrum.is>
Subject: RE: Jforum: J in math classroom
Date: Tue, 13 Apr 1999 20:30:11 +-100
Content-Type: multipart/mixed; boundary="---- =_NextPart_000_01BE866F.24A5F3A0"
------ =_NextPart_000_01BE866F.24A5F3A0
From: Cliff Reiter[SMTP:reiterc@lafvax.lafayette.edu]
> J in the Math Classroom: Visualization, Number Theory and Linear Algebra
> Messiah College Grantham, PA 17027
> June 14-18, 1999
> Presented by Clifford A. Reiter Professor at Lafayette College
Very interesting
You should send it to c.l.a as well as some math news groups if you have not
already done so.
/Gosi
------ =_NextPart_000_01BE866F.24A5F3A0
Content-Type: application/ms-tnef
Content-Transfer-Encoding: base64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------ =_NextPart_000_01BE866F.24A5F3A0--
--------------------------------------------------------------------------------
J Forum: for information about this list, see http://www.jsoftware.com/forum.htm
From owner-jsoftware@lists.interlog.com Wed Apr 14 08:50:52 1999
Date: Wed, 14 Apr 1999 08:34:58 -0400
From: Cliff Reiter <reiterc@lafvax.lafayette.edu>
Subject: Re: Jforum: J in math classroom
Organization: Lafayette College
Content-type: text/plain; charset=us-ascii
References: <01BE866F.23C29880@ls-27.skima.is>
Actually Gosi,
Thanks for
I can read, but not post to usenet :). Simply a matter
that usenet is a very low priority on a campus with very
few unix boxes.
You are welcome to repost it there on my behalf unless you
know of other ways I can get/send usenet. Any free usenet
servers? I know about free web-pages/e-mail, but haven't
seen free usenet.
Best,
Cliff
PS Sorry for the broadcast - but I suspect 1 or 2 others
wondered the same thing.
> >J in the Math Classroom: Visualization, Number Theory and Linear Algebra
> > Messiah College Grantham, PA 17027
> > June 14-18, 1999
> > Presented by Clifford A. Reiter Professor at Lafayette College
>
> Very interesting
>
> You should send it to c.l.a as well as some math news groups if you have not
> already done so.
>
> /Gosi
--
Clifford A. Reiter
Mathematics Department, Lafayette College
Easton, PA 18042 USA, 610-330-5277
http://www.lafayette.edu/~reiterc
--------------------------------------------------------------------------------
J Forum: for information about this list, see http://www.jsoftware.com/forum.htm
From owner-jsoftware@lists.interlog.com Wed Apr 14 10:02:49 1999
Date: Wed, 14 Apr 1999 09:38:10 -0400
From: Eric Iverson <eiverson@interlog.com>
Organization: Iverson Software Inc.
Subject: Re: Jforum: Mapped files (was: Re: Time series formats in J?)
References: <003c01be85a6$59729760$e8b0ba89@f3nbp> <3713E6C0.69A94A7F@balcab.ch> <371406D5.3E54@interlog.com> <37147F25.ABE5FACD@balcab.ch>
d.alis wrote:
> I interpreted what Roger wrote to mean that the noun 'data', could be mapped onto a
> number of different files. I dont see where this is done in either of the two labs.
> The mapped file database lab treats table columns as. mapped-files that are
> independent of each other.
>
> In the mapped file database lab the nouns 'lic' and 'make' are two mapped files ,
> but the expression
> data=. lic,.make
> doesn't seem to generate 'data' as a mapped file.
> What should I do to create data so that it maps to the two files 'lic.jmf' and
> 'make.jmf' ?
> Is this actually possible?
I don't think Roger was suggesting that you combine lic and make into a
single noun data. Just directly access each column of data as its own
noun.
Mapping allows a J noun to map to a file or part of a file. It allows
several nouns to map to the same file (same, different, or overlapping
parts). But it does not allow a noun to map to several files the way you
suggest. I don't think this is necessary or desirable for the kind of
data base application we are talking about.
--------------------------------------------------------------------------------
J Forum: for information about this list, see http://www.jsoftware.com/forum.htm
From owner-jsoftware@lists.interlog.com Wed Apr 14 16:05:12 1999
Date: Wed, 14 Apr 1999 15:52:47 -0400
From: Kirk B Iverson <kbi@interlog.com>
Subject: Re: Jforum: J in math classroom
In-Reply-To: <37148B72.5574@lafvax.lafayette.edu>
References: <01BE866F.23C29880@ls-27.skima.is>
At 08:34 1999-04-14 -0400, Cliff wrote:
[...]
> You are welcome to repost it there on my behalf unless you
>know of other ways I can get/send usenet. Any free usenet
>servers? I know about free web-pages/e-mail, but haven't
>seen free usenet.
You can use www.dejanews.com for free web-based Usenet access.
A list of some alternatives can be found at:
http://dir.yahoo.com/Computers_and_Internet/Internet/Usenet/Public_Access_Usenet_Sites/
/K
--------------------------------------------------------------------------------
J Forum: for information about this list, see http://www.jsoftware.com/forum.htm
From owner-jsoftware@lists.interlog.com Wed Apr 14 17:23:02 1999
From: mellemf@nimo.com
X-Lotus-FromDomain: NMPC
Date: Wed, 14 Apr 1999 15:29:10 -0400
Subject: Jforum: I am missing a thought here?
Content-type: text/plain; charset=us-ascii
Content-Disposition: inline
av=: +/%#
a =: 2 3 4 5
av a .NB ok
3.5
+/%# 2 3 4 5 .NB not OK (Why?)
0.25
+/%# (2 3 4 5) .NB not OK
0.25
(+/%#) 2 3 4 5 .NB ok
3.5
--------------------------------------------------------------------------------
J Forum: for information about this list, see http://www.jsoftware.com/forum.htm
From owner-jsoftware@lists.interlog.com Wed Apr 14 18:11:42 1999
Date: Wed, 14 Apr 1999 17:59:05 -0400
From: David Ness <DNess@home.com>
Subject: Re: Jforum: I am missing a thought here?
References: <85256753.006BB313.00@mail.nimo.com>
Yes. +/%# 2 3 4 5 is read as +/(%(#(2 3 4 5))) or
+/(%(4)) or +/ 0.25 or 0.25
Remember: Operations are _right to left_ and applied in order unless
parenthesis (or definition) intervenes
To get the `trident effect' you want to (+/%#) or use a definition...
mellemf@nimo.com wrote:
>
> av=: +/%#
> a =: 2 3 4 5
> av a .NB ok
> 3.5
> +/%# 2 3 4 5 .NB not OK (Why?)
> 0.25
> +/%# (2 3 4 5) .NB not OK
> 0.25
> (+/%#) 2 3 4 5 .NB ok
> 3.5
>
> --------------------------------------------------------------------------------
> J Forum: for information about this list, see http://www.jsoftware.com/forum.htm
--------------------------------------------------------------------------------
J Forum: for information about this list, see http://www.jsoftware.com/forum.htm
From owner-jsoftware@lists.interlog.com Wed Apr 14 18:14:36 1999
From: "Chris Burke" <cdburke@interlog.com>
References: <85256753.006BB313.00@mail.nimo.com>
Subject: Re: Jforum: I am missing a thought here?
Date: Wed, 14 Apr 1999 18:03:11 -0400
charset="iso-8859-1"
Substituting definitions for names in a sentence may change the order
of execution. Use parens to ensure you get the same order of execution
as before. Of course this is true in any language, not just J.
Consider:
a=. 5 - 3
a + 4
6
5 - 3 + 4
_2
(5 - 3) + 4
6
> av=: +/%#
> a =: 2 3 4 5
> av a .NB ok
> 3.5
> +/%# 2 3 4 5 .NB not OK (Why?)
> 0.25
> +/%# (2 3 4 5) .NB not OK
> 0.25
> (+/%#) 2 3 4 5 .NB ok
> 3.5
--------------------------------------------------------------------------------
J Forum: for information about this list, see http://www.jsoftware.com/forum.htm
From ljdickey Thu Apr 15 06:32:18 1999
Subject: Re: Jforum: I am missing a thought here?
See the section about "forks".
From ljdickey Thu Apr 15 06:41:55 1999
Subject: Re: Jforum: I am missing a thought here?
The notation
f g h a
means
f ( g ( h (a) ) )
and we read it as
f of g of h of a
You are wanting to apply the combination
f g h
as a single unit (a fork) to the data so that it is treated as
(f a) g ( h a )
From owner-jsoftware@lists.interlog.com Thu Apr 15 09:30:11 1999
From: "Bjorn G. Helgason" <gosi@centrum.is>
Subject: RE: Jforum: J in math classroom
Date: Thu, 15 Apr 1999 13:11:04 +-100
Content-Type: multipart/mixed; boundary="---- =_NextPart_000_01BE8741.CB1DBBA0"
------ =_NextPart_000_01BE8741.CB1DBBA0
From: Cliff Reiter[SMTP:reiterc@lafvax.lafayette.edu]
> You are welcome to repost it there on my behalf
Thanks !
I will. I think it is important to get the message out that there are quality
courses available.
/Gosi
------ =_NextPart_000_01BE8741.CB1DBBA0
Content-Type: application/ms-tnef
Content-Transfer-Encoding: base64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=
------ =_NextPart_000_01BE8741.CB1DBBA0--
--------------------------------------------------------------------------------
J Forum: for information about this list, see http://www.jsoftware.com/forum.htm
From owner-jsoftware@lists.interlog.com Thu Apr 15 17:37:49 1999
X-Lotus-FromDomain: JPMORGAN@SMTP
From: "Jacob Moskowitz" <moskowitz_jacob@jpmorgan.com>
Date: Thu, 15 Apr 1999 17:17:25 -0400
Subject: Jforum: Plot Animation ?
Content-type: text/plain; charset=us-ascii
Content-Disposition: inline
I was hoping to generate animated plots with J, however
it seems that "frames" are skipped, producing a jerky halting result:
See the attached example.
Any assistance will be greatly appreciated.
I am running NT 4.0 SP4
Thanks
- Jake Moskowitz
JP Morgan
moskowitz_jacob@jpmorgan.com
__________________________________________________
Nx =: 300
dt =: 0.03
load 'trig'
load 'D:\apps\J\SYSTEM\CLASSES\PLOT\PLOT.IJS'
ix =: i.Nx
x =: 0.01*ix
NB. Attempt to show a travelling sine wave.
MovePlot =: 3 : 0
t =. 0.0
Nt =. <. ( y. % dt )
M =. sin (10*x)
NB. Fill array with sine wave (vector) at successive time increments:
while. ( t < y. ) do.
t =. t + dt
M =. ( M, sin ( (10*x) + t ) )
end.
M =. ( Nt,Nx ) $ M
NB. Now spit it out:
it =. 0
while. ( it < Nt ) do.
plot it{M
it =. it + 1
end.
)
MovePlot 5.0
This communication is for informational purposes only. It is not intended as
an offer or solicitation for the purchase or sale of any financial instrument
or as an official confirmation of any transaction, unless specifically agreed
otherwise. All market prices, data and other information are not warranted as
to completeness or accuracy and are subject to change without notice. Any
comments or statements made herein do not necessarily reflect those of
J.P. Morgan & Co. Incorporated, its subsidiaries and affiliates.
--------------------------------------------------------------------------------
J Forum: for information about this list, see http://www.jsoftware.com/forum.htm
From owner-jsoftware@lists.interlog.com Thu Apr 15 18:11:58 1999
From: "Chris Burke" <cdburke@interlog.com>
References: <85256754.00740E6B.00@nyc-ntgw-n01.ny.jpmorgan.com>
Subject: Re: Jforum: Plot Animation ?
Date: Thu, 15 Apr 1999 18:02:24 -0400
charset="iso-8859-1"
> I was hoping to generate animated plots with J, however
> it seems that "frames" are skipped, producing a jerky halting
result:
> See the attached example.
>
> Any assistance will be greatly appreciated.
>
> I am running NT 4.0 SP4
>
> Thanks
>
> - Jake Moskowitz
By default, J paints an isigraph window by simply marking it for
painting when the system is otherwise idle. If J continues execution
for a while, then the painting will be delayed.
To override this, issue a glpaint'' command after you do the plot,
i.e.
...
while. ( it < Nt ) do.
plot it{M
glpaint''
--------------------------------------------------------------------------------
J Forum: for information about this list, see http://www.jsoftware.com/forum.htm
From owner-jsoftware@lists.interlog.com Fri Apr 16 03:57:35 1999
Date: Fri, 16 Apr 1999 03:40:06 -0400
From: Kirk B Iverson <kbi@interlog.com>
Subject: Re: Jforum: Capturing data from mainframe host directly int J
using FTP
In-Reply-To: <FAAFBBC5B8FED011976340005003670603432A03@exchange02.usaa00
1.com>
At 17:51 1999-04-08 -0500, Charles Fisk <charles.fisk@usaa.com> wrote:
>I am building an application on the PC that captures data from the
>mainframe, does calculations on the data and
>displays graphs of the results.
>
>I built a prototype in APL 2 that
>gathers data from an MVS PDS. The way I am doing this is
>- by using AP100 which invokes the FTP command prompt in our NT systems.
>- The FTP command gets the data and puts it in a temporary directory.
>- APL2 then reads the data into the workspace.
>
>This is a bit cumbersome as it involves
>filing the ftp commands,
>executing the file with the ftp commands which gets the mainframe data and
>files it on the PC
>reading in the individual files into the APL workspace.
>
>I was wondering if I could do a better job with J.
>
>A sample of the FTP commands created by APL2 on the PC is shown here:
>
>open jesl900
>userID
>password
>cd 'PLP.ACT.AUCMI.USAA'
>get FL001 c:\temp\FL001.prn // get files from the PDS and put them on
>the PC
>get FL010 c:\temp\FL010.prn
>get FL018 c:\temp\FL018.prn
>get FL020 c:\temp\FL020.prn
>bye
>
>The commands are then placed in a file and executed by sending the following
>APL AP100 command:
> 'ftp -s:C:\temp\gather.bat>c:\temp\response.txt' // execute ftp to get
>data. PIPE ftp reply to response.txt
You can use the socket interface in J to connect to the FTP server on
the mainframe and grab the files directly into your workspace.
I wrote some sample code to do this:
http://www.interlog.com/~kbi/ftp/ftp.html
It includes a verb, ftpget, which takes a left argument naming the FTP
server and user/password, and a right argument containing the names of
the files to be grabbed. The result is the data from each file, along
with a log of the FTP transaction.
It's a bit crude in that it doesn't do much error-checking (eg bad
password or file name), but that stuff could be added easily.
/K
--------------------------------------------------------------------------------
J Forum: for information about this list, see http://www.jsoftware.com/forum.htm
From owner-jsoftware@lists.interlog.com Fri Apr 16 06:40:03 1999
Date: Fri, 16 Apr 1999 12:31:21 +0300
From: Anssi Seppala <anssi.seppala@enease.fi>
Subject: Jforum: J memory requirement estimation
In-Reply-To: <003c01be85a6$59729760$e8b0ba89@f3nbp>
With Mapped files I want to estimate what are the memory requirements to
make any operations. I wonder what this means:
createjmf_jmf_ jdatafn;10000000 NB. Create a file of 10MB
map_jmf_ 'jdata';jdatafn NB. jdata is the name of variable
jdata=:20 4000 24$i.100 NB. Assign an 8MB array to mapped variable
7!:0 '' NB. Ask J memory allocated
4740608
7!:2 '+/{.jdata' NB. Ask memory needed to a sum over the array
526144
7!:2 'jdata=: 0 (<10;2000;12) } jdata' NB. Ask memory needed to
assign one value in array.
16780416
It seems that assigning one value in a mapped array this requires 16MB
memory! It is obviously not true. So how can I estimate such operations
reliably?
Other question is that can I limit the memory allocation so that I can
reliably develope applicaitons to run in limited memory (WinCE)?
Anssi
--------------------------------------------------------------------------------
J Forum: for information about this list, see http://www.jsoftware.com/forum.htm
From owner-jsoftware@lists.interlog.com Fri Apr 16 08:28:18 1999
Date: Fri, 16 Apr 1999 08:15:44 -0400
From: Jake <jm@optonline.net>
Subject: Re: Jforum: Plot Animation ?
References: <85256754.007BC6B2.00@nyc-ntgw-n01.ny.jpmorgan.com>
Thanks !!
cdburke@interlog.com wrote:
>
> > I was hoping to generate animated plots with J, however
> > it seems that "frames" are skipped, producing a jerky halting
> result:
> > See the attached example.
> >
> > Any assistance will be greatly appreciated.
> >
> > I am running NT 4.0 SP4
> >
> > Thanks
> >
> > - Jake Moskowitz
>
> By default, J paints an isigraph window by simply marking it for
> painting when the system is otherwise idle. If J continues execution
> for a while, then the painting will be delayed.
>
> To override this, issue a glpaint'' command after you do the plot,
> i.e.
>
> ...
> while. ( it < Nt ) do.
> plot it{M
> glpaint''
>
> ---------------------------------------------------------------------------
> -----
> J Forum: for information about this list, see
> http://www.jsoftware.com/forum.htm
--
- Jake Moskowitz
--------------------------------------------------------------------------------
J Forum: for information about this list, see http://www.jsoftware.com/forum.htm
From owner-jsoftware@lists.interlog.com Fri Apr 16 08:46:07 1999
From: "Bjorn G. Helgason" <gosi@centrum.is>
Subject: RE: Jforum: Capturing data from mainframe host directly int J using FTP
Date: Fri, 16 Apr 1999 12:27:48 +-100
Content-Type: multipart/mixed; boundary="---- =_NextPart_000_01BE8805.700F0880"
------ =_NextPart_000_01BE8805.700F0880
From: Kirk B Iverson[SMTP:kbi@interlog.com]
> You can use the socket interface in J to connect to the FTP server on
> the mainframe and grab the files directly into your workspace.
Wow that sounds good !!
/Gosi
------ =_NextPart_000_01BE8805.700F0880
Content-Type: application/ms-tnef
Content-Transfer-Encoding: base64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------ =_NextPart_000_01BE8805.700F0880--
--------------------------------------------------------------------------------
J Forum: for information about this list, see http://www.jsoftware.com/forum.htm
From owner-jsoftware@lists.interlog.com Fri Apr 16 09:46:11 1999
Date: Fri, 16 Apr 1999 16:32:16 +0300
From: Anssi Seppala <anssi.seppala@enease.fi>
Subject: Jforum: WinCE libraries and Win libraries
The J CE version is missing utilites from J4.02a. Before I try testing each
of them I would like to know if there is already some information
available how JFILES.IJS, JMF.IJS work in Win CE?
Thanks,
Anssi
--------------------------------------------------------------------------------
J Forum: for information about this list, see http://www.jsoftware.com/forum.htm
From owner-jsoftware@lists.interlog.com Fri Apr 16 11:11:43 1999
Date: Fri, 16 Apr 1999 10:50:20 -0400
From: Jake <jm@optonline.net>
Subject: Jforum: Density Plot Color Gradations
References: <85256755.002ED9D4.00@nyc-ntgw-n01.ny.jpmorgan.com>
The color density plot seems to employ only about 9 colors,
even with video set to 65,000 colors and above (running NT4 SP4).
I've tried playing with the BANDCOLOR setting with no effect:
BANDCOLOR=: RGCLR
BANDCOLOR=: RGBCLR
BANDCOLOR=: RBCLR
Thanks for any assistance.
- Jake Moskowitz
(516) 889-4566
(212) 235-3787
jm@nassau.cv.net
--------------------------------------------------------------------------------
J Forum: for information about this list, see http://www.jsoftware.com/forum.htm
From owner-jsoftware@lists.interlog.com Fri Apr 16 12:53:08 1999
Date: Fri, 16 Apr 1999 12:35:46 -0400 (EDT)
From: "Brian M. Schott" <dscbms@panther.Gsu.EDU>
Subject: Jforum: How do you load J onto a cassiopeia CE "razor"
In-Reply-To: <85256754.00740E6B.00@nyc-ntgw-n01.ny.jpmorgan.com>
Casio has recently announced there E-15 handheld model with
16mb of ram and a mips vr4111 processor. They had earlier
also announced a more powerful model (E-100) which is not
yet shipping (either). These models look very appealing to
me and seem to be able to handle J4. Some relevant URLs are
listed below.
My question is how does one load the J4 executable onto one
of the WinCe devices? I assume I need a Windows computer
connected to the Web or a CDRom drive and J on CD. I also
assume a cable is supplied with these WinCE models that
connects to a port in a Windows device. I also assume that
the owners manual of the WinCE device contains some
instructions for downloading from the WinCe machine. Are all
of my assumptions correct?
Even if my assumptions are correct, I have a special problem
in that I have only a Power Macintosh 8600/200. I have
Softwindows 95 on the Mac, but do not know about the ports
needed. I may need to use a friend's Windows box for
downloading to the WinCE device.
http://www.casio.com/hpc/detail.cfm?PID=1248
http://www.cewindows.net/index.htm
(B=)
--------------------------------------------------------------------------------
J Forum: for information about this list, see http://www.jsoftware.com/forum.htm
From owner-jsoftware@lists.interlog.com Fri Apr 16 16:38:10 1999
Date: Fri, 16 Apr 1999 16:25:30 -0400
From: Cliff Reiter <reiterc@lafvax.lafayette.edu>
Subject: Jforum: x. q: y. on exact integers
Organization: Lafayette College
Content-type: text/plain; charset=iso-8859-1
References: <Pine.GSO.3.95.990416112458.12613B-100000@panther.Gsu.EDU>
I ran across the following situation where dyadic q: surprised
me by not quickly finding factors of 2 in large integers.
1 q: 1000 Asks what the power of 2 is in 1000
3
n-1 An exact integer
46684149435289146246
4|n-1 It has just one factor of 2
2
1 q: n-1 Seems to hang or take a huge amount of time
�break
� 1 q:n-1
Best,
Cliff
--
Clifford A. Reiter
Mathematics Department, Lafayette College
Easton, PA 18042 USA, 610-330-5277
http://www.lafayette.edu/~reiterc
--------------------------------------------------------------------------------
J Forum: for information about this list, see http://www.jsoftware.com/forum.htm
From owner-jsoftware@lists.interlog.com Fri Apr 16 16:51:17 1999
Date: Fri, 16 Apr 1999 16:39:51 -0400 (EDT)
From: "Brian M. Schott" <dscbms@panther.Gsu.EDU>
Subject: Re: Jforum: x. q: y. on exact integers
In-Reply-To: <37179CBA.3CDE@lafvax.lafayette.edu>
Content-Transfer-Encoding: 8bit
X-MIME-Autoconverted: from QUOTED-PRINTABLE to 8bit by plus.interlog.com id QAA00670
My Powermac 8600/200 gives the following.
1 q: n-1
|out of memory
| 1 q:n-1
(B=)
On Fri, 16 Apr 1999, Cliff Reiter wrote:
+ n-1 An exact integer
+46684149435289146246
+ 4|n-1 It has just one factor of 2
+2
+ 1 q: n-1 Seems to hang or take a huge amount of time
+�break
+� 1 q:n-1
+
--------------------------------------------------------------------------------
J Forum: for information about this list, see http://www.jsoftware.com/forum.htm
From owner-jsoftware@lists.interlog.com Fri Apr 16 17:51:40 1999
Date: Fri, 16 Apr 1999 17:37:48 -0400
From: Eric Iverson <eiverson@interlog.com>
Organization: Iverson Software Inc.
Subject: Re: Jforum: WinCE libraries and Win libraries
References: <4.1.19990416162552.00b20270@pop.kolumbus.fi>
Anssi Seppala wrote:
>
> The J CE version is missing utilites from J4.02a. Before I try testing each
> of them I would like to know if there is already some information
> available how JFILES.IJS, JMF.IJS work in Win CE?
I think if you move the jfiles.ijs script over you will find that it
works. The wince is very limited in space and we wanted to keep the
standard distributed scripts to a minimum. Given the limited space you
might be better off working with simple flat files, rather than
component files.
Unfortunately jmf.ijs mapped files won't work on CE unless you do a bit
a work. The problem is that the win32 subset on wince is not exactly the
same as that on nt and win9x. There are some differences in the win32
api calls used in jmf.ijs. I'm sure it could be made to work, but we
have not done so at this time. At first glance it might seem that mapped
files are very useful in the limited memory of wince. I'm not sure that
is the case. Memory is so limited that it might be better to just keep
data in the simplest possible flat files and use iterative and
traditional file io to access them.
If you decide to tackle trying to get jmf working on wince, I'd be glad
to try to help.
--------------------------------------------------------------------------------
J Forum: for information about this list, see http://www.jsoftware.com/forum.htm
From owner-jsoftware@lists.interlog.com Fri Apr 16 17:51:48 1999
Date: Fri, 16 Apr 1999 17:28:45 -0400
From: Eric Iverson <eiverson@interlog.com>
Organization: Iverson Software Inc.
Subject: Re: Jforum: J memory requirement estimation
References: <4.1.19990416120101.00a60100@pop.kolumbus.fi>
Anssi Seppala wrote:
>
> With Mapped files I want to estimate what are the memory requirements to
> make any operations. I wonder what this means:
>
> createjmf_jmf_ jdatafn;10000000 NB. Create a file of 10MB
> map_jmf_ 'jdata';jdatafn NB. jdata is the name of variable
> jdata=:20 4000 24$i.100 NB. Assign an 8MB array to mapped variable
> 7!:0 '' NB. Ask J memory allocated
> 4740608
> 7!:2 '+/{.jdata' NB. Ask memory needed to a sum over the array
> 526144
> 7!:2 'jdata=: 0 (<10;2000;12) } jdata' NB. Ask memory needed to assign one value in array.
> 16780416
>
The expression:
jdata=: 0 (<10;2000;12) } jdata
is not being done in place, and is generating a temporary 8mb array. I
don't think this should be the case, but it is and that is why you are
seeing the large memory requirement. I'll look into this some more and
will get back to you.
> Other question is that can I limit the memory allocation so that I can
> reliably develope applicaitons to run in limited memory (WinCE)?
There isn't much you can do here. The only tool other than 7!:0 is the
limit on any single allocation.
9!:20 y Query memory limit
9!:21 y Set memory limit (rounded to power of 2)
Setting this low will help you develop wince applications, but you
really have to test on the target.
--------------------------------------------------------------------------------
J Forum: for information about this list, see http://www.jsoftware.com/forum.htm
From owner-jsoftware@lists.interlog.com Fri Apr 16 19:14:14 1999
From: "Chris Burke" <cdburke@interlog.com>
References: <4.1.19990416162552.00b20270@pop.kolumbus.fi>
Subject: Re: Jforum: WinCE libraries and Win libraries
Date: Fri, 16 Apr 1999 18:58:19 -0400
charset="iso-8859-1"
J files is already in Wince and works fine.
JMF is not yet, see the msg from Eric.
----- Original Message -----
From: Anssi Seppala <anssi.seppala@enease.fi>
Sent: Friday, April 16, 1999 9:32 AM
Subject: Jforum: WinCE libraries and Win libraries
> The J CE version is missing utilites from J4.02a. Before I try
testing each
> of them I would like to know if there is already some information
> available how JFILES.IJS, JMF.IJS work in Win CE?
>
> Thanks,
> Anssi
--------------------------------------------------------------------------------
J Forum: for information about this list, see http://www.jsoftware.com/forum.htm
From owner-jsoftware@lists.interlog.com Fri Apr 16 20:12:36 1999
From: "Andrew Nikitin" <a_nikitin@gcc.ne.mediaone.net>
Subject: Re: Jforum: x. q: y. on exact integers
Date: Fri, 16 Apr 1999 19:55:14 -0400
charset="iso-8859-1"
-----Original Message-----
From: Brian M. Schott <dscbms@panther.Gsu.EDU>
Date: Friday, April 16, 1999 4:57 PM
>My Powermac 8600/200 gives the following.
>
> 1 q: n-1
>|out of memory
>| 1 q:n-1
>
Another spot with q:
*/p:i.39x
962947420735983927056946215901134429196419130606213075415963491270
q: */p:i.39x
|nonce error
| q:*/p:i.39
p:38 39
167 173
BTW, what does this 'nonce error' mean?
nsg
--------------------------------------------------------------------------------
J Forum: for information about this list, see http://www.jsoftware.com/forum.htm
From owner-jsoftware@lists.interlog.com Fri Apr 16 20:28:09 1999
Date: Fri, 16 Apr 1999 20:19:34 -0400
From: David Ness <DNess@home.com>
Subject: Re: Jforum: x. q: y. on exact integers
References: <021901be8864$9245aee0$1302a8c0@nevada>
Andrew Nikitin wrote:
>
<snip of earlier>
>
> BTW, what does this 'nonce error' mean?
>
> nsg
A very good question. I always thought it meant `only once, for a
special occasion'. But given the number of `nonce errors' I have
seen ...
--------------------------------------------------------------------------------
J Forum: for information about this list, see http://www.jsoftware.com/forum.htm
From owner-jsoftware@lists.interlog.com Fri Apr 16 21:42:49 1999
From: "Seymour Glass" <glasss@mindspring.com>
Subject: RE: Jforum: x. q: y. on exact integers
Date: Fri, 16 Apr 1999 21:32:07 -0400
charset="iso-8859-1"
In-Reply-To: <3717D396.63148815@Home.Com>
Importance: Normal
Nonce means 'the present time or occasion'. I think a nonce error
is an error in the current implementation that may be fixed someday.
So, q: */p:i.39x is not a domain error, but if numbers that big
aren't supported yet, it would be a nonce error.
I'm guessing, based on the nonce errors I have seen.
Henry Rich
> -----Original Message-----
> From: owner-jsoftware@lists.interlog.com
> [mailto:owner-jsoftware@lists.interlog.com]On Behalf Of David Ness
> Sent: Friday, April 16, 1999 20:20
> To: forum@jsoftware.com
> Subject: Re: Jforum: x. q: y. on exact integers
>
>
> Andrew Nikitin wrote:
> >
> <snip of earlier>
> >
> > BTW, what does this 'nonce error' mean?
> >
> > nsg
>
> A very good question. I always thought it meant `only once, for a
> special occasion'. But given the number of `nonce errors' I have
> seen ...
>
> ------------------------------------------------------------------
> --------------
> J Forum: for information about this list, see
> http://www.jsoftware.com/forum.htm
>
--------------------------------------------------------------------------------
J Forum: for information about this list, see http://www.jsoftware.com/forum.htm
From owner-jsoftware@lists.interlog.com Fri Apr 16 23:52:04 1999
From: Eemcd@aol.com
Date: Fri, 16 Apr 1999 23:43:23 EDT
Subject: Jforum: nonce error
In the original APL\360 the error report 'Nonce Error' meant that at some
time in the future it was intended that the expression would give a result,
but just now, or for the nonce, it wasn't available. When the APL system,
which originally was available only within IBM, was turned into a generally
available product, the APL group was required to remove the message, because
it was a preannouncement of a feature not available, and IBM was under a
court order not to preannounce.
J has revived nonce error to have the same meaning: not yet, but perhaps some
time in the future this facility will be available.
Eugene McDonnell
Palo Alto
--------------------------------------------------------------------------------
J Forum: for information about this list, see http://www.jsoftware.com/forum.htm
From owner-jsoftware@lists.interlog.com Sat Apr 17 02:27:39 1999
Date: Sat, 17 Apr 1999 09:15:36 +0300
From: Anssi Seppala <anssi.seppala@enease.fi>
Subject: Jforum: How to use 9!:20 an 9!:21?
In-Reply-To: <3717AB8D.2DA@interlog.com>
References: <4.1.19990416120101.00a60100@pop.kolumbus.fi>
Hi,
How can I set the memory limit? I get only errors:
9!:20 ''
1073741824
9 !: 21 10485760
|rank error
| 9!:21 10485760
9 !: 21 '10485760'
|domain error
| 9!:21'10485760'
At 17:28 16.4.1999 -0400, you wrote:
>There isn't much you can do here. The only tool other than 7!:0 is the
>limit on any single allocation.
>
>9!:20 y Query memory limit
>9!:21 y Set memory limit (rounded to power of 2)
>
>Setting this low will help you develop wince applications, but you
>really have to test on the target.
--------------------------------------------------------------------------------
J Forum: for information about this list, see http://www.jsoftware.com/forum.htm
From owner-jsoftware@lists.interlog.com Sat Apr 17 02:42:13 1999
From: "Roger Hui" <RHui@interlog.com>
Subject: Jforum: Re: How to use 9!:20 an 9!:21?
Date: Fri, 16 Apr 1999 23:27:30 -0400
charset="iso-8859-1"
Anssi Seppala writes on Saturday, April 17:
>How can I set the memory limit? I get only errors:
>
>9!:20 ''
>1073741824
>9 !: 21 10485760
>|rank error
>| 9!:21 10485760
>9 !: 21 '10485760'
>|domain error
>| 9!:21'10485760'
The right argument to the conjunction !: needs to be
separated from the argument to the resultant verb. e.g.
9!:21 (10e6)
9!:21] 10e6
(9!:21) 10e6
The phrase 9!:21 10485760 is equivalent to
9!:(21 10485760) and the system responded
correctly with a rank error.
--------------------------------------------------------------------------------
J Forum: for information about this list, see http://www.jsoftware.com/forum.htm
From owner-jsoftware@lists.interlog.com Sat Apr 17 03:01:57 1999
From: "Roger Hui" <RHui@interlog.com>
Subject: Jforum: Re: x. q: y. on exact integers
Date: Fri, 16 Apr 1999 23:53:02 -0400
charset="iso-8859-1"
Content-Transfer-Encoding: 8bit
Cliff Reiter writes on Friday, April 16:
>I ran across the following situation where dyadic q: surprised
>me by not quickly finding factors of 2 in large integers.
>
> 1 q: 1000 Asks what the power of 2 is in 1000
>3
> n-1 An exact integer
>46684149435289146246
> 4|n-1 It has just one factor of 2
>2
> 1 q: n-1 Seems to hang or take a huge amount of time
>�break
>� 1 q:n-1
There is currently only spotty support for extended integers in q: .
The algorithm used does not alway work. As well, there is no
special casing in the dyad q: -- it proceeds by first applying q:
to the right argument, and does not exploit the fact that you are
just asking for factors of 2.
--------------------------------------------------------------------------------
J Forum: for information about this list, see http://www.jsoftware.com/forum.htm
From owner-jsoftware@lists.interlog.com Sat Apr 17 09:03:29 1999
Date: Sat, 17 Apr 1999 08:51:23 -0400
From: Cliff Reiter <reiterc@lafvax.lafayette.edu>
Subject: Jforum: residue on mixed integers
Organization: Lafayette College
Content-type: text/plain; charset=us-ascii
References: <008e01be8885$f64caa80$a0b0ba89@f3nbp.lafayette.edu>
I ran across the following situation where dyadic | surprised
me by requiring exact integers for small moduli.
I figure this is an artifact of 2^3 likely being promoted
to floating type but I was still caught off guard.
I was on my search, now finished, for the power of 2 exactly
dividing exact integers. Here's the surprise:
8| 46684149435289146246x
6
(2^3)| 46684149435289146246x I expected 6
0
2^3
8
(2^3x)| 46684149435289146246x
6
--
Clifford A. Reiter
Mathematics Department, Lafayette College
Easton, PA 18042 USA, 610-330-5277
http://www.lafayette.edu/~reiterc
--------------------------------------------------------------------------------
J Forum: for information about this list, see http://www.jsoftware.com/forum.htm
From owner-jsoftware@lists.interlog.com Sat Apr 17 12:16:37 1999
From: "Roger Hui" <RHui@interlog.com>
Subject: Jforum: Re: residue on mixed integers
Date: Sat, 17 Apr 1999 09:06:47 -0400
charset="iso-8859-1"
Cliff Reiter writes on Saturday, April 17:
>I ran across the following situation where dyadic | surprised
>me by requiring exact integers for small moduli.
>I figure this is an artifact of 2^3 likely being promoted
>to floating type but I was still caught off guard.
>I was on my search, now finished, for the power of 2 exactly
>dividing exact integers. Here's the surprise:
>
> 8| 46684149435289146246x
>6
> (2^3)| 46684149435289146246x I expected 6
>0
> 2^3
>8
> (2^3x)| 46684149435289146246x
>6
This is a consequence of the type conversion rules specified
in the table in section II G of the dictionary. Specifically,
for the scalar dyadic functions, if one argument is floating point
and the other is extended, the extended argument is converted
to floating point and then the function proceeds.
If you want to stay within the exact domains (extended integer
or rational), affix an "x" to all integer constants. A future version
of J may implement a global switch which, when on, makes
all integer constants extended.
--------------------------------------------------------------------------------
J Forum: for information about this list, see http://www.jsoftware.com/forum.htm
From owner-jsoftware@lists.interlog.com Sat Apr 17 12:31:55 1999
From: Eemcd@aol.com
Date: Sat, 17 Apr 1999 12:21:27 EDT
Subject: Jforum: addendum to nonce
As a counterpart to the nonce error, I once proposed that in APL we provide a
'Yore Error' to describe a facility that had once worked but which had been
removed. For example, the monadic case of base value, which in the initial
APL\360 had been base-2 value, but which had been removed in order to make
room for upgrade/downgrade. It would indicate 'we used to have this in days
of yore, but no longer'. In J this monadic case has been restored, after more
than thirty years of having been withdrawn.
Eugene McDonnell
Palo Alto
--------------------------------------------------------------------------------
J Forum: for information about this list, see http://www.jsoftware.com/forum.htm
From owner-jsoftware@lists.interlog.com Sat Apr 17 12:53:32 1999
From: "Seymour Glass" <glasss@mindspring.com>
Subject: RE: Jforum: Re: Time series formats in J?
Date: Sat, 17 Apr 1999 12:42:55 -0400
charset="iso-8859-1"
In-Reply-To: <003c01be85a6$59729760$e8b0ba89@f3nbp>
Importance: Normal
> Anssi Seppala:
>
>
> >My applications manage time dependent data, mostly hourly or half-hourly.
> >Usually I handle it in n x 24 matrix format or n x 48 format...
> > What functions are the most suitable?
>
>
Roger Hui:
> I would store the data in inverted form (by column), with one of
> the columns
> being a time column. The time column would be day numbers and minutes
> i.e. (day number**/24 60)+minutes (or seconds, if the resolution
> is finer).
> If there is a lot of data I would use mapped files, one file per column.
>
> Summarizing is applying a function f that classifies the time
> (those that are
> the same day, same hour, same minute, etc.) and then using key (/.) to
> add up the data that belong to the same class.
>
> (f time) +//. data
>
It helps if the result of f is a scalar. With 2 <: #$x., the
key adverb /. runs in quadratic time (as I recollect), whereas
on rank-1 x. it runs faster, maybe as (* ^.) #x.
So, if f just returns (year,month,day), you get the slow version.
The slowdown results from the requirement of tolerant comparison.
As of 4.01, if you use exact comparison, you get the fast
behavior even for large rank (nice work again, Roger!). So
I have the definition
key0_z_ =: ((]:@#~) =!.0)~
and I would write
(f time) +/ key0 data
!. does not currently apply to /.; maybe someday I will be able to
replace key0 with /.!.0 .
Of course, these observations may not apply to versions 4.02 and later.
Henry Rich
--------------------------------------------------------------------------------
J Forum: for information about this list, see http://www.jsoftware.com/forum.htm
From owner-jsoftware@lists.interlog.com Sat Apr 17 13:18:36 1999
From: "Roger Hui" <RHui@interlog.com>
Subject: Jforum: Re: J memory requirement estimation
Date: Sat, 17 Apr 1999 10:08:12 -0400
charset="iso-8859-1"
Anssi Seppala writes on Friday, April 16:
>> jdata=:20 4000 24$i.100 NB. Assign an 8MB array to mapped variable
>> 7!:0 '' NB. Ask J memory allocated
>> 4740608
>> 7!:2 '+/{.jdata' NB. Ask memory needed to a sum over the array
>> 526144
>> 7!:2 'jdata=: 0 (<10;2000;12) } jdata' NB. Ask memory needed to assign one value in array.
>> 16780416
Eric Iverson:
>The expression:
> jdata=: 0 (<10;2000;12) } jdata
>is not being done in place, and is generating a temporary 8mb array. I
>don't think this should be the case, but it is and that is why you are
>seeing the large memory requirement. I'll look into this some more and
>will get back to you.
The amend _was_ being done in place, but it does something just
as bad as not doing it in place, by using a profligate expression
to compute the index positions to be amended (FYI: ind{i.$jdata).
This performance bug has now been corrected and will be
in the next release.
jdata=:20 4000 24$i.100
7!:2 'jdata=: 0 (<10;2000;12) } jdata'
3328
Thanks for reporting the error.
--------------------------------------------------------------------------------
J Forum: for information about this list, see http://www.jsoftware.com/forum.htm
From owner-jsoftware@lists.interlog.com Sat Apr 17 14:22:26 1999
From: "Roger Hui" <RHui@interlog.com>
Subject: Jforum: Re: Time series formats in J?
Date: Sat, 17 Apr 1999 11:01:57 -0400
charset="iso-8859-1"
Henry Rich writes on Saturday, April 17:
>Roger Hui:
>
>> I would store the data in inverted form (by column), with one of
>> the columns
>> being a time column. The time column would be day numbers and minutes
>> i.e. (day number**/24 60)+minutes (or seconds, if the resolution
>> is finer).
>> If there is a lot of data I would use mapped files, one file per column.
>>
>> Summarizing is applying a function f that classifies the time
>> (those that are
>> the same day, same hour, same minute, etc.) and then using key (/.) to
>> add up the data that belong to the same class.
>>
>> (f time) +//. data
>>
>
>It helps if the result of f is a scalar. With 2 <: #$x., the
>key adverb /. runs in quadratic time (as I recollect), whereas
>on rank-1 x. it runs faster, maybe as (* ^.) #x.
>
>So, if f just returns (year,month,day), you get the slow version.
>
>The slowdown results from the requirement of tolerant comparison.
>As of 4.01, if you use exact comparison, you get the fast
>behavior even for large rank (nice work again, Roger!). So
>I have the definition
The explanation is incorrect. Any slow down results
from using tolerant comparsion in the dyad i. on higher
ranked arguments. If you do precisely as I suggested,
viz., the time column is an encoded time
(day number * */24 60)+minutes, then (f time)
as left argument to "key" would be fast for any
reasonable f.
This has been the case since no later than J3.02,
1996-06-24. The dyad i. has remained substantially
unchanged since then.
--------------------------------------------------------------------------------
J Forum: for information about this list, see http://www.jsoftware.com/forum.htm
From owner-jsoftware@lists.interlog.com Sun Apr 18 01:20:59 1999
From: "Seymour Glass" <glasss@mindspring.com>
Subject: RE: Jforum: Re: Time series formats in J?
Date: Sun, 18 Apr 1999 01:09:15 -0400
charset="iso-8859-1"
Importance: Normal
In-Reply-To: <009801be88e4$b733d120$9eb0ba89@f3nbp>
> -----Original Message-----
> From: owner-jsoftware@lists.interlog.com
> [mailto:owner-jsoftware@lists.interlog.com]On Behalf Of Roger Hui
> Sent: Saturday, April 17, 1999 11:02
> To: J Forum
> Subject: Jforum: Re: Time series formats in J?
>
>
> Henry Rich writes on Saturday, April 17:
>
> >Roger Hui:
> >
> >> I would store the data in inverted form (by column), with one of
> >> the columns
> >> being a time column. The time column would be day numbers and minutes
> >> i.e. (day number**/24 60)+minutes (or seconds, if the resolution
> >> is finer).
> >> If there is a lot of data I would use mapped files, one file
> per column.
> >>
> >> Summarizing is applying a function f that classifies the time
> >> (those that are
> >> the same day, same hour, same minute, etc.) and then using key (/.) to
> >> add up the data that belong to the same class.
> >>
> >> (f time) +//. data
> >>
> >
> >It helps if the result of f is a scalar. With 2 <: #$x., the
> >key adverb /. runs in quadratic time (as I recollect), whereas
> >on rank-1 x. it runs faster, maybe as (* ^.) #x.
> >
> >So, if f just returns (year,month,day), you get the slow version.
> >
> >The slowdown results from the requirement of tolerant comparison.
> >As of 4.01, if you use exact comparison, you get the fast
> >behavior even for large rank (nice work again, Roger!). So
> >I have the definition
>
>
> The explanation is incorrect. Any slow down results
> from using tolerant comparsion in the dyad i. on higher
> ranked arguments. If you do precisely as I suggested,
> viz., the time column is an encoded time
> (day number * */24 60)+minutes, then (f time)
> as left argument to "key" would be fast for any
> reasonable f.
You are right, as ever. I didn't read your prescription
carefully, where you encode the day as a day number.
Moreover, I realize that I'm not sure that the slow cases
are quadratic, as I stated; I know only that they are
slower than the exact comparison.
However, it is true, isn't it, that a 'reasonable' f should
return a scalar. A simple-minded f that returns
(day,hour), for example, might hit the slow case.
I don't know whether the slow case applies to integer arguments
(which don't use the tolerance), so it might do to
apply <. to a vector f. .
Henry Rich
--------------------------------------------------------------------------------
J Forum: for information about this list, see http://www.jsoftware.com/forum.htm
From owner-jsoftware@lists.interlog.com Sun Apr 18 15:49:28 1999
Date: Sun, 18 Apr 1999 15:39:14 -0400
From: Jake <jm@optonline.net>
Subject: Jforum: Multiple args, some non-scalar in user-defined verb ?
References: <85256755.007904AA.00@nyc-ntgw-n01.ny.jpmorgan.com>
I gather that the way to have multiple arguments in user-defined
function, is to index the arg-list as a vector, eg:
F =: 3 : 0
mo =. 0{y.
sgn =. 1{y.
B =. 2{y.
x =. 3{y.
t =. 4{y.
. . .
But is there a clean way set this up, if one of the args can be a
vector of arbirary length ?
I can imagine messy ways of handling this, extracting the length of
the vector, etc, but suspect that there must be an elegant way ...
Thanks.
- Jake Moskowitz
(516) 889-4566
(212) 235-3787
jm@nassau.cv.net
--------------------------------------------------------------------------------
J Forum: for information about this list, see http://www.jsoftware.com/forum.htm
From owner-jsoftware@lists.interlog.com Sun Apr 18 16:02:31 1999
Date: Sun, 18 Apr 1999 15:51:47 -0400
From: David Ness <DNess@home.com>
Subject: Re: Jforum: Multiple args, some non-scalar in user-defined verb ?
References: <85256755.007904AA.00@nyc-ntgw-n01.ny.jpmorgan.com> <371A34E2.E72311B0@nassau.cv.net>
You'll probably get some more expert answers, but I can at least
try:
Normally, a construct like
F =: 3 : 0
'mo sgn B x t' =. y.
...
will work well. The usual `J' answer to `things of arbitrary length'
is `box it', so if,
a =: 3; 4 5; 6 7 8
then
'x y z' =: a
probably does what you want it to do.
Jake wrote:
>
> I gather that the way to have multiple arguments in user-defined
> function, is to index the arg-list as a vector, eg:
>
> F =: 3 : 0
> mo =. 0{y.
> sgn =. 1{y.
> B =. 2{y.
> x =. 3{y.
> t =. 4{y.
>
> . . .
>
> But is there a clean way set this up, if one of the args can be a
> vector of arbirary length ?
>
> I can imagine messy ways of handling this, extracting the length of
> the vector, etc, but suspect that there must be an elegant way ...
>
> Thanks.
>
> - Jake Moskowitz
>
--------------------------------------------------------------------------------
J Forum: for information about this list, see http://www.jsoftware.com/forum.htm
From owner-jsoftware@lists.interlog.com Sun Apr 18 19:18:24 1999
Date: Sun, 18 Apr 1999 19:09:56 -0400
From: Dan King <danking@sympatico.ca>
Subject: Jforum: Explanation Please
NB. Can anyone explain the following script? Why does the first line return
a matrix?
$( ,@>:@i. ) 11 NB. ravel of increment of an integer list of a scalar
11 1
NB. whereas
$( ,@i. ) 11 NB. ravel of an integer list of a scalar
11
>:11
12
$ >: i. 11 NB. explicit code
11
--------------------------------------------------------------------------------
J Forum: for information about this list, see http://www.jsoftware.com/forum.htm
From owner-jsoftware@lists.interlog.com Sun Apr 18 20:01:04 1999
From: "Seymour Glass" <glasss@mindspring.com>
Subject: RE: Jforum: Explanation Please
Date: Sun, 18 Apr 1999 19:50:45 -0400
charset="iso-8859-1"
Importance: Normal
In-Reply-To: <3.0.1.32.19990418190956.00706000@pop1.sympatico.ca>
The problem can be simplified to:
$ ,@>: (0 1 2 3)
4 1
understood as follows:
You have a verb (,@>:). What is its rank? You see that
the rank of u@v is (mv lv rv), and the rank of monadic >:
is 0; so the rank of (,@>:) is 0 .
So, (,@>:) is being applied to 0-cells (scalars), and
each one is being incremented and then raveled into a
vector; the vectors are assembled into the final result,
which is a list of vectors, i. e. of rank 2.
Obviously, you want something like ,@>: but with infinite
rank. That's what @: is for:
$ ,@:>: (0 1 2 3)
4
In the original problem,
$( ,@:>:@i. ) 11
11
This is one of those things that baffles every newcomer
to J for a while.
Henry Rich
> -----Original Message-----
> From: owner-jsoftware@lists.interlog.com
> [mailto:owner-jsoftware@lists.interlog.com]On Behalf Of Dan King
> Sent: Sunday, April 18, 1999 19:10
> To: forum@jsoftware.com
> Subject: Jforum: Explanation Please
>
>
> NB. Can anyone explain the following script? Why does the first
> line return
> a matrix?
>
> $( ,@>:@i. ) 11 NB. ravel of increment of an integer list
> of a scalar
> 11 1
>
> NB. whereas
> $( ,@i. ) 11 NB. ravel of an integer list
> of a scalar
> 11
> >:11
> 12
> $ >: i. 11 NB. explicit code
> 11
>
>
>
>
>
>
> ------------------------------------------------------------------
> --------------
> J Forum: for information about this list, see
> http://www.jsoftware.com/forum.htm
>
--------------------------------------------------------------------------------
J Forum: for information about this list, see http://www.jsoftware.com/forum.htm
From owner-jsoftware@lists.interlog.com Mon Apr 19 07:48:35 1999
Date: Mon, 19 Apr 1999 12:33:43 +0100
From: Ewart Shaw <strgh@warwick.ac.uk>
Organization: Statistics Dept, University of Warwick
Subject: Jforum: <.@f with extended precision arguments
1) <.@f sometimes gives incorrect results, like the following:
3 <.@%: 123456789012345678x^3 NB. dyadic <.@% unimplemented?
123456789012345742
3 <.@%: 1234567890123456789x^3 NB. but this looks like a bug!
21486409917946268784
2) Is there a canonical list of verbs f for which <.@f works under
extended
precision?
Many thanks, Ewart Shaw
--------------------------------------------------------------------------------
J Forum: for information about this list, see http://www.jsoftware.com/forum.htm
From owner-jsoftware@lists.interlog.com Mon Apr 19 09:23:32 1999
Date: Mon, 19 Apr 1999 09:13:38 -0400
From: "Kenneth E. Iverson" <kei@interlog.com>
Organization: Iverson Software Inc.
Subject: Re: Jforum: Explanation Please
References: <000001be89f6$46e00840$b191f7a5@seymourg>
Seymour Glass wrote:
>
> The problem can be simplified to:
>
> $ ,@>: (0 1 2 3)
> 4 1
>
> understood as follows:
>
> You have a verb (,@>:). What is its rank? You see that
> the rank of u@v is (mv lv rv), and the rank of monadic >:
> is 0; so the rank of (,@>:) is 0 .
...
>
> This is one of those things that baffles every newcomer
> to J for a while.
Perhaps because the close composition provided by @ was available
before the open composition provided by @:, use of the former
appears to be favoured. It might be better (at least for beginners)
to favour @: and use @ only when necessary, as in the transposed
inverse provided by:
a=:?.3 2 2$9x
|: @ %. a
_1r5 1r5
3r10 _1r20
1 _1
0 1r6
7r44 _1r11
_3r44 2r11
Trains may also be substituted for open composition by using
the cap [: as follows:
f=: [:|: [:%. ]
f a
_1r5 1 7r44
1r5 _1 _1r11
3r10 0 _3r44
_1r20 1r6 2r11
m=:[: NB. mnemonic for "monadic application"
g=: m|: m%. ]
g a
_1r5 1 7r44
1r5 _1 _1r11
3r10 0 _3r44
_1r20 1r6 2r11
--------------------------------------------------------------------------------
> J Forum: for information about this list, see http://www.jsoftware.com/forum.htm
--------------------------------------------------------------------------------
J Forum: for information about this list, see http://www.jsoftware.com/forum.htm
From owner-jsoftware@lists.interlog.com Mon Apr 19 13:28:58 1999
From: "Chris Burke" <cdburke@interlog.com>
References: <000001be89f6$46e00840$b191f7a5@seymourg> <371B2C02.7011@interlog.com>
Subject: Re: Jforum: Explanation Please
Date: Mon, 19 Apr 1999 13:13:43 -0400
charset="iso-8859-1"
----- Original Message -----
From: Kenneth E. Iverson <kei@interlog.com>
Sent: Monday, April 19, 1999 9:13 AM
Subject: Re: Jforum: Explanation Please
> ...
> Perhaps because the close composition provided by @ was available
> before the open composition provided by @:, use of the former
> appears to be favoured. It might be better (at least for beginners)
> to favour @: and use @ only when necessary, as in the transposed
> inverse provided by:
FYI, the standard library defines "on" as @: , for example:
$ , on >: 0 1 2 3
4
--------------------------------------------------------------------------------
J Forum: for information about this list, see http://www.jsoftware.com/forum.htm
From owner-jsoftware@lists.interlog.com Tue Apr 20 07:46:06 1999
From: "Bjorn G. Helgason" <gosi@centrum.is>
Subject: RE: Jforum: addendum to nonce
Date: Mon, 19 Apr 1999 12:44:15 +-100
Content-Type: multipart/mixed; boundary="---- =_NextPart_000_01BE8B22.506B22A0"
------ =_NextPart_000_01BE8B22.506B22A0
From: Eemcd@aol.com[SMTP:Eemcd@aol.com]
> As a counterpart to the nonce error, I once proposed that in APL we
> provide a 'Yore Error' to describe a facility that had once worked but
> which had been removed.
> For example, the monadic case of base value, which in the initial
> APL\360 had been base-2 value, but which had been removed in order to
> make room for upgrade/downgrade. It would indicate 'we used to have
> this in days of yore, but no longer'. In J this monadic case has been
> restored, after more than thirty years of having been withdrawn.
"You're Error" would cover a good many cases for me ;>
It would indicate I had written some stupid expression which will never work.
/Gosi
------ =_NextPart_000_01BE8B22.506B22A0
Content-Type: application/ms-tnef
Content-Transfer-Encoding: base64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------ =_NextPart_000_01BE8B22.506B22A0--
--------------------------------------------------------------------------------
J Forum: for information about this list, see http://www.jsoftware.com/forum.htm
From owner-jsoftware@lists.interlog.com Tue Apr 20 07:46:03 1999
From: "Bjorn G. Helgason" <gosi@centrum.is>
Subject: Jforum: Linux in Iceland
Date: Mon, 19 Apr 1999 12:21:46 +-100
There is a growing interest in Linux in Iceland. Tomorrow is a conference on Linux.
Check out:
www.linux.is
/gosi
--------------------------------------------------------------------------------
J Forum: for information about this list, see http://www.jsoftware.com/forum.htm
From owner-jsoftware@lists.interlog.com Tue Apr 20 07:46:22 1999
From: "Bjorn G. Helgason" <gosi@centrum.is>
Subject: RE: Jforum: nonce error
Date: Mon, 19 Apr 1999 12:37:25 +-100
Content-Type: multipart/mixed; boundary="---- =_NextPart_000_01BE8B22.486E7520"
------ =_NextPart_000_01BE8B22.486E7520
From: Eemcd@aol.com[SMTP:Eemcd@aol.com]
> In the original APL\360 the error report 'Nonce Error' meant that at some
> time in the future it was intended that the expression would give a result,
> but just now, or for the nonce, it wasn't available.
> ...... skipped a few lines ....
> J has revived nonce error to have the same meaning: not yet, but
> perhaps some time in the future this facility will be available.
Thanks for that explanation !!!
Most interesting. You might want to send that bit to c.l.a.
/Gosi
------ =_NextPart_000_01BE8B22.486E7520
Content-Type: application/ms-tnef
Content-Transfer-Encoding: base64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------ =_NextPart_000_01BE8B22.486E7520--
--------------------------------------------------------------------------------
J Forum: for information about this list, see http://www.jsoftware.com/forum.htm
From owner-jsoftware@lists.interlog.com Tue Apr 20 09:11:30 1999
Date: Tue, 20 Apr 1999 08:54:46 -0400
From: David Ness <DNess@home.com>
Subject: Re: Jforum: addendum to nonce
References: <01BE8B22.506B22A0@ls-2.skima.is>
I hesitate to say it, but `You're Error' is `Your Error' in more ways
than one...
Bjorn G. Helgason wrote:
>
> From: Eemcd@aol.com[SMTP:Eemcd@aol.com]
> > As a counterpart to the nonce error, I once proposed that in APL we
> > provide a 'Yore Error' to describe a facility that had once worked but
> > which had been removed.
> > For example, the monadic case of base value, which in the initial
> > APL\360 had been base-2 value, but which had been removed in order to
> > make room for upgrade/downgrade. It would indicate 'we used to have
> > this in days of yore, but no longer'. In J this monadic case has been
> > restored, after more than thirty years of having been withdrawn.
>
> "You're Error" would cover a good many cases for me ;>
> It would indicate I had written some stupid expression which will never work.
>
> /Gosi
>
--------------------------------------------------------------------------------
J Forum: for information about this list, see http://www.jsoftware.com/forum.htm
From owner-jsoftware@lists.interlog.com Tue Apr 20 10:29:57 1999
From: <paul_gauthier@dynasys.tm.fr>
Subject: Jforum: Windows Syntax
Date: Tue, 20 Apr 1999 16:14:43 +0200
I had the following small annoying problem more than once and I would
like to know the wizard answer to the following:
wd 'winexec "write.exe c:\Test\Document 1.doc";'
Does not work because the space between Documentand the digit 1 provides
a smooth answer.
Anyone can upgrade me on the exact Windows special trick I should
use.../Paul
--------------------------------------------------------------------------------
J Forum: for information about this list, see http://www.jsoftware.com/forum.htm
From ljdickey Tue Apr 20 13:37:22 1999
Subject: RE: Jforum: addendum to nonce
Dear Gosi,
You wrote:
> "You're Error" would cover a good many cases for me ;>
> It would indicate I had written some stupid expression which will never work.
Good joke. I like it.
But the spelling is
"Your Error."
The word "You're" is a contraction meaning "You are".
> eJ8+Ih4KAQaQCAAEAAAAAAABAAEAAQeQBgAIAAAA5AQAAAAAAADoAAENgAQAAgAAAAIAAgABBJAG
> ABwBAAABAAAADAAAAAMAADADAAAACwAPDgAAAAACAf8PAQAAAEUAAAAAAAAAgSsfpL6jEBmdbgDd
> AQ9UAgAAAABmb3J1bUBqc29mdHdhcmUuY29tAFNNVFAAZm9ydW1AanNvZnR3YXJlLmNvbQAAAAAe
> AAIwAQAAAAUAAABTTVRQAAAAAB4AAzABAAAAFAAAAGZvcnVtQGpzb2Z0d2FyZS5jb20AAwAVDAEA
> AAADAP4PBgAAAB4AATABAAAAFgAAACdmb3J1bUBqc29mdHdhcmUuY29tJwAAAAIBCzABAAAAGQAA
> AFNNVFA6Rk9SVU1ASlNPRlRXQVJFLkNPTQAAAAADAAA5AAAAAAsAQDoBAAAAAgH2DwEAAAAEAAAA
> AAAAAxs0AQiABwAYAAAASVBNLk1pY3Jvc29mdCBNYWlsLk5vdGUAMQgBBIABAB4AAABSRTogSmZv
> cnVtOiBhZGRlbmR1bSB0byBub25jZQA2CgEFgAMADgAAAM8HBAATAAwALAAPAAEANQEBIIADAA4A
> AADPBwQAEwAMACgAAgABACQBAQmAAQAhAAAANjcwOEIwNEE1MUY2RDIxMThCOUEyQ0ExRkZDMDAw
> MDAAEAcBA5AGAIAEAAASAAAACwAjAAEAAAADACYAAAAAAAsAKQABAAAAAwA2AAAAAABAADkAwH1c
> 81mKvgEeAHAAAQAAAB4AAABSRTogSmZvcnVtOiBhZGRlbmR1bSB0byBub25jZQAAAAIBcQABAAAA
> FgAAAAG+ilnzU0qwCGj2URHSi5osof/AAAAAAB4AHgwBAAAABQAAAFNNVFAAAAAAHgAfDAEAAAAQ
> AAAAZ29zaUBjZW50cnVtLmlzAAMABhBjMgebAwAHECICAAAeAAgQAQAAAGUAAABGUk9NOkVFTUNE
> QEFPTENPTVNNVFA6RUVNQ0RAQU9MQ09NQVNBQ09VTlRFUlBBUlRUT1RIRU5PTkNFRVJST1IsSU9O
> Q0VQUk9QT1NFRFRIQVRJTkFQTFdFUFJPVklERUFZT1JFAAAAAAIBCRABAAAAAQMAAP0CAAAbBgAA
> TFpGdbS5ozn/AAoBDwIVAqgF6wKDAFAC8gkCAGNoCsBzZXQyNwYABsMCgzIDxQIAcHJCcRHic3Rl
> bQKDM/cC5AcTAoM0A0UTNQdtAoP2NRLMFMV9CoAIzwnZAoAHCoENsQtgbmcxMDOyOQr/MTYMIRNQ
> bxPQ6mMFQEYDYToK4RxbHUYJDIIgRRPgY2RAYQUG8C4FoG1bU01UFFA6H+tdCotsaTOeNg3wC1UV
> YQvxID4eF5cUUQvyHUZBBCBhIAWgnnUCMASQCrEFQHRvJpDQaGUgbgIgYybwBJDBA2ByLCBJICcj
> HUHUcG8RsGQmwWEFQAuA4RSwUEwgdybwCocjP2ckTx2CHUF2aQ2wJbEn1lkFsCbwRSeCJyaSDbA1
> BPJiLSJmANADEGl0Xnko1BGAKMAnI3cFsGt5KLFidQVAKd8q7yTLd7xoaRFwL+MuwAnwIBnQdQRg
> dgmALjEvMj8kvEahBbFleGFtC1BlJ8C7JtIEYG4wADRAJdBhEbDtJ/BmMOA6MnYHQApQJ8D/NCQp
> MSbSC4AvYAcxNZ82r3MkvSlwXFwi4TR4OqIt/xHgOvUw8jQvNTMpIgWwBIF/JpI8vz3PJMsAwDCw
> NQBvZwNwLwAFsXVwCcAwAGVwL2RvdxuASBI1gEl7QcEIYGxDMjnhKQAm8Cf9KaF1KKMmsBGANVBE
> D0UfzyS8JtAEACkiZGETsDpi7nktgUGEJxAgGWAbgASQ9idI8QOgSk4kOasRgAQg/zTDS09MU0yf
> HTcZ0BPALYH6ZCfAYQGAQ7EEYC2RKOHNO9JpACAvgHllEZE6YrtK8QuAZzS0A/Am0GRIEP1IgC4K
> j0wIIf8jDltPIw6eIlQvHWQtcEu9dSdff/9VNi20Xn5JNQWgNVAFwCXA/mdHYCjAA4EvgDoiBCBH
> ovEHgCA7PlmVSR8n0S/y7ncFEAJANOFzA3Am8BPA/0fgLQA4wRNQB5AAkAIgQdXlA/BsAyBuZWWS
> MIJZhmlZlS9HKJBpWZUY8QABbqAAAAADABAQAAAAAAMAERAAAAAAQAAHMICiw1xZir4BQAAIMICi
> w1xZir4BHgA9AAEAAAAFAAAAUkU6IAAAAAC3Vw==
From owner-jsoftware@lists.interlog.com Tue Apr 20 14:48:03 1999
From: bjorn@rvk.is
X-Lotus-FromDomain: RVK
Date: Tue, 20 Apr 1999 18:33:15 +0000
Subject: Jforum: Icelandic J Primer on the web
Content-type: text/plain; charset=us-ascii
Content-Disposition: inline
I created a PDF copy of my translation of the J Primer. It will be put on
the web page
http://www.linux.is/source.php3?run=forritun_J
shortly along with references to to www.jsoftware.com and
www.juggle.gaertner.de
J is on the main page together with C and Perl. It is under "forritunarmal"
which means "programming languages". And as you all know then J is A
Programming Language (APL for short).
If you have not already done so I suggest that you create a PDF format for
all the J manuals.
It is quite good to use the PDF format because it can be used on any
platform and you do not need more than the free acrobat reader to work with
it. It can contain pictures and no problems with fonts or anything. You can
also have anymation in PDF files if wanted. I find the PDF much better and
easier to use than the HTML. You only need one file. No files on the side
for bitmaps and stuff. Another bonus is that the user can easily print the
file in whole or in parts. The user can not change anything.
There was a very interesting conference on Linux here in Iceland today. I
am hoping there will be similar growth in interest in J soon.
/Gosi
--------------------------------------------------------------------------------
J Forum: for information about this list, see http://www.jsoftware.com/forum.htm
From owner-jsoftware@lists.interlog.com Tue Apr 20 15:14:29 1999
Date: Tue, 20 Apr 1999 14:58:24 -0400
From: Eric Iverson <eiverson@interlog.com>
Organization: Iverson Software Inc.
Subject: Re: Jforum: Windows Syntax
References: <c=FR%a=_%p=DynaSys_-_Schilt%l=DYNASYS01-990420141443Z-129@mail.dynasys.tm.fr>
paul_gauthier@dynasys.tm.fr wrote:
> I had the following small annoying problem more than once and I would
> like to know the wizard answer to the following:
> wd 'winexec "write.exe c:\Test\Document 1.doc";'
> Does not work because the space between Document and the digit 1
A file name with a blank needs to be bracketed by "s so that the winexec
command can parse its parameters. In your example winexec thinks it has
2 parameters. This is tricky because wd uses "s to parse its parameters.
The easiest solution in this case is to use * so that the wd parameter
for winexec is all that follows as this allows you to include the
necessary "s around the filename. That is:
wd 'winexec *write.exe "c:\Test\Document 1.doc"'
In more complicated examples you might need to use the delimiter 255{a.
.
Note that write.exe will have to be in your path, else it won't be found
and you will need to provide a full path in the argument.
--------------------------------------------------------------------------------
J Forum: for information about this list, see http://www.jsoftware.com/forum.htm
From owner-jsoftware@lists.interlog.com Tue Apr 20 22:03:00 1999
Date: Tue, 20 Apr 1999 21:47:47 -0400 (EDT)
From: "Brian M. Schott" <dscbms@panther.Gsu.EDU>
Subject: Re: Jforum: J for Windows CE
In-Reply-To: <002201be252f$43f9fd60$2d0114d1@xps200>
In the very last line below, Chris used the term "handheld"
wince. Is there a distinction between "handheld" and "palm"
held? I am looking at the Cassiopeia E-15 which has no
keyboard but plenty of Ram and a mips processor.
(B=)
On Fri, 11 Dec 1998, Chris Burke wrote:
+J has now been released for Windows CE.
+
+This is the complete J402 language interpreter, plus a session manager
+based on the session manager and window driver in J for Windows 9x/NT.
+Scripts are easily portable between Windows CE and Windows 9x/NT, so that
+this is the ideal travelling companion for the J programmer.
+
+The system has the standard J development library, including the plot
+package and grid control. In general, window driver facilities supported by
+the Windows CE operating system are available in J for Windows CE.
+
+The new LAPACK AddOn library introduced with J402 is available and runs
+well on Windows CE.
+
+For best results, 16mb or more is recommended. J can run on a smaller
+machine, and indeed runs nicely on a machine with only 4mb, but may not
+have enough room for all the demos and labs, and may allow only very small
+apps.
+
+J executables, scripts, and files can be stored on a flashcard to extend
+the memory of the base system.
+
+The J executables take about 750K, and the full set of distributed scripts
+another 750K. The LAPACK Addon takes about 2.5MB.
+
+There are executables for the SH3 and MIPS processors for handheld machines.
--------------------------------------------------------------------------------
J Forum: for information about this list, see http://www.jsoftware.com/forum.htm
From owner-jsoftware@lists.interlog.com Wed Apr 21 04:45:23 1999
From: <paul_gauthier@dynasys.tm.fr>
Subject: RE: Jforum: Windows Syntax
Date: Wed, 21 Apr 1999 10:32:32 +0200
Content-Transfer-Encoding: 8bit
X-MIME-Autoconverted: from quoted-printable to 8bit by plus.interlog.com id EAA05281
Thank you very much. But, as you know, the annoying syntax hell is worst
when another language is involved. For instance let's take two paths
with "doc 1.doc" as examples:
First path: word"C:\Systeme\doc 1.doc"
Second path: word"Syst�me\doc 1.doc"
Then, using olemethod as in the word.js example, I can open the first
path but not the second.
(I modified the verb word to take a right argument of a file name)
P.S.: Those french accentuated characters are probably causing the
domain error I get. Even using 255{a. does not help...
Regards/Paul
>-----Message d'origine-----
>De: Eric Iverson [SMTP:eiverson@interlog.com]
>
> wd 'winexec *write.exe "c:\Test\Document 1.doc"'
>
>In more complicated examples you might need to use the delimiter 255{a.
>.
>
>Note that write.exe will have to be in your path, else it won't be found
>and you will need to provide a full path in the argument.
>
>
>-----------------------------------------------------------------------------
>---
>J Forum: for information about this list, see
>http://www.jsoftware.com/forum.htm
--------------------------------------------------------------------------------
J Forum: for information about this list, see http://www.jsoftware.com/forum.htm
From owner-jsoftware@lists.interlog.com Wed Apr 21 08:56:31 1999
Date: Wed, 21 Apr 1999 07:40:02 -0500
From: Don Guinn <donguinn@hal-pc.org>
Subject: Jforum: NB.
What is the significance of the letters "NB" in the comment verb?
--------------------------------------------------------------------------------
J Forum: for information about this list, see http://www.jsoftware.com/forum.htm
From owner-jsoftware@lists.interlog.com Wed Apr 21 08:57:26 1999
Date: Wed, 21 Apr 1999 07:38:13 -0500
From: Don Guinn <donguinn@hal-pc.org>
Subject: Jforum: Monadic Atop and Bond are different
According to the dictionary monadic uses of @ and & are equivilent.
(See Atop, second sentence.) Replacing @ with & in the last example
says otherwise.
k=: *:@+
k/ 1 2 3 4
6770404
l=: *:&+
l/ 1 2 3 4
|domain error: l
| l/1 2 3 4
This was run on:
9!:14 ''
4.01/1998-07-05/05:40
donguinn@hal-pc.org
--------------------------------------------------------------------------------
J Forum: for information about this list, see http://www.jsoftware.com/forum.htm
From owner-jsoftware@lists.interlog.com Wed Apr 21 09:01:36 1999
Date: Wed, 21 Apr 1999 08:35:25 -0400
From: Cliff Reiter <reiterc@lafvax.lafayette.edu>
Subject: Jforum: Control Keys in GUI
Organization: Lafayette College
Content-type: text/plain; charset=us-ascii
References:
<c=FR%a=_%p=DynaSys_-_Schilt%l=DYNASYS01-990421083232Z-63@mail.dynasys.tm.fr>
Hi,
I am pretty inexperienced in J GUI programming, and I am
too embarassed to ask my 12 year son for help, so I ask this
forum for a bit of help: How do I best put controls on control
keys in a J application? For example, if I want
ctrl-O to be a shortcut for the "file-open" menu, I see
how to add that to the menu, but I don't know how to best
make the ctrl-O event trigger the correct response. Should I have
a "char" event capture the keyboard and have cases for all
keys (I am thinking of the Open-Gl lab where there is
keyboard control -- can this approach handle ctrl keys?)
or is there a cleaner general approach? Pointers to
examples in the scripts and labs appreciated.
Best,
Cliff
--
Clifford A. Reiter
Mathematics Department, Lafayette College
Easton, PA 18042 USA, 610-330-5277
http://www.lafayette.edu/~reiterc
--------------------------------------------------------------------------------
J Forum: for information about this list, see http://www.jsoftware.com/forum.htm
From owner-jsoftware@lists.interlog.com Wed Apr 21 09:02:47 1999
From: "Chris Burke" <cdburke@interlog.com>
References: <Pine.GSO.3.95.990420214357.2846A-100000@panther.Gsu.EDU>
Subject: Re: Jforum: J for Windows CE
Date: Wed, 21 Apr 1999 08:39:36 -0400
charset="iso-8859-1"
> In the very last line below, Chris used the term "handheld"
> wince. Is there a distinction between "handheld" and "palm"
> held? I am looking at the Cassiopeia E-15 which has no
> keyboard but plenty of Ram and a mips processor.
Yes, J requires a WinCE with a keyboard.
--------------------------------------------------------------------------------
J Forum: for information about this list, see http://www.jsoftware.com/forum.htm
From owner-jsoftware@lists.interlog.com Wed Apr 21 09:36:30 1999
Date: Wed, 21 Apr 1999 15:19:13 +0200 (MET DST)
From: Martin Neitzel <neitzel@gaertner.de>
Subject: Re: Jforum: Monadic Atop and Bond are different
> According to the dictionary monadic uses of @ and & are equivilent.
Correct.
> (See Atop, second sentence.) Replacing @ with & in the last example
> says otherwise.
Nope, because u/ leads to the _dyad_ meaning of u, not the mondic one.
That is:
> l=: *:&+
> l/ 1 2 3 4
is equivalent to:
(1 *:&+ (2 *:&+ (3 *:&+ 4)))
which tries to NAND the complex conjugates of items and/or intermediate
results.
Martin
--------------------------------------------------------------------------------
J Forum: for information about this list, see http://www.jsoftware.com/forum.htm
From owner-jsoftware@lists.interlog.com Wed Apr 21 09:38:45 1999
Date: Wed, 21 Apr 1999 15:28:21 +0200 (MET DST)
From: Martin Neitzel <neitzel@gaertner.de>
Subject: Re: Jforum: NB.
> What is the significance of the letters "NB" in the comment verb?
NOTA BENE ("please note carefully")
(No, it's _not_ "Niobium" ;-)
Martin
--------------------------------------------------------------------------------
J Forum: for information about this list, see http://www.jsoftware.com/forum.htm
From owner-jsoftware@lists.interlog.com Wed Apr 21 09:54:12 1999
Date: Wed, 21 Apr 1999 09:38:53 -0400
From: Eric Iverson <eiverson@interlog.com>
Organization: Iverson Software Inc.
Subject: Re: Jforum: J for Windows CE
References: <Pine.GSO.3.95.990420214357.2846A-100000@panther.Gsu.EDU>
Brian M. Schott wrote:
> In the very last line below, Chris used the term "handheld"
> wince. Is there a distinction between "handheld" and "palm"
> held? I am looking at the Cassiopeia E-15 which has no
> keyboard but plenty of Ram and a mips processor.
J is officially supported on wince handheld systems at version 2.0+ for
mips and sh3 processors.
There is a new MS wince development environment for palmtops and for
handheld Pros such as the HP Jornada. The Jornada also uses a new
processor (Intel ARM).
We have not moved to the latest MS wince environment. We do not support
machines like the Jornada. We do not officially support palmtop
machines.
It is possible that J will run on a palmtop (without taking advantage of
some palmtop features). Or it may not work at all until we move to the
new MS development system. You'll just have to try it.
We have no immediate plans of moving to the new development environment
as we want to let the dust settle a bit. Perhaps around the end of the
year.
--------------------------------------------------------------------------------
J Forum: for information about this list, see http://www.jsoftware.com/forum.htm
From owner-jsoftware@lists.interlog.com Wed Apr 21 10:05:12 1999
Date: Wed, 21 Apr 1999 15:23:34 +0200 (MET DST)
From: Martin Neitzel <neitzel@gaertner.de>
Subject: Re: Jforum: Control Keys in GUI
> How do I best put controls on control keys in a J application?
> For example, if I want ctrl-O to be a shortcut for ...
... you'll just have to define a monad named "yourappname_octrl_fkey".
The standard wd_handler will trigger it automatically.
Martin
--------------------------------------------------------------------------------
J Forum: for information about this list, see http://www.jsoftware.com/forum.htm
From owner-jsoftware@lists.interlog.com Wed Apr 21 10:06:54 1999
From: <paul_gauthier@dynasys.tm.fr>
Subject: RE: Jforum: NB.
Date: Wed, 21 Apr 1999 15:48:37 +0200
Content-Transfer-Encoding: 8bit
X-MIME-Autoconverted: from quoted-printable to 8bit by plus.interlog.com id JAA16494
NB stands for Nota Bene which is latin for Note Well or Notez Bien in
french.
>-----Message d'origine-----
>De: Don Guinn [SMTP:donguinn@hal-pc.org]
>Date: mercredi 21 avril 1999 14:40
>�: J Forum
>Objet: Jforum: NB.
>
>What is the significance of the letters "NB" in the comment verb?
>
>
>
>-----------------------------------------------------------------------------
>---
>J Forum: for information about this list, see
>http://www.jsoftware.com/forum.htm
--------------------------------------------------------------------------------
J Forum: for information about this list, see http://www.jsoftware.com/forum.htm
From owner-jsoftware@lists.interlog.com Wed Apr 21 10:14:59 1999
Date: Wed, 21 Apr 1999 09:56:38 -0400
From: Eric Iverson <eiverson@interlog.com>
Organization: Iverson Software Inc.
Subject: Re: Jforum: Control Keys in GUI
References: <199904211323.PAA03874@ohura.gaertner.de>
Cliff Reiter wrote:
> How do I best put controls on control keys in a J application?
> For example, if I want ctrl-O to be a shortcut for ...
To elaborate a bit on Martin's response. Take a look at the documention
in the User Manual | Form Edit | Menu Dialog. This refers you to the
Code Dialog documentation.
--------------------------------------------------------------------------------
J Forum: for information about this list, see http://www.jsoftware.com/forum.htm
From owner-jsoftware@lists.interlog.com Wed Apr 21 10:18:06 1999
Date: Wed, 21 Apr 1999 09:51:39 -0400
From: Cliff Reiter <reiterc@lafvax.lafayette.edu>
Subject: Re: Jforum: Control Keys in GUI
Organization: Lafayette College
Content-type: text/plain; charset=us-ascii
References: <199904211323.PAA03874@ohura.gaertner.de>
Martin Neitzel wrote:
>
> > How do I best put controls on control keys in a J application?
> > For example, if I want ctrl-O to be a shortcut for ...
>
> ... you'll just have to define a monad named "yourappname_octrl_fkey".
> The standard wd_handler will trigger it automatically.
>
> Martin
This is sweet - thanks
Cliff
--
Clifford A. Reiter
Mathematics Department, Lafayette College
Easton, PA 18042 USA, 610-330-5277
http://www.lafayette.edu/~reiterc
--------------------------------------------------------------------------------
J Forum: for information about this list, see http://www.jsoftware.com/forum.htm
From owner-jsoftware@lists.interlog.com Wed Apr 21 10:20:03 1999
Date: Wed, 21 Apr 1999 10:01:21 -0400
From: David Ness <DNess@home.com>
Subject: Re: Jforum: NB.
References: <371DC722.80289D31@hal-pc.org>
I had always assumed `NB' was `Nota Bene' ...
Don Guinn wrote:
>
> What is the significance of the letters "NB" in the comment verb?
>
>
--------------------------------------------------------------------------------
J Forum: for information about this list, see http://www.jsoftware.com/forum.htm
From ljdickey Wed Apr 21 11:38:05 1999
Subject: Re: Jforum: Control Keys in GUI
>
> Hi,
> I am pretty inexperienced in J GUI programming, and I am
> too embarassed to ask my 12 year son for help, so I ask this
> forum for a bit of help:
How do you know your son does not read the Jforum?
Lee
From owner-jsoftware@lists.interlog.com Wed Apr 21 12:11:31 1999
From: Bill Harris <billh@lsid.hp.com>
Subject: Re: Jforum: NB.
Date: Wed, 21 Apr 1999 08:38:28 -0800 (PDT)
In-Reply-To: <371DDA31.7241419F@Home.Com> from "David Ness" at Apr 21, 99 10:01:21 am
> I had always assumed `NB' was `Nota Bene' ...
>
> Don Guinn wrote:
> >
> > What is the significance of the letters "NB" in the comment verb?
Learning J will make one an educated (in the classical sense) person!
(NB, nonce, copula, ...)
Bill
--
Bill Harris Year 2000 Program Office
mailto: bill_harris@am.exch.hp.com Hewlett-Packard Company
phone: (425) 335-2200 M/S 330C
fax: (425) 335-2483 8600 Soper Hill Road
web: http://hpweb.lsid.hp.com:8080/~billh/ Everett, WA 98205-1298
--------------------------------------------------------------------------------
J Forum: for information about this list, see http://www.jsoftware.com/forum.htm
From owner-jsoftware@lists.interlog.com Wed Apr 21 12:42:13 1999
Date: Wed, 21 Apr 1999 12:18:16 -0400
From: Murray Eisenberg <murray@math.umass.edu>
Organization: Mathematics & Statistics, Univ. of Mass./Amherst
Subject: Re: Jforum: NB.
References: <371DC722.80289D31@hal-pc.org>
Short for standard abbreviation n.b. (or N.B.) meaning "nota bene",
Latin for, literally, "note well".
Don Guinn wrote:
>
> What is the significance of the letters "NB" in the comment verb?
>
> --------------------------------------------------------------------------------
> J Forum: for information about this list, see http://www.jsoftware.com/forum.htm
--
Murray Eisenberg murray@math.umass.edu
Mathematics & Statistics Dept. phone 413 549-1020 (H)
Univ. of Massachusetts 413 545-2859 (W)
Amherst, MA 01003-4515
--------------------------------------------------------------------------------
J Forum: for information about this list, see http://www.jsoftware.com/forum.htm
From reiterc@lafvax.lafayette.edu Wed Apr 21 14:08:09 1999
Date: Wed, 21 Apr 1999 14:08:30 -0400
From: Cliff Reiter <reiterc@lafvax.lafayette.edu>
Subject: Re: Jforum: Control Keys in GUI
Reply-to: reiterc@lafvax.lafayette.edu
Organization: Lafayette College
Content-type: text/plain; charset=us-ascii
References: <199904211538.LAA06203@math.uwaterloo.ca>
Lee Dickey wrote:
>
> >
> > Hi,
> > I am pretty inexperienced in J GUI programming, and I am
> > too embarassed to ask my 12 year son for help, so I ask this
> > forum for a bit of help:
>
> How do you know your son does not read the Jforum?
>
> Lee
Hi Lee,
Actually, he was on the J forum for a couple months before
me and when I mentioned I'ld joined he sort of said
"duh dad". But he only has a free e-mail account and can
only read e-mail a couple time a month when he follows
me to the office on Friday nights - so the volume overwelmed
him and last I heard he had quit. :)
Best,
Cliff
--
Clifford A. Reiter
Mathematics Department, Lafayette College
Easton, PA 18042 USA, 610-330-5277
http://www.lafayette.edu/~reiterc
From owner-jsoftware@lists.interlog.com Thu Apr 22 07:29:18 1999
From: "Hans Fahlin" <hans.fahlin@alfredberg.se>
Subject: RE: Jforum: J for Windows CE
Date: Thu, 22 Apr 1999 13:15:15 +0200
X-Msmail-Priority: Normal
In-Reply-To: <371DD4ED.1BCB@interlog.com>
X-Mimeole: Produced By Microsoft MimeOLE V4.72.3110.3
Importance: Normal
charset="iso-8859-1"
Eric Iverson wrote on 21 april 1999 15:39
> We have no immediate plans of moving to the new development environment
> as we want to let the dust settle a bit. Perhaps around the end of the
> year.
>
[sigh of disappointment].
For what it is worth, you have at least one J user, i.e. me, who is eagerly
waiting for a palmtop version of J.
Regards,
Hans
--------------------------------------------------------------------------------
J Forum: for information about this list, see http://www.jsoftware.com/forum.htm
From owner-jsoftware@lists.interlog.com Thu Apr 22 08:51:59 1999
Date: Thu, 22 Apr 1999 14:57:50 +0300
From: Anssi Seppala <anssi.seppala@enease.fi>
Subject: RE: Jforum: J for Windows CE
In-Reply-To: <003a01be8cb1$65e06f60$628b67c2@absanc9098.sto.alfredberg.s
e>
References: <371DD4ED.1BCB@interlog.com>
I have been running J on handheld HP620LX. J works fine there. If you want
J, buy a handheld that runs J. There is a lot of machines available that
run J WinCE.
Anssi
At 13:15 22.4.1999 +0200, you wrote:
>For what it is worth, you have at least one J user, i.e. me, who is eagerly
>waiting for a palmtop version of J.
--------------------------------------------------------------------------------
J Forum: for information about this list, see http://www.jsoftware.com/forum.htm
From owner-jsoftware@lists.interlog.com Thu Apr 22 09:18:15 1999
From: "Hans Fahlin" <hans.fahlin@alfredberg.se>
Subject: RE: Jforum: J for Windows CE
Date: Thu, 22 Apr 1999 14:58:20 +0200
X-Msmail-Priority: Normal
In-Reply-To: <4.1.19990422145004.00a92870@pop.kolumbus.fi>
X-Mimeole: Produced By Microsoft MimeOLE V4.72.3110.3
Importance: Normal
charset="iso-8859-1"
Anssi Seppala wrote
> I have been running J on handheld HP620LX. J works fine there.
> If you want
> J, buy a handheld that runs J. There is a lot of machines available that
> run J WinCE.
>
> Anssi
>
> At 13:15 22.4.1999 +0200, you wrote:
> >For what it is worth, you have at least one J user, i.e. me, who
> is eagerly
> >waiting for a palmtop version of J.
I am sorry if I did not make myself sufficiently clear. I am eagerly
awaiting a P/PC version of J, i.e. a version that is supported on the Palm
WinCE devices, such as the Casio Cassiopeia E-10, E-100 or the HP Jornada
420. The size of these devices is comparable to that of the 3COM Palmpilot,
they are very small and they can conveniently be carried with you at all
times.
Regards,
Hans
--------------------------------------------------------------------------------
J Forum: for information about this list, see http://www.jsoftware.com/forum.htm
From owner-jsoftware@lists.interlog.com Thu Apr 22 09:36:12 1999
Date: Thu, 22 Apr 1999 09:23:03 +0100
From: maxray <maxray@cloud9.net>
Subject: Re: Jforum: J for Windows CE
References: <371DD4ED.1BCB@interlog.com> <4.1.19990422145004.00a92870@pop.kolumbus.fi>
Hi,
Which version J are you running on the handheld ?
Thanks,
Engin
Anssi Seppala wrote:
> I have been running J on handheld HP620LX. J works fine there. If you want
> J, buy a handheld that runs J. There is a lot of machines available that
> run J WinCE.
>
> Anssi
>
> At 13:15 22.4.1999 +0200, you wrote:
> >For what it is worth, you have at least one J user, i.e. me, who is eagerly
> >waiting for a palmtop version of J.
>
> --------------------------------------------------------------------------------
> J Forum: for information about this list, see http://www.jsoftware.com/forum.htm
--------------------------------------------------------------------------------
J Forum: for information about this list, see http://www.jsoftware.com/forum.htm
From owner-jsoftware@lists.interlog.com Thu Apr 22 09:55:24 1999
Date: Thu, 22 Apr 1999 15:34:18 +0200 (MET DST)
From: Martin Neitzel <neitzel@gaertner.de>
Subject: RE: Jforum: J for Windows CE
I am also quite interested in J on the Palm Pilot (or other pen-based,
small, pocketable units like the Cassiopeia or Nino). I can say for
sure that in the Pilot's PalmOS pen strokes will by default by converted
into ordinary "key events". If the same is true on WindowsCE -- which I
don't know -- AND if one of the pen-driven CE devices happens to be
based on a J-supported processor, then it should be worthwhile just to
give it a try. I would be most interested to hear reports about any
experiments in this regard.
Martin Neitzel
--------------------------------------------------------------------------------
J Forum: for information about this list, see http://www.jsoftware.com/forum.htm
From owner-jsoftware@lists.interlog.com Thu Apr 22 10:55:05 1999
Date: Thu, 22 Apr 1999 16:44:07 +0300
From: Anssi Seppala <anssi.seppala@enease.fi>
Subject: Re: Jforum: J for Windows CE
In-Reply-To: <371EDC66.AD429CE@cloud9.net>
References: <371DD4ED.1BCB@interlog.com>
<4.1.19990422145004.00a92870@pop.kolumbus.fi>
Content-Transfer-Encoding: 8bit
X-MIME-Autoconverted: from quoted-printable to 8bit by plus.interlog.com id KAA25656
J 4.02a
At 09:23 22.4.1999 +0100, you wrote:
>Hi,
>
>Which version J are you running on the handheld ?
>
>Thanks,
>
>Engin
>
>
>Anssi Seppala wrote:
>
>> I have been running J on handheld HP620LX. J works fine there. If you want
>> J, buy a handheld that runs J. There is a lot of machines available that
>> run J WinCE.
>>
>> Anssi
>>
>> At 13:15 22.4.1999 +0200, you wrote:
>> >For what it is worth, you have at least one J user, i.e. me, who is eagerly
>> >waiting for a palmtop version of J.
>>
>>
>----------------------------------------------------------------------------
>----
>> J Forum: for information about this list, see
>http://www.jsoftware.com/forum.htm
>
>
>----------------------------------------------------------------------------
>----
>J Forum: for information about this list, see
>http://www.jsoftware.com/forum.htm
-------------------------------------------------------------------
Anssi Sepp�l�
Konsultti, s�hk�markkinat ja s�hk�njakelu: Sovellukset ja ohjelmistot
e-mail: anssi.seppala@enease.fi tel: +358 9 45400550
Enease Oy fax: +358 9 45400551
Mannerheimintie 66 A mob: +358 400 760077
FIN - 00260 HELSINKI, Finland http://www.kolumbus.fi/enease/
--------------------------------------------------------------------------------
J Forum: for information about this list, see http://www.jsoftware.com/forum.htm
From owner-jsoftware@lists.interlog.com Thu Apr 22 10:57:07 1999
Date: Thu, 22 Apr 1999 17:03:25 +0300
From: Anssi Seppala <anssi.seppala@enease.fi>
Subject: RE: Jforum: J for Windows CE
In-Reply-To: <004901be8cbf$cc335170$628b67c2@absanc9098.sto.alfredberg.s
e>
References: <4.1.19990422145004.00a92870@pop.kolumbus.fi>
There has been a discussion around this wince matter. I think that there is
now too much different handheld machines to support. The J manufacturer has
decided to start with such systems that are most similar to 95/NT systems
(keyboard, mouse, screen).
As an application developer I like to have common code in 95/NT and CE as
much as possible to save work developing versions on various platforms.
Also it is very difficult to imagine how to work with J productively
without a keyboard! How to work with scripts and how to program a GUI to
show results? Can you bring ideas here?
Finally the handhelds are only one flavour of small computers. I think that
one main target for any producer of small sized systems is to be
integrated to washing machines etc. mass products. Would it be a hit
product to develop a J program to control a washing machine or a GSM
telephone?
Anssi
At 14:58 22.4.1999 +0200, you wrote:
>I am eagerly
>awaiting a P/PC version of J, i.e. a version that is supported on the Palm
>WinCE devices, such as the Casio Cassiopeia E-10, E-100 or the HP Jornada
>420. The size of these devices is comparable to that of the 3COM Palmpilot,
>they are very small and they can conveniently be carried with you at all
>times.
>
>Regards,
>
>Hans
>
>
>
>----------------------------------------------------------------------------
>----
>J Forum: for information about this list, see
>http://www.jsoftware.com/forum.htm
--------------------------------------------------------------------------------
J Forum: for information about this list, see http://www.jsoftware.com/forum.htm
From owner-jsoftware@lists.interlog.com Thu Apr 22 11:19:56 1999
Date: Thu, 22 Apr 1999 10:03:02 -0500
From: Don Guinn <donguinn@hal-pc.org>
Subject: Re: Jforum: Monadic Atop and Bond are different
References: <199904211319.PAA03858@ohura.gaertner.de>
Martin Neitzel wrote:
> > According to the dictionary monadic uses of @ and & are equivilent.
>
> Correct.
>
> > (See Atop, second sentence.) Replacing @ with & in the last example
> > says otherwise.
>
> Nope, because u/ leads to the _dyad_ meaning of u, not the mondic one.
> That is:
>
> > l=: *:&+
> > l/ 1 2 3 4
>
> is equivalent to:
> (1 *:&+ (2 *:&+ (3 *:&+ 4)))
>
> which tries to NAND the complex conjugates of items and/or intermediate
> results.
>
> Martin
The dictionary says that monadic uses of u@v and u&v are equivilent, not u.
"l/ 1 2 3 4" is monadic use of l and consequently, monadic use of u&v. If & forces
dyadic use of u where @ does not, then this is contrary to what the dictionary says.
So, which is right?
Don Guinn
--------------------------------------------------------------------------------
J Forum: for information about this list, see http://www.jsoftware.com/forum.htm
From ljdickey Thu Apr 22 11:42:01 1999
Subject: RE: Jforum: J for Windows CE
> ... . Would it be a hit
> product to develop a J program to control a washing machine or a GSM
> telephone?
In a competition, a J programmer might be the first to complete the task!
Lee
From owner-jsoftware@lists.interlog.com Thu Apr 22 12:00:30 1999
Date: Thu, 22 Apr 1999 11:40:05 -0400
From: Daniel Torres <torres@SAmerica.com>
Organization: Leo Burnett Worldwide, Latin America Headquarters
Subject: Re: Jforum: Monadic Atop and Bond are different
References: <199904211319.PAA03858@ohura.gaertner.de> <371F3A26.7D47246B@hal-pc.org>
Content-Type: multipart/mixed;
boundary="------------81D9A08B308498B2F98E3537"
This is a multi-part message in MIME format.
--------------81D9A08B308498B2F98E3537
Don Guinn wrote:
>
> Martin Neitzel wrote:
>
> > > According to the dictionary monadic uses of @ and & are equivilent.
> >
> > Correct.
> >
> > > (See Atop, second sentence.) Replacing @ with & in the last example
> > > says otherwise.
> >
> > Nope, because u/ leads to the _dyad_ meaning of u, not the mondic one.
> > That is:
> >
> > > l=: *:&+
> > > l/ 1 2 3 4
> >
> > is equivalent to:
> > (1 *:&+ (2 *:&+ (3 *:&+ 4)))
> >
> > which tries to NAND the complex conjugates of items and/or intermediate
> > results.
> >
> > Martin
>
> The dictionary says that monadic uses of u@v and u&v are equivilent, not u.
>
> "l/ 1 2 3 4" is monadic use of l and consequently, monadic use of u&v. If & forces
> dyadic use of u where @ does not, then this is contrary to what the dictionary says.
>
> So, which is right?
>
> Don Guinn
It's the "/", not the "&", which is forcing the diadic use of "u".
Remember that "/" modifies the verb to its left to produce a different,
though related, verb. The new verb is used dyadically.
Daniel
--------------81D9A08B308498B2F98E3537
name="torres.vcf"
Content-Description: Card for Daniel Torres
Content-Disposition: attachment;
filename="torres.vcf"
begin:vcard
n:Torres;Daniel
tel;fax:1-978.383-5817
tel;home:1-305.441-0369
tel;work:1-305.461-6829
x-mozilla-html:TRUE
adr:;;600 Biltmore Way APT 412;Coral Gables;FL;33134-7529;USA
version:2.1
email;internet:torres@SAmerica.com
fn:Daniel Torres
end:vcard
--------------81D9A08B308498B2F98E3537--
--------------------------------------------------------------------------------
J Forum: for information about this list, see http://www.jsoftware.com/forum.htm
From ljdickey Thu Apr 22 12:29:32 1999
Subject: Re: Jforum: Monadic Atop and Bond are different
Don Guinn writes
> The dictionary says that monadic uses of u@v and u&v are equivilent, not u.
>
> "l/ 1 2 3 4" is monadic use of l and consequently, monadic use of u&v.
> If & forces
> dyadic use of u where @ does not, then this is contrary to what the
> dictionary says.
>
> So, which is right?
I read the first clause in the above to mean that
(u&v) y
and
(u@v) y
yield the same results for all y.
I don't understand the phrase "not u".
About the rest of the quoted material, I think the
confusion is about the meaning of
f / 1 2 3 4
whether this means that
"f" is the verb and "/ 1 2 3 4" is the noun,
or
"f /" is the verb and "1 2 3 4" is the noun.
The symbol "/" represents an adverb which modifies everything
to its left, namely f, and is not part of the noun. The first
reading above is not the way J works. The meaning is
"f /" is a monadic verb taking "1 2 3 4"
as its argument, and the meaning of "(f /) 1 2 3 4" is
1 f 2 f 3 f 4 .
a dyadic usage of f.
Lee Dickey
From owner-jsoftware@lists.interlog.com Thu Apr 22 12:45:01 1999
Date: Thu, 22 Apr 1999 17:48:32 +0200 (MET DST)
From: Martin Neitzel <neitzel@gaertner.de>
Subject: Re: Jforum: Monadic Atop and Bond are different
> "l/ 1 2 3 4" is monadic use of l
No. It's a monadic use of (l/) which in turn resolves to the dyadic
meaning of (l). Separate clearly between a verb such as "l" and verbs
derived from it such as "l/". The adverb Insert derives the monadic
reduction from a dyad. (Like the _dyad _ Addition in "+/ 1 2 3 4".)
Martin
--------------------------------------------------------------------------------
J Forum: for information about this list, see http://www.jsoftware.com/forum.htm
From owner-jsoftware@lists.interlog.com Thu Apr 22 12:49:21 1999
From: Lee Dickey <ljdickey@math.uwaterloo.ca>
Date: Thu, 22 Apr 1999 12:29:32 -0400 (EDT)
Subject: Re: Jforum: Monadic Atop and Bond are different
Don Guinn writes
> The dictionary says that monadic uses of u@v and u&v are equivilent, not u.
>
> "l/ 1 2 3 4" is monadic use of l and consequently, monadic use of u&v.
> If & forces
> dyadic use of u where @ does not, then this is contrary to what the
> dictionary says.
>
> So, which is right?
I read the first clause in the above to mean that
(u&v) y
and
(u@v) y
yield the same results for all y.
I don't understand the phrase "not u".
About the rest of the quoted material, I think the
confusion is about the meaning of
f / 1 2 3 4
whether this means that
"f" is the verb and "/ 1 2 3 4" is the noun,
or
"f /" is the verb and "1 2 3 4" is the noun.
The symbol "/" represents an adverb which modifies everything
to its left, namely f, and is not part of the noun. The first
reading above is not the way J works. The meaning is
"f /" is a monadic verb taking "1 2 3 4"
as its argument, and the meaning of "(f /) 1 2 3 4" is
1 f 2 f 3 f 4 .
a dyadic usage of f.
Lee Dickey
--------------------------------------------------------------------------------
J Forum: for information about this list, see http://www.jsoftware.com/forum.htm
From owner-jsoftware@lists.interlog.com Thu Apr 22 12:55:51 1999
From: "Hans Fahlin" <hans.fahlin@alfredberg.se>
Subject: RE: Jforum: J for Windows CE
Date: Thu, 22 Apr 1999 18:03:49 +0200
X-Msmail-Priority: Normal
In-Reply-To: <4.1.19990422164544.00b973e0@pop.kolumbus.fi>
X-Mimeole: Produced By Microsoft MimeOLE V4.72.3110.3
Importance: Normal
charset="iso-8859-1"
Anssi Seppala wrote
>
> Also it is very difficult to imagine how to work with J productively
> without a keyboard! How to work with scripts and how to program a GUI to
> show results? Can you bring ideas here?
In my view, J is not only a powerful tool for developing full blown
applications. J is also a tool of thought. The high level of abstraction
made possible by its efficient notation allows for the reflection upon and
the solution of difficult problems. For example, the Sample Topics in the
Dictionary and Cliff Reiter's book on fractals and J contain a multitude of
cases where I have gained additional insight into a problem by seeing it
expressed, and executed, in J.
I do not see a P/PC J as a vehicle for application development. I see it as
a handy companion, a resource that I can take out of my pocket and use to
solve small and sometimes not so small problems encountered in my daily
life. A super-calculator, if you will. A tool to disentangle my sometimes
entangled thoughts.
Being used to pen input from using my Palmpilot since almost three years
now, I do not see a problem, but rather convenience, when imagining my use
of J in the context I have tried to describe.
By all means, I can also imagine writing small applications on e.g. my PC to
use on the P/PC. Here, the primary purpose is to use J as the extremely
quick application development system it is, and have those applications
accessible from wherever I am, as they will be executable in an extremely
portable system. There are, of course, limits to how big or complex such an
application could be.
The success of the Palmpilot and the growing number of WinCE P/PCs are proof
enough of the demand for ultra-portable computing devices. The business
decision to write and support a version of J for them is not easy, and not
mine to make. I can only express my demand, as a current J customer, for
such a J version, and hope, that if others share my ideas, maybe JSoftware
might decide to give it a try.
Regards,
Hans
--------------------------------------------------------------------------------
J Forum: for information about this list, see http://www.jsoftware.com/forum.htm
From owner-jsoftware@lists.interlog.com Thu Apr 22 14:30:28 1999
From: "Simon Garland" <simon_garland@bluewin.ch>
References: <199904221629.MAA05883@goedel.math.uwaterloo.ca>
Subject: Re: Jforum: Monadic Atop and Bond are different
Date: Thu, 22 Apr 1999 20:10:49 +0200
charset="iso-8859-1"
X-Mimeole: Produced By Microsoft MimeOLE V5.00.2014.211
something that always helps me sort out things like this are the utility
functions in the SYSTEM\PACKAGES\MISC\FGH.IJS script.
f_ means monadic f, _f_ is dyadic, heres an extract from a session log:
-snip-
load'E:\j402\SYSTEM\PACKAGES\MISC\FGH.IJS'
f
ff :.fi
f 1
f_ 1
f/1 2 3 4
1 _f_ (2 _f_ (3 _f_ 4))
(f /) 1 2 3 4
1 _f_ (2 _f_ (3 _f_ 4))
i=.f&g
i/1 2 3 4
(g_ 1) _f_ (g_ ((g_ 2) _f_ (g_ ((g_ 3) _f_ (g_ 4)))))
j=.f@g
j/1 2 3 4
f_ (1 _g_ (f_ (2 _g_ (f_ (3 _g_ 4)))))
-end snip-
/simon
----- Original Message -----
From: Lee Dickey <ljdickey@math.uwaterloo.ca>
Sent: Thursday, April 22, 1999 6:29 PM
Subject: Re: Jforum: Monadic Atop and Bond are different
> Don Guinn writes
> > The dictionary says that monadic uses of u@v and u&v are equivilent, not
u.
> >
> > "l/ 1 2 3 4" is monadic use of l and consequently, monadic use of u&v.
> > If & forces
> > dyadic use of u where @ does not, then this is contrary to what the
> > dictionary says.
> >
> > So, which is right?
>
> I read the first clause in the above to mean that
> (u&v) y
> and
> (u@v) y
> yield the same results for all y.
> I don't understand the phrase "not u".
>
> About the rest of the quoted material, I think the
> confusion is about the meaning of
> f / 1 2 3 4
> whether this means that
> "f" is the verb and "/ 1 2 3 4" is the noun,
> or
> "f /" is the verb and "1 2 3 4" is the noun.
>
> The symbol "/" represents an adverb which modifies everything
> to its left, namely f, and is not part of the noun. The first
> reading above is not the way J works. The meaning is
> "f /" is a monadic verb taking "1 2 3 4"
> as its argument, and the meaning of "(f /) 1 2 3 4" is
> 1 f 2 f 3 f 4 .
> a dyadic usage of f.
>
> Lee Dickey
>
>
> --------------------------------------------------------------------------
------
> J Forum: for information about this list, see
http://www.jsoftware.com/forum.htm
>
--------------------------------------------------------------------------------
J Forum: for information about this list, see http://www.jsoftware.com/forum.htm
From owner-jsoftware@lists.interlog.com Thu Apr 22 14:32:57 1999
Date: Thu, 22 Apr 1999 11:11:00 -0700
From: David Leibs <leibs@neometron.com>
Subject: Re: Jforum: J for Windows CE
References: <005201be8cd9$b5df3a50$628b67c2@absanc9098.sto.alfredberg.se>
Just so you don't feel like a voice crying in the wilderness I too have
want the same handy super-calculator.
I know we will have it someday. We will probably just have to wait a
little bit longer.
Since J is one of those great viruses (Technology/Notation/ThoughtWare)
that comes along every so often that actually makes people smarter its
emergence onto the near future Palmpilot like machines is probably
guaranteed.
Hans Fahlin wrote:
> In my view, J is not only a powerful tool for developing full blown
> applications. J is also a tool of thought. The high level of abstraction
> made possible by its efficient notation allows for the reflection upon and
> the solution of difficult problems. For example, the Sample Topics in the
> Dictionary and Cliff Reiter's book on fractals and J contain a multitude of
> cases where I have gained additional insight into a problem by seeing it
> expressed, and executed, in J.
>
> I do not see a P/PC J as a vehicle for application development. I see it as
> a handy companion, a resource that I can take out of my pocket and use to
> solve small and sometimes not so small problems encountered in my daily
> life. A super-calculator, if you will. A tool to disentangle my sometimes
> entangled thoughts.
>
> Being used to pen input from using my Palmpilot since almost three years
> now, I do not see a problem, but rather convenience, when imagining my use
> of J in the context I have tried to describe.
>
> By all means, I can also imagine writing small applications on e.g. my PC to
> use on the P/PC. Here, the primary purpose is to use J as the extremely
> quick application development system it is, and have those applications
> accessible from wherever I am, as they will be executable in an extremely
> portable system. There are, of course, limits to how big or complex such an
> application could be.
>
> The success of the Palmpilot and the growing number of WinCE P/PCs are proof
> enough of the demand for ultra-portable computing devices. The business
> decision to write and support a version of J for them is not easy, and not
> mine to make. I can only express my demand, as a current J customer, for
> such a J version, and hope, that if others share my ideas, maybe JSoftware
> might decide to give it a try.
>
> Regards,
>
> Hans
>
>
>
> --------------------------------------------------------------------------------
> J Forum: for information about this list, see http://www.jsoftware.com/forum.htm
--------------------------------------------------------------------------------
J Forum: for information about this list, see http://www.jsoftware.com/forum.htm
From owner-jsoftware@lists.interlog.com Thu Apr 22 15:10:02 1999
From: "Nichols, Peter" <pnichols@sprinc.com>
Subject: RE: Jforum: J for Windows CE
Date: Thu, 22 Apr 1999 13:46:34 -0500
charset="iso-8859-1"
Yes, I agree. I have a number of small applications I have written that
require minimal keyboard input to use. Having a machine that would run
these and is small enough to carry around in my pocket would be sweet. I
guess my question for the folks whose time I'm requesting is how much of a
problem is it to port down? And what other enhancements are being created
for the next J release? If we, the J community request this, what other
features might we then be choosing to wait for?
-----Original Message-----
From: David Leibs [SMTP:leibs@neometron.com]
Sent: Thursday, April 22, 1999 1:11 PM
To: forum@JSoftware.Com
Subject: Re: Jforum: J for Windows CE
Just so you don't feel like a voice crying in the wilderness I too
have
want the same handy super-calculator.
I know we will have it someday. We will probably just have to wait
a
little bit longer.
Since J is one of those great viruses
(Technology/Notation/ThoughtWare)
that comes along every so often that actually makes people smarter
its
emergence onto the near future Palmpilot like machines is probably
guaranteed.
Hans Fahlin wrote:
> In my view, J is not only a powerful tool for developing full
blown
> applications. J is also a tool of thought. The high level of
abstraction
> made possible by its efficient notation allows for the reflection
upon and
> the solution of difficult problems. For example, the Sample Topics
in the
> Dictionary and Cliff Reiter's book on fractals and J contain a
multitude of
> cases where I have gained additional insight into a problem by
seeing it
> expressed, and executed, in J.
>
> I do not see a P/PC J as a vehicle for application development. I
see it as
> a handy companion, a resource that I can take out of my pocket and
use to
> solve small and sometimes not so small problems encountered in my
daily
> life. A super-calculator, if you will. A tool to disentangle my
sometimes
> entangled thoughts.
>
> Being used to pen input from using my Palmpilot since almost three
years
> now, I do not see a problem, but rather convenience, when
imagining my use
> of J in the context I have tried to describe.
>
> By all means, I can also imagine writing small applications on
e.g. my PC to
> use on the P/PC. Here, the primary purpose is to use J as the
extremely
> quick application development system it is, and have those
applications
> accessible from wherever I am, as they will be executable in an
extremely
> portable system. There are, of course, limits to how big or
complex such an
> application could be.
>
> The success of the Palmpilot and the growing number of WinCE P/PCs
are proof
> enough of the demand for ultra-portable computing devices. The
business
> decision to write and support a version of J for them is not easy,
and not
> mine to make. I can only express my demand, as a current J
customer, for
> such a J version, and hope, that if others share my ideas, maybe
JSoftware
> might decide to give it a try.
>
> Regards,
>
> Hans
>
>
>
>
----------------------------------------------------------------------------
----
> J Forum: for information about this list, see
http://www.jsoftware.com/forum.htm
----------------------------------------------------------------------------
----
J Forum: for information about this list, see
http://www.jsoftware.com/forum.htm
--------------------------------------------------------------------------------
J Forum: for information about this list, see http://www.jsoftware.com/forum.htm
From owner-jsoftware@lists.interlog.com Thu Apr 22 18:12:52 1999
Date: 22 Apr 99 13:41:54 -0800
Subject: Jforum: QR Decomposition (128!:0)
From: "Piet de Jong" <piet.dejong@commerce.ubc.ca>
I'm having trouble with the QR decomposition in J defined through the
foreign conjuction 128!:0 . The manual states that the Q matrix in 128!:0
is Hermitian -- but this seems not to be the case. ( I think the manual
means "orthogonal" when it states "hermitian")
In particular the Q matrix does not appear to be orthogonal when the matrix
has rows of zeros. Here is a session:
X=. 2 2 $ 1 3 0 0 NB. A 2x2 matrix with a bottom row of zeros
Q=.>@{.@(128!:0) NB. The Q matrix of the QR decomposition
R=.>@{:@(128!:0) NB. The R matrix of the QR decomposition
Q X
1 1
0 0
This is neither Hermitian nor orthogonal. However if mp denotes the matrix
product
(] -: Q mp R) X
1
Orthogonality seems to be the case if the X matrix is of full column rank.
The background to this problem is that I want the 128!:0 foreign
conjunction to work for cases where X is of deficient column rank. For
example when it has more columns then rows.
Any help appreciated. Please note that I am using J3x on a PowerMac.
--------------------------------------------------------------------------------
J Forum: for information about this list, see http://www.jsoftware.com/forum.htm
From owner-jsoftware@lists.interlog.com Thu Apr 22 23:21:01 1999
From: "Roger Hui" <RHui@interlog.com>
Subject: Jforum: Re: QR Decomposition (128!:0)
Date: Thu, 22 Apr 1999 15:44:38 -0400
charset="iso-8859-1"
X-Mimeole: Produced By Microsoft MimeOLE V4.72.3110.3
Piet de Jong writes on Thursday, April 22:
>I'm having trouble with the QR decomposition in J defined through the
>foreign conjuction 128!:0 . The manual states that the Q matrix in 128!:0
>is Hermitian -- but this seems not to be the case. ( I think the manual
>means "orthogonal" when it states "hermitian")
>
>In particular the Q matrix does not appear to be orthogonal when the matrix
>has rows of zeros. Here is a session:
>
> X=. 2 2 $ 1 3 0 0 NB. A 2x2 matrix with a bottom row of zeros
> Q=.>@{.@(128!:0) NB. The Q matrix of the QR decomposition
> R=.>@{:@(128!:0) NB. The R matrix of the QR decomposition
> Q X
>1 1
>0 0
>
>This is neither Hermitian nor orthogonal. However if mp denotes the matrix
>product
>
> (] -: Q mp R) X
>1
>
>Orthogonality seems to be the case if the X matrix is of full column rank.
>The background to this problem is that I want the 128!:0 foreign
>conjunction to work for cases where X is of deficient column rank. For
>example when it has more columns then rows.
0. The manual means Hermitian. A real Hermitian matrix is orthogonal,
but some Hermitian matrices are not real (have non-zero imaginary parts).
1. The monad 128!:0 only works on non-singular matrices.
--------------------------------------------------------------------------------
J Forum: for information about this list, see http://www.jsoftware.com/forum.htm
From owner-jsoftware@lists.interlog.com Fri Apr 23 04:04:45 1999
Date: Fri, 23 Apr 1999 03:48:41 -0400
From: Fraser Jackson <Fraser_Jackson@compuserve.com>
Subject: Jforum: Re: QR Decomposition (128!:0)
Content-Disposition: inline
Content-Transfer-Encoding: 8bit
X-MIME-Autoconverted: from quoted-printable to 8bit by plus.interlog.com id DAA25159
The behaviour of 128!:0 has been changed and in an earlier version it did
permit singular matrices. This enabled programming regression models
without worrying about whether the data matrix was singular. it is a
common situation to want to estimate a model across a range of data sets,
and sometimes the data happens to be singular because in that data set
there was no variation in one of the variables.
I have seen no reference to this change, but certainly many programs in J
versions 3.x do not work now.
The earlier algorithm usually gave sensible results in this case with
undefined values in the regression output where this was appropriate.
However I did have problems with some cases, and I suspect that Roger has
changed the algorithm to ensure some possible ambiguity in the behaviour of
the earlier one. I too would like 128!:0 to deal with singular cases as
Piet de Jong would, though always in cases with at least as many rows as
columns.
Fraser.
--------------------------------------------------------------------------------
J Forum: for information about this list, see http://www.jsoftware.com/forum.htm
From owner-jsoftware@lists.interlog.com Fri Apr 23 06:11:27 1999
Date: Fri, 23 Apr 1999 10:52:57 +0100 (BST)
From: J E H Shaw <strgh@csv.warwick.ac.uk>
Subject: Jforum: T. and t: on polynomials with rational coeffs.
If a is a rational/extended integer vector, then a&p T.
keeps the extended precision, whereas a&p t: (weighted Taylor coeff.)
converts to real. Would it be easy and sensible to keep the extended
precision in this particular case? Here's an example of when this
could be useful (simple calculation/demonstration of Bernoulli numbers):
a=. }. % ! i. 10x NB. Taylor series of (^ - 1:) % ]
(%@(a&p.)) T. 9 NB. -> coeffs 1 _1r2 1r12 0 _1r720 etc.
(%@(a&p.)) t: i. 9 NB. result is real, not rational
x: (%@(a&p.)) t: 6 NB. correctly gives 1r42 (6th B.n.)
x: (%@(a&p.)) t: 8 NB. _65088267767r1952648033004 not _1r30
Regards, Ewart Shaw
--------------------------------------------------------------------------------
J Forum: for information about this list, see http://www.jsoftware.com/forum.htm
From owner-jsoftware@lists.interlog.com Fri Apr 23 10:35:33 1999
Date: Fri, 23 Apr 1999 10:11:15 -0400
From: Eric Iverson <eiverson@interlog.com>
Organization: Iverson Software Inc.
Subject: Re: Jforum: J for Windows CE
References: <005201be8cd9$b5df3a50$628b67c2@absanc9098.sto.alfredberg.se> <371F6634.D5273BB7@neometron.com>
David Leibs, Hans Fahlin, et. al.
If you want the power of J at your fingertips today, then I urge you to
take a look at the lightest and smallest of the wince handhelds. They
are only a little bigger than the palmtops. They provide quite a bit
more power and capacity with a small increase in size. You can use pen
input with them as well as keyboard. They are available with J today.
Or you can wait a year. If you choose to wait a year, then I would have
to conclude that the value of either a handheld or palmheld J to you is
somewhat limited.
--------------------------------------------------------------------------------
J Forum: for information about this list, see http://www.jsoftware.com/forum.htm
From owner-jsoftware@lists.interlog.com Fri Apr 23 12:11:35 1999
Date: Fri, 23 Apr 1999 11:51:35 -0400
From: Daniel Torres <torres@SAmerica.com>
Organization: Leo Burnett Worldwide, Latin America Headquarters
Subject: Re: Jforum: J for Windows CE
References: <005201be8cd9$b5df3a50$628b67c2@absanc9098.sto.alfredberg.se> <371F6634.D5273BB7@neometron.com> <37207F83.5C58@interlog.com>
Content-Type: multipart/mixed;
boundary="------------41859F7A804F0F5102D7601A"
This is a multi-part message in MIME format.
--------------41859F7A804F0F5102D7601A
Then, of course, if you're a *real* power user, take a look at the Jeep
of WinCEs: "http://www.itronix.com/T5200/specs.html".
Daniel
Eric Iverson wrote:
>
> David Leibs, Hans Fahlin, et. al.
>
> If you want the power of J at your fingertips today, then I urge you to
> take a look at the lightest and smallest of the wince handhelds. They
> are only a little bigger than the palmtops. They provide quite a bit
> more power and capacity with a small increase in size. You can use pen
> input with them as well as keyboard. They are available with J today.
>
..
>
--------------41859F7A804F0F5102D7601A
name="torres.vcf"
Content-Description: Card for Daniel Torres
Content-Disposition: attachment;
filename="torres.vcf"
begin:vcard
n:Torres;Daniel
tel;fax:1-978.383-5817
tel;home:1-305.441-0369
tel;work:1-305.461-6829
x-mozilla-html:TRUE
adr:;;600 Biltmore Way APT 412;Coral Gables;FL;33134-7529;USA
version:2.1
email;internet:torres@SAmerica.com
fn:Daniel Torres
end:vcard
--------------41859F7A804F0F5102D7601A--
--------------------------------------------------------------------------------
J Forum: for information about this list, see http://www.jsoftware.com/forum.htm
From owner-jsoftware@lists.interlog.com Fri Apr 23 12:20:36 1999
From: "Roger Hui" <RHui@interlog.com>
Subject: Jforum: Re: This is a Fork!
Date: Fri, 23 Apr 1999 09:02:15 -0400
charset="iso-8859-1"
"MellemF" writes on Friday, April 23:
>I think the DISPLAY FORM for simply entered strings of verbs
>should show the order of execution.
>
>Or to para phrase Crocodile Dundee:
>
> +/%# That's not a fork ; This is a fork! (+/%#).
If you chose the "Parens" form of display (, Mate :-),
you get a display that is similar to what you requested.
Only the outermost parens are suppressed. Thus:
+/ % #
(+/) % #
[: %: +/ % #
[: %: ((+/) % #)
The options are in the Edit...|Configure... menu.
--------------------------------------------------------------------------------
J Forum: for information about this list, see http://www.jsoftware.com/forum.htm
From owner-jsoftware@lists.interlog.com Fri Apr 23 13:16:17 1999
Date: Fri, 23 Apr 1999 12:57:05 -0400 (EDT)
From: "J. Patrick Harrington" <jph@astro.umd.edu>
Subject: Jforum: Re: J for Windows CE
There has been some discussion of running J on
handheld/palmtop machines recently, so maybe some
would find my experience with one such device of
interest.
In 1992 I started using J on an HP 95LX and later
on an HP 100LX. These are fine little machines, and
and they seem to run forever on a pair of AA cells.
However, Iverson Software has dropped support for DOS
and my HP is stuck at J 2.02 -- circa 1994. So the code
I wrote for my other machines began to diverge from
the HP. Thus I was very interested in the announcement
last year that J would be released in a version for
Windows CE. I started to look at the newer handhelds.
Eugene McDonnell posted (11 Dec 98) something about the
Philips Velo 500, and this seemed to be a good J CE
platform: not much larger than the old HP 100LX, a
reasonable battery life (though this is a problem
with all these machines), and it was supposed to be
faster than most of this class (it has a 75 MHz MIPS
processor). The screen is a nice 640 x 240 pixel size,
but since it's only 16-level greyscale (to be hot, it
has to be color now), the prices have dropped -- now
listed at $550 for the 24 MB version by Mobile Planet;
(I paid less for mine through buycomp.com, but now they
just list the 16 MB ($400), and that's not even in stock.
I think that Philips is dropping this machine.)
You should know something about the CPUs in these
machines (SH3, MIPS, now StrongArm). They are very
different from the Intel `86 chips, and thus no DOS
or Windows programs need apply. A lot of programs you
might want just don't exist yet. For example, you don't
want to use the built-in Pocket Word to edit program
files! I'd like a "vi" type editor, but there is no such
beast for CE. So far I've found "Notepad" clones and a
(beta) Micro Emacs. And some things you want may *never*
exist, because CE 2.0 will soon be replaced by CE 3.0,
to be replaced by ..., etc. and leave your machine behind.
And you should know a few things about Windows CE.
It is set up to exist as an appendage of a PC running
Windows. Although CE has "terminal communication
software" built in, there is no way of using it to
talk to a non-windows machine through the serial port
without 3rd party software! What you must do is install
a package called CE Services on your PC, and then
you get the directory tree of your handheld displayed
on the PC and you can move files back and forth. This
is easy if your machine runs Windows 95, since there
are no security issues with 95 (i.e., no security).
However, my PC's are dual boot Linux and Windows NT.
Since NT has some idea of security, and the CE Services
software is trying to set up a network link with your
handheld, NT regards this with deep suspicion. Even
with Velo support explaining things they left out of the
instructions and faxing you pages of instructions with
stuff about setting up guest accounts, and what odd IP
addresses, and which service packs to install, and
reboot, and reinstall, etc. I *still* have only got my
Velo to link with one of my two machines. But once
you are connected to one PC, you can then install a
3rd party terminal program (I'm using vxHpc) and use
e.g. the kermit protocol to transfer files between your
Unix box and the handheld.
So if you only have a Windows NT PC, you are in for
a taste of Microsoft hell. If you don't have any Windows
machine, you may be S.O.L. While you can connect to the
internet through the built-in modem and Pocket Internet
Explorer (sort of neat), I don't think Pocket IE supports
ftp downloads. There are good 3rd party ftp programs like
"Scotty" that let you download software, but you've got
a bootstrap problem. (You can move stuff with a flash
card, but once again, the software to support the flash
has to be put into the handheld first.)
OK. So you've got your handheld CE synchronized with
your PC running CE Services. Then you just execute the
.exe file you've downloaded from Iverson Software and it
does the installation of J 402a on your handheld. (There
are two ways of installing stuff on CE machines: either
by running an .exe installation file on the PC, or by loading
a .CAB file into the handheld. When you run the .CAB file
on the handheld, it does the installation.)
After all this initial frustration, I do like running J
on the Velo. My programs all seem to run, and much faster
that the old HP 100. And the plot routines are nice. I've
just compared a routine to integrate a few stiff differential
equations on the Velo with my Sun SPARC 20. I find times
of 1.97 sec on the SPARC and 13 sec on the Velo. Another
test, inverting a matrix in rational (exact) math:
a=. x: 8 8 $ 64 ? 100
ai=. %. a
This takes 3.95 sec on the SPARC and about 30 sec on the
Velo. I say about because something weird has happened:
doing 6!:2 'ai=. %. a' on the Velo gives me results like
_27 or 30 or _31. I.e., integers, some of which are negative!
But the absolute values are about right. Is this a bug?
Anyway, I'm glad J is available on this platform, and I hope
there is enough interest to keep the product out there. I like
being able to use the LAPAC library on the Velo. (Hey, what
about the FFTW package? Will that be made available for CE?)
My problem now is that the Velo has more features than the J
I normally use (either Sun SPARC or Linux). I also have J for
Windows 95/NT, but I almost never boot up in NT except to run
Photoshop. I really wish that J would support X11 for some
of the GUI stuff. I can see why the folks at Iverson Software
have to give 1st place to the Windows market, and linking in
with the Windows GUI makes J viable for those who write software
professionally. But there are many of us in the scientific/academic
world who use J as a research tool, and to us Linux is more
and more attractive. It would be great if J support for that
platform could extend beyond the basic J language.
J. Patrick Harrington
-------------------------------------------------------------------
P.S.
Curious things about the Velo 500: There are four
little doors on the bottom. Behind them are:
(1) the coin backup battery
(2) A space for a flash storage card. However, the Velo
does not take a standard Type II card, or one of the
"compact flash" cards, but the small (and rare)
"miniature" flash card. Although these were introduced
by Intel and others, they seemed to be available only
from Philips -- and now Philips doesn't seem to carry
them any more. They only hold 4MB, but are really cute
things (38 x 33 x 3.5 mm). They have no pins, but flat
contacts that press against a strange rubbery strip in
the Velo. When I got a miniature flash card, I found that
my "rubbery strip" was defective, and eventually caused
a cold boot of the machine with a total loss of all my
programs/data. I sent it back to Philips; they didn't
try to fix it, just sent me a new Velo. The new one now
works fine with the miniature cards.
(3) A slot for added RAM -- also a "miniature" 8MB DRAM thing
that brings the Velo storage/program memory up to 24 MB.
My Velo came with this installed.
(4) A slot with contacts to plug in the V-module accessory,
which makes the unit thicker, but allows you to use the
Type II PC cards. In particular, you can use the regular
San Disk flash cards. It turns out that standard flash
cards are the best solution to the problem of importing/
exporting large amounts data to/from the Velo. While this
makes the Velo less portable, the V-module can snapped on
and off as needed.
--------------------------------------------------------------------------------
J Forum: for information about this list, see http://www.jsoftware.com/forum.htm
From owner-jsoftware@lists.interlog.com Fri Apr 23 14:41:51 1999
From: bjorn@rvk.is
X-Lotus-FromDomain: RVK
Date: Fri, 23 Apr 1999 18:24:53 +0000
Subject: Jforum: J on the web
Content-type: text/plain; charset=us-ascii
Content-Disposition: inline
I have managed to get J referred to on this web page
http://www.ht.is/lofatolvur.html
/Gosi
--------------------------------------------------------------------------------
J Forum: for information about this list, see http://www.jsoftware.com/forum.htm
From owner-jsoftware@lists.interlog.com Fri Apr 23 14:51:04 1999
Date: 23 Apr 99 10:59:49 -0700
Subject: Jforum: QR Decomposition (128!:0)
From: "Piet de Jong" <piet.dejong@commerce.ubc.ca>
X-Fontfamily: Geneva
X-Fontsize: 12
Content-Transfer-Encoding: quoted-printable
According to the definitions I see in the various texts, a Hermitian
matrix is one which equals its conjugate transpose (ie symmetric in
the real case).
The Q matrix in the QR decomposition provided by 128!:0 is not
generally Hermitian. (I'm using the implementation in J3x). However
it does appear to be orthonormal in for example the case where the
matrix operated on is of full column rank.
Maybe I am mistaken but in the math literature, the QR decomposition
of X refers to finding matrices Q and R such that X=3DQR where Q*Q=3DI
and R is square triangular and Q* is the conjugate transpose of Q.
Thus Q is orthonormal (not nec. Hermitian).
Generally QR decompositions are constructed for matrices of full
column rank and hence not necessarily square. (as implied by Roger's
nonsingularity remark).
The J3x manual states that the domain of 128!:0 is the same as that of
%. The domain of %. seems to include non square matrices. This
appears to be at variance with Roger's remarks.
In statistics the QR decomposition is very useful in regression where
the X matrix is the regressor matrix, with generally more rows then
columns. The QR decomposition then performs "safe" regression.
Deficient column rank occurs often (eg overparametrized designs,
perfect multicollinearity) and hence having a QR decomposition that
covers this case is very useful.
To summarize, the 128!:0 foreign conjuction in J3 seems to provide the
QR decomposition as defined in various texts provided X is of full
column rank. It does not appear to work as described in the
dictionary (eg "Hermitian" Q) or as in Roger's remark (X must be
nonsingular).
-- Piet
--------------------------------------------------------------------------------
J Forum: for information about this list, see http://www.jsoftware.com/forum.htm
From owner-jsoftware@lists.interlog.com Fri Apr 23 15:11:40 1999
From: "Roger Hui" <RHui@interlog.com>
Subject: Jforum: Re: T. and t: on polynomials with rational coeffs.
Date: Fri, 23 Apr 1999 11:52:25 -0400
charset="iso-8859-1"
Ewart Shaw writes on Friday, April 23:
>If a is a rational/extended integer vector, then a&p T.
>keeps the extended precision, whereas a&p t: (weighted Taylor coeff.)
>converts to real. Would it be easy and sensible to keep the extended
>precision in this particular case? Here's an example of when this
>could be useful (simple calculation/demonstration of Bernoulli numbers):
>
> a=. }. % ! i. 10x NB. Taylor series of (^ - 1:) % ]
> (%@(a&p.)) T. 9 NB. -> coeffs 1 _1r2 1r12 0 _1r720 etc.
> (%@(a&p.)) t: i. 9 NB. result is real, not rational
> x: (%@(a&p.)) t: 6 NB. correctly gives 1r42 (6th B.n.)
> x: (%@(a&p.)) t: 8 NB. _65088267767r1952648033004 not _1r30
This behaviour is an oversight and a bug. It has now been corrected
and will be in the next release. Thanks for reporting this error.
%@(a&p.) t: 6x
1r42
%@(a&p.) t: 8x
_1r30
You need to use this with some care, viz. the argument
you give has to be extended to get an extended result.
If you give an non-extended argument and then
extend the result, you don't get the answer you want.
(It's a correct answer, just not what you want.)
x: %@(a&p.) t: 8
_65117383323r1953521499700
The explanation for this is apparent by looking at
the function:
%@(a&p.) t:
! * %@(1 1r2 1r6 1r24 1r120 1r720 1r5040 1r40320 1r362880&p.)t.
With 8 (instead of 8x), !8 gives a floating point result,
and that forces the rational result of the right tine to be
converted to floating point before the multiplication.
--------------------------------------------------------------------------------
J Forum: for information about this list, see http://www.jsoftware.com/forum.htm
From owner-jsoftware@lists.interlog.com Fri Apr 23 17:35:48 1999
From: jquint@cdcna.cdcc.com
Subject: RE: Jforum: Re: J for Windows CE
Date: Fri, 23 Apr 1999 17:15:14 -0400
charset="iso-8859-1"
I also got a Velo 500 with 24MB RAM just to run J. My experience with has
been somewhat more positive, see below...
----------
From: J. Patrick Harrington [SMTP:jph@astro.umd.edu]
Sent: Friday, April 23, 1999 12:57 PM
To: forum@JSoftware.Com
Cc: jph@astro.umd.edu
Subject: Jforum: Re: J for Windows CE
There has been some discussion of running J on
handheld/palmtop machines recently, so maybe some
would find my experience with one such device of
interest.
In 1992 I started using J on an HP 95LX and later
on an HP 100LX. These are fine little machines, and
and they seem to run forever on a pair of AA cells.
However, Iverson Software has dropped support for DOS
and my HP is stuck at J 2.02 -- circa 1994. So the code
I wrote for my other machines began to diverge from
the HP. Thus I was very interested in the announcement
I used a HP200LX to run J6.2 and I had similar difficulties because of the
divergence. Yet, I avoided must of the problems by programming exclusively
in tacit form (yes, everything can be done tacitly but that does not mean
that everything should be done tacitly). The main problem for me was the
speed and space limitations.
last year that J would be released in a version for
Windows CE. I started to look at the newer handhelds.
Eugene McDonnell posted (11 Dec 98) something about the
Philips Velo 500, and this seemed to be a good J CE
platform: not much larger than the old HP 100LX, a
reasonable battery life (though this is a problem
with all these machines), and it was supposed to be
faster than most of this class (it has a 75 MHz MIPS
processor). The screen is a nice 640 x 240 pixel size,
but since it's only 16-level greyscale (to be hot, it
has to be color now), the prices have dropped -- now
listed at $550 for the 24 MB version by Mobile Planet;
(I paid less for mine through buycomp.com, but now they
just list the 16 MB ($400), and that's not even in stock.
I think that Philips is dropping this machine.)
You should know something about the CPUs in these
machines (SH3, MIPS, now StrongArm). They are very
different from the Intel `86 chips, and thus no DOS
or Windows programs need apply. A lot of programs you
There is a WinCE DOS emulator (see http://www.pyram-id.demon.co.uk/XTCE.html
<http://www.pyram-id.demon.co.uk/XTCE.html> ) but I have not tried it yet.
might want just don't exist yet. For example, you don't
want to use the built-in Pocket Word to edit program
files! I'd like a "vi" type editor, but there is no such
beast for CE. So far I've found "Notepad" clones and a
(beta) Micro Emacs. And some things you want may *never*
exist, because CE 2.0 will soon be replaced by CE 3.0,
to be replaced by ..., etc. and leave your machine behind.
And you should know a few things about Windows CE.
It is set up to exist as an appendage of a PC running
Windows. Although CE has "terminal communication
software" built in, there is no way of using it to
talk to a non-windows machine through the serial port
without 3rd party software! What you must do is install
a package called CE Services on your PC, and then
you get the directory tree of your handheld displayed
on the PC and you can move files back and forth. This
is easy if your machine runs Windows 95, since there
are no security issues with 95 (i.e., no security).
However, my PC's are dual boot Linux and Windows NT.
Since NT has some idea of security, and the CE Services
software is trying to set up a network link with your
handheld, NT regards this with deep suspicion. Even
with Velo support explaining things they left out of the
instructions and faxing you pages of instructions with
stuff about setting up guest accounts, and what odd IP
addresses, and which service packs to install, and
reboot, and reinstall, etc. I *still* have only got my
Velo to link with one of my two machines. But once
you are connected to one PC, you can then install a
3rd party terminal program (I'm using vxHpc) and use
e.g. the kermit protocol to transfer files between your
Unix box and the handheld.
So if you only have a Windows NT PC, you are in for
a taste of Microsoft hell. If you don't have any Windows
The Windows 95 connection is easy (I did it alone at home). I did not even
tried to set the Windows NT connection at the office; the tech guys set it
up and it works fine.
machine, you may be S.O.L. While you can connect to the
internet through the built-in modem and Pocket Internet
Explorer (sort of neat), I don't think Pocket IE supports
ftp downloads. There are good 3rd party ftp programs like
"Scotty" that let you download software, but you've got
a bootstrap problem. (You can move stuff with a flash
card, but once again, the software to support the flash
has to be put into the handheld first.)
OK. So you've got your handheld CE synchronized with
your PC running CE Services. Then you just execute the
.exe file you've downloaded from Iverson Software and it
does the installation of J 402a on your handheld. (There
are two ways of installing stuff on CE machines: either
by running an .exe installation file on the PC, or by loading
a .CAB file into the handheld. When you run the .CAB file
on the handheld, it does the installation.)
After all this initial frustration, I do like running J
on the Velo. My programs all seem to run, and much faster
that the old HP 100. And the plot routines are nice. I've
just compared a routine to integrate a few stiff differential
equations on the Velo with my Sun SPARC 20. I find times
of 1.97 sec on the SPARC and 13 sec on the Velo. Another
test, inverting a matrix in rational (exact) math:
a=. x: 8 8 $ 64 ? 100
ai=. %. a
This takes 3.95 sec on the SPARC and about 30 sec on the
Velo. I say about because something weird has happened:
doing 6!:2 'ai=. %. a' on the Velo gives me results like
_27 or 30 or _31. I.e., integers, some of which are negative!
But the absolute values are about right. Is this a bug?
Anyway, I'm glad J is available on this platform, and I hope
there is enough interest to keep the product out there. I like
being able to use the LAPAC library on the Velo. (Hey, what
about the FFTW package? Will that be made available for CE?)
Yes, where is it?
My problem now is that the Velo has more features than the J
I normally use (either Sun SPARC or Linux). I also have J for
Windows 95/NT, but I almost never boot up in NT except to run
Photoshop. I really wish that J would support X11 for some
of the GUI stuff. I can see why the folks at Iverson Software
have to give 1st place to the Windows market, and linking in
with the Windows GUI makes J viable for those who write software
professionally. But there are many of us in the scientific/academic
world who use J as a research tool, and to us Linux is more
and more attractive. It would be great if J support for that
platform could extend beyond the basic J language.
J. Patrick Harrington
-------------------------------------------------------------------
P.S.
Curious things about the Velo 500: There are four
little doors on the bottom. Behind them are:
(1) the coin backup battery
(2) A space for a flash storage card. However, the Velo
does not take a standard Type II card, or one of the
"compact flash" cards, but the small (and rare)
"miniature" flash card. Although these were introduced
by Intel and others, they seemed to be available only
from Philips -- and now Philips doesn't seem to carry
them any more. They only hold 4MB, but are really cute
I got a 8MB compact flash card made by a third party (Simple Technology) and
I formatted with the Philips software available at their site with no
problems. Some people say the have gone as far as 16MB. This is where I
put J (and Lapack).
things (38 x 33 x 3.5 mm). They have no pins, but flat
contacts that press against a strange rubbery strip in
the Velo. When I got a miniature flash card, I found that
my "rubbery strip" was defective, and eventually caused
a cold boot of the machine with a total loss of all my
programs/data. I sent it back to Philips; they didn't
try to fix it, just sent me a new Velo. The new one now
works fine with the miniature cards.
(3) A slot for added RAM -- also a "miniature" 8MB DRAM thing
that brings the Velo storage/program memory up to 24 MB.
My Velo came with this installed.
(4) A slot with contacts to plug in the V-module accessory,
which makes the unit thicker, but allows you to use the
Type II PC cards. In particular, you can use the regular
San Disk flash cards. It turns out that standard flash
cards are the best solution to the problem of importing/
exporting large amounts data to/from the Velo. While this
makes the Velo less portable, the V-module can snapped on
and off as needed.
This is a nice accessory. I had a regular flash card that was, it still is,
handy for swapping files between my laptop and the HP200LX (and now from and
to the Velo 500). It also can be used to connect a monitor, etc,
Getting back to J...
I am overall a very satisfied user of J in the WinCE platform. Yet, at
least couple of things could be a lot better:
(a) I miss those HP500LX neat displays of nested items (I want, like a
good beginner according to the J documentation, to see frequently the box
representation of tacit verbs) instead of the ugly characters pretending to
form boxes. I find myself copying and pasting to pocket Word with the ISIJ
font activated to reduce the aggravation. I still do not understand why the
session font is not optional! Does anybody know?
(b) It is far from clear, at least to me, what is and what is not
supported in the WinCE version vs. the Windows 95/NT versions. Apparently
not all the scripts that can work in WinCE were included in the package due
to the typical RAM and storage limitations of the environment. But, as Eric
Iverson pointed out today, the new machines have more power and capacity!
Maybe there should be two sets of files for distribution, the current one
and a full package.
So, J for Windows CE might not be perfect, but it is close enough.
----------------------------------------------------------------------------
----
J Forum: for information about this list, see
http://www.jsoftware.com/forum.htm
--------------------------------------------------------------------------------
J Forum: for information about this list, see http://www.jsoftware.com/forum.htm
From ljdickey Fri Apr 23 18:48:55 1999
Subject: Re: Jforum: QR Decomposition (128!:0)
Piet wrote:
> Maybe I am mistaken but in the math literature, the QR decomposition
> of X refers to finding matrices Q and R such that X=QR where Q*Q=I
> and R is square triangular and Q* is the conjugate transpose of Q.
> Thus Q is orthonormal (not nec. Unitary).
To help me write about this, I want to use the functions
dot =. sum . *
where
sum =. +/
The function "dot" is ordinary matrix multiplication.
Some of what is written above is OK is long as X starts out being
real. Where I see Q*Q=I, we take that to mean I = Q dot Q, but we
should say I=Q dot |:Q and, yes, the transpose is necessary.
Here is reasoning to show that if we start with a matrix X with some
(non-real) complex numbers in it, it might be possible, but not
necessarily, to find a Q and R where Q is orthogonal.
I give two examples, X and Y, of complex matrices. For X, there is an
orthogonal matrix Q so that X = Q dot R. For the second, Y, there is
NO orthogonal matrix Q so that Y = Q dot R.
For openers, we can't have Q and R both be real, because then Q dot R
would be real, and we are talking about starting with a matrix X with
some non-real, complex numbers in it.
Now let us suppose that Q is orthogonal and R has one complex number in
some place, say it is the non-real complex number c in the i.th entry
on the main diagonal. Now think about the equation
X = Q dot R .
From the definition of matrix multiplication and the fact that R is a
diagonal matrix, we see that the i.th column of X equals c times the
i.th column of Q .
Here is an example like that. Take X to be
X =.(0.2) * 2 2 $ _4 3j3 3 4j4
You can see that the second column of X is (3 4) * 1j1, so you have a
chance of finding real matrix Q so that X = Q dot R where Q is
orthogonal (and real) and R has some complex number in the second place
on the main diagonal. In fact, for this particular example, the matrix
Q given by
(0.2) * _4 3 3 4
serves us well, and R =. 2 2 $ 1 0 0 1j1.
Now make just a one tiny change in the numbers of the second column of X.
Let us take Y to be
Y =.(0.2) * 2 2 $ _4 3j3 3 4j8
If we try to find a complex number c so that the column 3j3 4j8 matches
c*(r1, r2), where r1 and r2 are real, we very soon discover that it is
impossible. Hence there the second column of Q can not be real. You
can not fit that Y with a real Q (with r1, r2 in the second column, and
c in the second entry of the diagonal of R.)
I hope this gives some insight.
Here are some definitions.
----------------------------
The notion of a unitary matrix is a generalization of
the notion of an orthogonal matrix.
It is useful to have the identity matrix at hand. Let
I be the identity matrix that has as many rows as there
are rows in the matrix in question.
A real matrix Q is said to be orthogonal
if it has an inverse and the inverse is the transpose of Q;
that is, if *./ ,I=Q dot (|: Q)
A matrix U is said to be unitary
if it has an inverse and the inverse is its conjugate transpose;
that is, if *./ ,I=U dot (+ |: U) .
Every orthogonal matrix is unitary.
Every real unitary matrix is orthogonal.
----------------------------
Just as the notions "orthogonal" and "unitary" go together, real and complex,
so also the notions "symmetric" and "Hermitian" go together, real and complex.
A matrix S is symmetric if *./ ,(|: S)=S
A matrix H is Hermitian if *./ ,( |: H)=+ H
Every symmetric matrix is Hermitian.
Every real Hermitian matrix is symmetric.
I hope this helps.
Lee Dickey
--
Prof. Leroy J. Dickey, Faculty of Mathematics, U of Waterloo, Canada N2L 3G1
ljdickey@math.UWaterloo.CA 1-519-888-4567, ext 5559
http://math.uwaterloo.ca/~ljdickey
From owner-jsoftware@lists.interlog.com Sat Apr 24 10:36:42 1999
From: Bjorn Helgason <gosi@centrum.is>
Subject: Jforum: No Browsing after this
Date: Sat, 24 Apr 1999 14:23:33 -0000
> What is the significance of the letters "NB" in the comment verb?
It is instruction for the parser "NB." is short for "No Browsing please."
/Gosi
--------------------------------------------------------------------------------
J Forum: for information about this list, see http://www.jsoftware.com/forum.htm
From owner-jsoftware@lists.interlog.com Sat Apr 24 10:38:40 1999
From: Bjorn Helgason <gosi@centrum.is>
Subject: Jforum: Me and my Errors
Date: Sat, 24 Apr 1999 10:45:07 -0000
David Ness wrote:
> I hesitate to say it, but `You're Error' is `Your Error' in more ways
> than one...
"You're Error" was meant to be short for "You are making an Error"
"Your Error" is ok too but I liked to make it look more like "Yore Error"
Thank you for showing interest in my errors. I bet you did not hesitate
very long ;>
/Gosi
--------------------------------------------------------------------------------
J Forum: for information about this list, see http://www.jsoftware.com/forum.htm
From owner-jsoftware@lists.interlog.com Sat Apr 24 11:20:33 1999
Date: Sat, 24 Apr 1999 11:07:26 -0400
From: David Ness <DNess@home.com>
Subject: Re: Jforum: Me and my Errors
References: <01BE8E5E.4C32A4D0@geisli-8.centrum.is>
Your errors (rare as they are) are often enlightening and/or
amusing. I'd say that `You're Error' in English is short for
`You _are_ Error' as opposed to `You _are making an_ Error', or
`You _are in_ Error'. As such, I suspect that saying it might get you
into a gunfight at the `OK Corral' (or was that the `Error Corral')
Bjorn Helgason wrote:
>
> David Ness wrote:
>
<snip of my earlier>
>
--------------------------------------------------------------------------------
J Forum: for information about this list, see http://www.jsoftware.com/forum.htm
From owner-jsoftware@lists.interlog.com Sat Apr 24 14:11:40 1999
Date: Sat, 24 Apr 1999 13:56:47 -0400
From: Eric Iverson <eiverson@interlog.com>
Organization: Iverson Software Inc.
Subject: Re: Jforum: Re: J for Windows CE
References: <50C877359DC3D21192DE0008C7B1CC530E1484@winntsvr.cdcna.com>
jquint@cdcna.cdcc.com wrote:
> I am overall a very satisfied user of J in the WinCE platform. Yet, at
> least couple of things could be a lot better:
>
> (a) I miss those HP500LX neat displays of nested items
The fact that isij isn't used in the wince ijx window is a mistake on my
part. I believed the documentation that said a non-unicode font wouldn't
work on the wince system and couldn't find any wince unicode fonts that
had box draw chars. It never occured to me to just try isij. Now that I
know it works, it will be a simple change for me to use isij in the ijx
window in the next wince release.
--------------------------------------------------------------------------------
J Forum: for information about this list, see http://www.jsoftware.com/forum.htm
From owner-jsoftware@lists.interlog.com Sat Apr 24 16:40:54 1999
From: Bjorn Helgason <gosi@centrum.is>
Subject: Jforum: Re Me and My errors
Date: Sat, 24 Apr 1999 20:02:35 -0000
> David Ness wrote:
> Your errors (rare as they are) are often enlightening and/or
> amusing.
Thank you !
They are meant to be.
> I'd say that `You're Error' in English is short for
> `You _are_ Error' as opposed to `You _are making an_ Error', or
> `You _are in_ Error'.
When you shorten something and someone else makes it longer again
there are bound to be problems.
IBM has an explanation for unexplanable problems.
It could be because there is shit behind the keyboard.
Which side of the keys does not really matter does it?
/Gosi
--------------------------------------------------------------------------------
J Forum: for information about this list, see http://www.jsoftware.com/forum.htm
From owner-jsoftware@lists.interlog.com Sat Apr 24 21:11:32 1999
From: Ajith Prasad <burap@nus.edu.sg>
Subject: RE: Jforum: Re: J for Windows CE
Date: Sun, 25 Apr 1999 08:52:05 +0800
charset="windows-1252"
I always thought "ISIJ" stood for Iverson Software Inc - J" font. Is that
correct or am I forcing an interpretation on a string of words that have no
specifc meaning or some other meaning?
-----Original Message-----
From: Eric Iverson
Sent: 4/25/99 1:56 AM
Subject: Re: Jforum: Re: J for Windows CE
jquint@cdcna.cdcc.com wrote:
> I am overall a very satisfied user of J in the WinCE platform. Yet,
at
> least couple of things could be a lot better:
>
> (a) I miss those HP500LX neat displays of nested items
The fact that isij isn't used in the wince ijx window is a mistake on my
part. I believed the documentation that said a non-unicode font wouldn't
work on the wince system and couldn't find any wince unicode fonts that
had box draw chars. It never occured to me to just try isij. Now that I
know it works, it will be a simple change for me to use isij in the ijx
window in the next wince release.
------------------------------------------------------------------------
--------
J Forum: for information about this list, see
http://www.jsoftware.com/forum.htm
--------------------------------------------------------------------------------
J Forum: for information about this list, see http://www.jsoftware.com/forum.htm
From owner-jsoftware@lists.interlog.com Sat Apr 24 23:31:32 1999
Date: Sat, 24 Apr 1999 23:18:46 -0400
From: Eric Iverson <eiverson@interlog.com>
Organization: Iverson Software Inc.
Subject: Re: Jforum: Re: J for Windows CE
References: <A12A4ACD69B4D011BB960020AFFBF1080123C242@exs02.ex.nus.edu.sg>
Ajith Prasad wrote:
>
> I always thought "ISIJ" stood for Iverson Software Inc - J" font.
That's right.
--------------------------------------------------------------------------------
J Forum: for information about this list, see http://www.jsoftware.com/forum.htm
From owner-jsoftware@lists.interlog.com Sun Apr 25 11:58:02 1999
Date: Sun, 25 Apr 1999 11:29:59 -0400 (EDT)
From: "Brian M. Schott" <dscbms@panther.Gsu.EDU>
Subject: Jforum: Re: J for Windows CE
In-Reply-To: <37228996.511B@interlog.com>
Yesterday I spent approximately 3 hours with two very
patient salesmen at Circuit City trying to launch the MIPS
version of J402a onto a Cassiopeia E-15. The primary goal
was not achieved but I learned some things that may be of
interest to others who desire the same result.
The E-15 (razor) manual is informative, but does not begin
to answer all the questions associated with copying the
software from the desktop computer and installing the .exe
file on the palmheld. First a utility called Windows CE
Services must be installed on the desktop computer and the
COM1 port must be configured to match the palmheld's serial
port. The configuration must be performed with a degree of
knowledge of navigating the Windows 95/98 control panels and
operating system to accomplish 19200 baud -- not a trivial
task for a nonWindows user.
Somewhat unintuitively, J402a must be installed first on the
desktop computer. This involves choices from among
unzipping, Windows unzipping, and others on a dialog window.
The palmheld has almost no file management ability within
its own user system. Windows CE Services on the desktop
computer must be used for most of the file manipulation on
the palmheld computer AND for transferring files between the
desktop and palmheld. (Btw, it may even be true that Windows
CE Services is needed to remove unneeded files from the
palmheld. I cannot be sure because this was never
attempted.) The download of the files from the desktop
computer to the palmheld takes approximately 15 minutes at
19200 baud. This dl included all of the files including the
executables and the program files. Our first download was
done mistakenly in a "synchronization" mode resulting in an
inability to access the dl'ed files. A second download was
more successful (but see way below). After much trial and
error one of the technical guys at the store was able to
produce a shortcut on the palmheld. (Remember, most of this
work has to be done on the desktop even though the results
appear on the palmheld. And some of the instructions were
unclear about where the next step was to performed -- on the
desktop or the palmheld. One particularly unclear
instruction required that the Explorer Window on the
palmheld to be used, but we could never find such a window
on the palmheld. Another unclear instruction recommended
that the shortcut icon be "sent to" somewhere (-- the
Explorer Window mentioned above -- when there was no "send
to" option available with the right click option.) But J
would not launch: an error message alert indicated that a
necessary program was not accessible. That was the most
encouraging result we could achieve; we were unable to
launch J402a on the palmheld.
The folks at Circuit City were extremely patient and
professional. They seemed quite experienced in using both
the Windows computer and handheld computers.
There is one small uncertainty; at the very end of the final
download, a customer in the store walked up and picked up
the palmheld computer, aborting the download. The
connection was remade as if the download was complete,
because the technical guy was sure that the important files
had been downloaded. But by this time I was rather
disillusioned about the limited file management features of
the operating system and the difficulty I would have
transferring results to my Macintosh to get hardcopy. So I
may be better off that the installation did not work -- I do
not know that would be satisfied with the razor's
limitations and I would have felt obligated to buy the unit
after they had spent so much time with me.
(B=)
--------------------------------------------------------------------------------
J Forum: for information about this list, see http://www.jsoftware.com/forum.htm
From owner-jsoftware@lists.interlog.com Sun Apr 25 12:35:03 1999
From: "Chris Burke" <cdburke@interlog.com>
References: <Pine.GSO.3.95.990425112720.15136A-100000@panther.Gsu.EDU>
Subject: Re: Jforum: Re: J for Windows CE
Date: Sun, 25 Apr 1999 12:22:53 -0400
charset="iso-8859-1"
----- Original Message -----
From: Brian M. Schott <dscbms@panther.Gsu.EDU>
Sent: Sunday, April 25, 1999 11:29 AM
Subject: Jforum: Re: J for Windows CE
> Yesterday I spent approximately 3 hours with two very
> patient salesmen at Circuit City trying to launch the MIPS
> version of J402a onto a Cassiopeia E-15. The primary goal
> was not achieved but I learned some things that may be of
> interest to others who desire the same result.
J requires a WinCE machine with a keyboard. See recent msgs to the
forum.
> The E-15 (razor) manual is informative, but does not begin
> to answer all the questions associated with copying the
> software from the desktop computer and installing the .exe
> file on the palmheld. First a utility called Windows CE
> Services must be installed on the desktop computer and the
> COM1 port must be configured to match the palmheld's serial
> port. The configuration must be performed with a degree of
> knowledge of navigating the Windows 95/98 control panels and
> operating system to accomplish 19200 baud -- not a trivial
> task for a nonWindows user.
Not trivial for a Windows user, either. Configuring WinCE is painfull!
Also, you can use a faster baud rate - up to 115K, and any free COM
port.
> Somewhat unintuitively, J402a must be installed first on the
> desktop computer. This involves choices from among
> unzipping, Windows unzipping, and others on a dialog window.
Not installed, but unzipped. You unzip the files on the PC, then copy
them to the WinCE machine. Since WinCE requires a PC link, this is a
reasonable way to distribute WinCE software.
--------------------------------------------------------------------------------
J Forum: for information about this list, see http://www.jsoftware.com/forum.htm
From owner-jsoftware@lists.interlog.com Sun Apr 25 21:44:14 1999
From: "Seymour Glass" <glasss@mindspring.com>
Subject: Jforum: Historical Tidbit
Date: Sun, 25 Apr 1999 21:17:45 -0400
charset="iso-8859-1"
In-Reply-To: <37228996.511B@interlog.com>
Importance: Normal
I have just started reading David Kahn's _The Codebreakers_.
It begins with a fascinating account of the American
efforts to decode the Japanese diplomatic ciphers during
World War II.
These ciphers were created by an enciphering machine,
which of course is nothing more than a mechanism for
converting plain language into gobbledygook.
The Americans called the machine PURPLE, but the Japanese
had their own name for it: they called it "J".
Henry Rich
--------------------------------------------------------------------------------
J Forum: for information about this list, see http://www.jsoftware.com/forum.htm
From owner-jsoftware@lists.interlog.com Mon Apr 26 02:36:45 1999
From: "Roger Hui" <RHui@interlog.com>
Subject: Jforum: Re: QR Decomposition (128!:0)
Date: Sun, 25 Apr 1999 23:24:20 -0400
charset="iso-8859-1"
Piet de Jong writes on Friday, April 23:
> According to the definitions I see in the various texts, a Hermitian matrix is one
> which equals its conjugate transpose (ie symmetric in the real case).
>
> The Q matrix in the QR decomposition provided by 128!:0 is not generally Hermitian.
> (I'm using the implementation in J3x). However it does appear to be orthonormal in for
> example the case where the matrix operated on is of full column rank.
>
> Maybe I am mistaken but in the math literature, the QR decomposition of X refers
> to finding matrices Q and R such that X=QR where Q*Q=I and R is square triangular
> and Q* is the conjugate transpose of Q. Thus Q is orthonormal (not nec. Hermitian).
>
> Generally QR decompositions are constructed for matrices of full column rank and
> hence not necessarily square. (as implied by Roger's nonsingularity remark).
>
> The J3x manual states that the domain of 128!:0 is the same as that of %. The domain
> of %. seems to include non square matrices. This appears to be at variance with Roger's remarks.
I see that I called the matrix Q Hermitian when I meant to say that it is orthonormal.
This bug will be corrected in the next version of the dictionary. Thank you for
pointing out this error.
The domain of %. is exactly the domain of 128!:0, and this domain is the set
of non-singular complex matrices. By non-singular, I mean the columns are
linearly independent ("full column rank"), and thus tall matrices as well as
square matrices (>:/$M) can be non-singular.
--------------------------------------------------------------------------------
J Forum: for information about this list, see http://www.jsoftware.com/forum.htm
From owner-jsoftware@lists.interlog.com Mon Apr 26 10:12:01 1999
Date: Mon, 26 Apr 1999 09:57:14 -0400
From: Murray Eisenberg <murray@math.umass.edu>
Organization: Mathematics & Statistics, Univ. of Mass./Amherst
Subject: Jforum: Current directory on startup
My J is in directory:
e:\langauges\j\j402
When I start J with, apparently, nothing else than the standard
profile.isj loaded and use the File | Open dialog to open a script, the
Open window always starts in directory
f:\user
-- on a different logical drive!
What causes this peculiar behavior, and how do I fix it?
--
Murray Eisenberg murray@math.umass.edu
Mathematics & Statistics Dept. phone 413 549-1020 (H)
Univ. of Massachusetts 413 545-2859 (W)
Amherst, MA 01003-4515
--------------------------------------------------------------------------------
J Forum: for information about this list, see http://www.jsoftware.com/forum.htm
From owner-jsoftware@lists.interlog.com Mon Apr 26 10:17:38 1999
Date: Mon, 26 Apr 1999 09:40:38 -0400
From: Murray Eisenberg <murray@math.umass.edu>
Organization: Mathematics & Statistics, Univ. of Mass./Amherst
Subject: Re: Jforum: Re: QR Decomposition (128!:0)
References: <002c01be8f94$6879d820$9eb0ba89@f3nbp>
Roger,
Do you mean "orthogonal" rather than "orthonormal"? Peculiarly, an
"orthogonal" matrix is one whose rows (or, equivalently, columns) form
an orthoNORMAL set of vectors.
Roger Hui wrote:
>
> I see that I called the matrix Q Hermitian when I meant to say that it is orthonormal.
--
Murray Eisenberg murray@math.umass.edu
Mathematics & Statistics Dept. phone 413 549-1020 (H)
Univ. of Massachusetts 413 545-2859 (W)
Amherst, MA 01003-4515
--------------------------------------------------------------------------------
J Forum: for information about this list, see http://www.jsoftware.com/forum.htm
From owner-jsoftware@lists.interlog.com Mon Apr 26 12:12:11 1999
Date: Mon, 26 Apr 1999 11:28:19 -0400
From: Eric Iverson <eiverson@interlog.com>
Organization: Iverson Software Inc.
Subject: Re: Jforum: Current directory on startup
References: <372470BA.31145619@math.umass.edu>
Murray Eisenberg wrote:
> My J is in directory:
> e:\langauges\j\j402
> When I start J with, apparently, nothing else than the standard
> profile.isj loaded and use the File | Open dialog to open a script, the
> Open window always starts in directory
> f:\user
> What causes this peculiar behavior, and how do I fix it?
I don't see how this can happen. The first file dialog should always
start in the J directory.
Does (did) it always start in f:\user?
Is f:\user a valid directory?
If you use file open to open a file in another directory, does the next
use properly remember that latest use.
--------------------------------------------------------------------------------
J Forum: for information about this list, see http://www.jsoftware.com/forum.htm
From owner-jsoftware@lists.interlog.com Mon Apr 26 12:34:48 1999
From: "Roger Hui" <RHui@interlog.com>
Subject: Jforum: Re: QR Decomposition (128!:0)
Date: Mon, 26 Apr 1999 09:16:53 -0400
charset="iso-8859-1"
Murray Eisenberg writes on Monday, April 26:
>Do you mean "orthogonal" rather than "orthonormal"? Peculiarly, an
>"orthogonal" matrix is one whose rows (or, equivalently, columns) form
>an orthoNORMAL set of vectors.
I mean the following: If X is a matrix in the domain of the
monad %. (i.e. if X is a non-singular complex matrix), and:
x=: +/ . *
c=: {:$X
'Q R' =: 128!:0 X
Then the following assertions hold:
1e_15 > >./ , | X - Q x R
1
($Q) -: $X
1
1e_15 > >./ , | (=i.c) - (+|:Q) x Q
1
($R) -: 2$c
1
(0~:R) <: <:/~ i.c
1 1 1 1 1 1 1
1 1 1 1 1 1 1
1 1 1 1 1 1 1
1 1 1 1 1 1 1
1 1 1 1 1 1 1
1 1 1 1 1 1 1
1 1 1 1 1 1 1
0 ~: (<0 1)|:R
1 1 1 1 1 1 1
--------------------------------------------------------------------------------
J Forum: for information about this list, see http://www.jsoftware.com/forum.htm
From owner-jsoftware@lists.interlog.com Mon Apr 26 13:44:32 1999
From: jquint@cdcna.cdcc.com
Subject: RE: Jforum: Re: J for Windows CE
Date: Mon, 26 Apr 1999 13:27:38 -0400
charset="iso-8859-1"
----- Original Message -----
----------
From: Eric Iverson [SMTP:eiverson@interlog.com]
<mailto:[SMTP:eiverson@interlog.com]>
Sent: Saturday, April 24, 1999 1:57 PM
To: forum@JSoftware.Com <mailto:forum@JSoftware.Com>
Subject: Re: Jforum: Re: J for Windows CE
jquint@cdcna.cdcc.com <mailto:jquint@cdcna.cdcc.com> wrote:
> I am overall a very satisfied user of J in the WinCE platform.
Yet, at
> least couple of things could be a lot better:
>
> (a) I miss those HP500LX neat displays of nested items
The fact that isij isn't used in the wince ijx window is a mistake on my
part. I believed the documentation that said a non-unicode font wouldn't
work on the wince system and couldn't find any wince unicode fonts that had
box draw chars. It never occured to me to just try isij. Now that I know it
works, it will be a simple change for me to use isij in the ijx window in
the next wince release.
----------------------------------------------------------------------------
----
J Forum: for information about this list, see
http://www.jsoftware.com/forum.htm <http://www.jsoftware.com/forum.htm>
Thanks Eric for responding so quickly. I will be looking forward for the
next version...
I also saw that my other comments were imbedded in the text of the J.
Patrick Harrington's original message and were most likely unnoticed by most
readers (the indentation looked fine in MS Exchange though). For what is
worth, here they are again:
On keeping compatible J code for old and new versions...
I used a HP200LX to run J6.2 and I had similar difficulties because of the
divergence. Yet, I avoided must of the problems by programming exclusively
in tacit form (yes, everything can be done tacitly but that does not mean
that everything should be done tacitly). The main problem for me was the
speed and space limitations.
On not being able to run old DOS programs under WinCE...
There is a WinCE DOS emulator (see http://www.pyram-id.demon.co.uk/XTCE.html
<http://www.pyram-id.demon.co.uk/XTCE.html> ) but I have not tried it yet.
On the difficulties of transferring files from and to a Windows NT machine.
The Windows 95 connection is easy (I did it alone at home). I did not even
tried to set the Windows NT connection at the office; the tech guys set it
up and it works fine.
On asking about the FFTW package for J WinCE version...
Yes, where is it?
On the 4MB limitation of the Velo 500 Compact Flash Card...
I got a 8MB compact flash card made by a third party (Simple Technology) and
I formatted with the Philips software available at their site with no
problems. Some people say the have gone as far as 16MB. This is where I
put J (and Lapack).
--------------------------------------------------------------------------------
J Forum: for information about this list, see http://www.jsoftware.com/forum.htm
From owner-jsoftware@lists.interlog.com Mon Apr 26 16:39:43 1999
Date: 26 Apr 99 12:25:11 -0800
Subject: Re: Jforum: Re: QR Decomposition (128!:0)
From: "Piet de Jong" <piet.dejong@commerce.ubc.ca>
X-Fontfamily: Geneva
X-Fontsize: 12
Content-Transfer-Encoding: quoted-printable
The terminology "nonsingular" as any matrix with full column rank is,
I think, nonstandard.
The issue remains as to what, if anything, can be usefully made of
the output of the QR decomposition foreign conjunction (128!:0) in
the case where the X matrix is not of full column rank ("singular").
In J3 on a Mac the Q matrix is no longer orthonormal. However Q x R
does return the original matrix X, at least in the cases I have tried.
Can the output Q and R be pieced together to a provide a valid QR
decomposition in the non full column rank case? That is what is the
definition of the Q and R matrices in this case? In J3 there is no
error reported if X is of less than full column rank. Maybe this has
changed in J4.
--------------------------------------------------------------------------------
J Forum: for information about this list, see http://www.jsoftware.com/forum.htm
From owner-jsoftware@lists.interlog.com Mon Apr 26 18:19:03 1999
From: Martin Barghoorn <barg@cs.tu-berlin.de>
Date: Mon, 26 Apr 1999 18:36:00 +0200 (MET DST)
Subject: Re: Jforum: Monadic Atop and Bond are different?APL-Journal
X-Sun-Charset: US-ASCII
>
> Nope, because u/ leads to the _dyad_ meaning of u, not the mondic one.
> That is:
>
> > l=: *:&+
> > l/ 1 2 3 4
>
> is equivalent to:
> (1 *:&+ (2 *:&+ (3 *:&+ 4)))
>
> which tries to NAND the complex conjugates of items and/or intermediate
> results.
>
> Martin
>
Hallo, lieber Namesvetter im fernen Niedersachsen,
Du bist immernoch der world-wide-aktivste, wie
ich immer wieder hocherfreut feststelle.
Doch wir muessen unser gutes altes APL-Journal
zu neuem (J) Leben erwecken, deshalb sollte hier Dein toller
J-Vortrag (oder Teile davon) von Bingen moeglichst bald eintreffen.
Sonst wird das Journal zu APL-lastig.
Also bitte, spute Dich
Dein (auch) Martin
--------------------------------------------------------------------------------
J Forum: for information about this list, see http://www.jsoftware.com/forum.htm
From owner-jsoftware@lists.interlog.com Mon Apr 26 21:17:23 1999
Date: Mon, 26 Apr 1999 20:59:29 -0400
From: Murray Eisenberg <murray@math.umass.edu>
Organization: Mathematics & Statistics, Univ. of Mass./Amherst
Subject: Re: Jforum: Current directory on startup
References: <372470BA.31145619@math.umass.edu> <37248613.1BD6@interlog.com>
I don't understand how it can happen, either!!
The path f:\user IS a legitimate directory (on a different logical
drive). And J 4.02a ALWAYS starts there when I first use File | Open in
a session.
If I use the File | Open dialog to open a file in, say,
e:\language\j\j402\user [subdirectory of my main e:\language\j\j402
directory that contains j.exe], then during the SAME J session the File
| Open dialog does remember that as the current directory.
But next time I start J, it's back to f:\user as the current directory
when I first use File | Open again.
This happens even if I start J directoy by double-clicking the file
j.exe in e:\language\j\j402. I tried that in desperation even though I
did carefully check that my Start menu shortcut to J had the entry to
start in e:\language\j\j402.
Below is verbatim the file profile.ijs (I reverted to the distributed
profile after being bewildered by the peculiar behavior.) And, just to
make sure that nothing of mine was being loaded, I checked that the only
locales defined are 'base', 'j', and 'z'; and I checked all the verbs
and nouns defined at startup, and they are the ones expected with only
the distributed profile being executed.
This behavior has certainly been present since I upgraded to Windows 98
in late January/early February of this year; I believe it was present
earlier, too, but I would not bet a large amount of money on it.
~~~~~~~~~ profile.ijs ~~~~~~
NB. standard J profile
NB.
NB. To set up your own profile, first make a copy of this file,
NB. then change the J startup parameter to point to the new file.
NB.
NB. 1. CONFIGFILE_j_ points to the configuration file. You can
NB. create your own file, and do not need to change the
NB. original config file (system\extras\config\config.ijs).
NB.
NB. The config file can be customized within J by running
NB. menu item: Edit|Configure...
NB.
NB. 2. The script boot.ijs then loads and runs the config file.
NB. You do not need to change this line.
NB.
NB. 3. After the boot, you can add your own commands to
NB. further customize the start up. By default there is a
NB. command to display the J version number.
NB. =========================================================
NB. set name of config file:
CONFIGFILE_j_=: 'system\extras\config\config.ijs'
NB. =========================================================
NB. boot up J:
0!:0 <'system\extras\util\boot.ijs'
NB. =========================================================
NB. add your own commands here
Eric Iverson wrote:
>
> Murray Eisenberg wrote:
> > My J is in directory:
> > e:\langauges\j\j402
> > When I start J with, apparently, nothing else than the standard
> > profile.isj loaded and use the File | Open dialog to open a script, the
> > Open window always starts in directory
> > f:\user
> > What causes this peculiar behavior, and how do I fix it?
>
> I don't see how this can happen. The first file dialog should always
> start in the J directory.
>
> Does (did) it always start in f:\user?
>
> Is f:\user a valid directory?
>
> If you use file open to open a file in another directory, does the next
> use properly remember that latest use.
--
Murray Eisenberg murray@math.umass.edu
Mathematics & Statistics Dept. phone 413 549-1020 (H)
Univ. of Massachusetts 413 545-2859 (W)
Amherst, MA 01003-4515
--------------------------------------------------------------------------------
J Forum: for information about this list, see http://www.jsoftware.com/forum.htm
From owner-jsoftware@lists.interlog.com Mon Apr 26 22:02:42 1999
From: "Roger Hui" <RHui@interlog.com>
Subject: Jforum: Re: QR Decomposition (128!:0)
Date: Mon, 26 Apr 1999 18:46:43 -0400
charset="iso-8859-1"
> The terminology "nonsingular" as any matrix with full column rank is,
> I think, nonstandard.
Well, in this sense, %. is also non-standard.
> The issue remains as to what, if anything, can be usefully made of
> the output of the QR decomposition foreign conjunction (128!:0)
> in the case where the X matrix is not of full column rank ("singular").
> In J3 on a Mac the Q matrix is no longer orthonormal. However
> Q x R does return the original matrix X, at least in the cases I have tried.
> Can the output Q and R be pieced together to a provide a valid QR
> decomposition in the non full column rank case? That is what is the
> definition of the Q and R matrices in this case? In J3 there is no error
> reported if X is of less than full column rank. Maybe this has changed in J4.
The specification for 128!:0 is that the argument must be
non-singular. If you want something different you are
welcome to figure it out.
--------------------------------------------------------------------------------
J Forum: for information about this list, see http://www.jsoftware.com/forum.htm
From owner-jsoftware@lists.interlog.com Mon Apr 26 23:31:59 1999
Date: Mon, 26 Apr 1999 23:16:03 -0400
From: Eric Iverson <eiverson@interlog.com>
Organization: Iverson Software Inc.
Subject: Re: Jforum: Current directory on startup
References: <372470BA.31145619@math.umass.edu> <37248613.1BD6@interlog.com> <37250BF1.7D02AFB0@math.umass.edu>
Murray Eisenberg wrote:
>
> I don't understand how it can happen, either!!
> ...
> This behavior has certainly been present since I upgraded to Windows 98
> in late January/early February of this year; I believe it was present
> earlier, too, but I would not bet a large amount of money on it.
It has nothing to do with how J starts or what profile does. The first
use of the open file dialog uses an empty string as the parameter to
initialize the path. Thereafter it uses whatever path you set in the
dialog. The empty string is defined to mean the current directory, which
J has set to be the J directory on startup. I wonder if this is a small
change in the file open dialog in win98. I don't have immediate access
to a win98 machine, but will check it tomorrow.
--------------------------------------------------------------------------------
J Forum: for information about this list, see http://www.jsoftware.com/forum.htm
From owner-jsoftware@lists.interlog.com Tue Apr 27 00:07:01 1999
Date: Mon, 26 Apr 1999 20:53:12 -0700
Subject: Re: Jforum: Re: QR Decomposition (128!:0)
From: "Piet de Jong" <piet.dejong@commerce.ubc.ca>
Content-type: text/plain; charset="US-ASCII"
----------
>From: "Roger Hui" <RHui@interlog.com>
>To: "J Forum" <Forum@JSoftware.Com>
>Subject: Jforum: Re: QR Decomposition (128!:0)
>Date: Mon, Apr 26, 1999, 3:46 PM
>
>> The terminology "nonsingular" as any matrix with full column rank is,
>> I think, nonstandard.
>
>Well, in this sense, %. is also non-standard.
>
The difference is that everybody uses "nonsingular" in a given sense which
you are trying to change, while nobody else, to my knowledge, uses %.
--------------------------------------------------------------------------------
J Forum: for information about this list, see http://www.jsoftware.com/forum.htm
From owner-jsoftware@lists.interlog.com Tue Apr 27 02:48:14 1999
From: "Roger Hui" <RHui@interlog.com>
Subject: Jforum: Re: QR Decomposition (128!:0)
Date: Mon, 26 Apr 1999 21:33:28 -0400
charset="iso-8859-1"
Piet de Jong writes on Monday, April 26:
>>> The terminology "nonsingular" as any matrix with full column rank is,
>>> I think, nonstandard.
>>
>>Well, in this sense, %. is also non-standard.
>>
>
>The difference is that everybody uses "nonsingular" in a given sense which
>you are trying to change, while nobody else, to my knowledge, uses %.
I disagree. In the "standard" usage, matrix inverse applies only
to square matrices, and invertible matrices are non-singular.
In the J (also APL) usage, matrix inverse has been extended
to apply to tall matrices (as well as applying to square matrices),
and invertible matrices are, I contend, non-singular.
--------------------------------------------------------------------------------
J Forum: for information about this list, see http://www.jsoftware.com/forum.htm
From owner-jsoftware@lists.interlog.com Tue Apr 27 02:50:26 1999
Date: Tue, 27 Apr 1999 07:36:22 +0100
From: Mike Day <m020u000@cwcom.net>
Subject: Re: Jforum: Historical Tidbit
References: <000001be8f82$9754f380$6d89f7a5@seymourg>
Not J until you program it, but if you're into cryptography
and the first computer - as some claim - visit
BletchleyPark@Cranfield.ac.uk
I recommend a physical visit too - they have open days every
fortnight or so.
Mike Day
Seymour Glass wrote:
>
> I have just started reading David Kahn's _The Codebreakers_.
> It begins with a fascinating account of the American
> efforts to decode the Japanese diplomatic ciphers during
> World War II.
>
> These ciphers were created by an enciphering machine,
> which of course is nothing more than a mechanism for
> converting plain language into gobbledygook.
>
> The Americans called the machine PURPLE, but the Japanese
> had their own name for it: they called it "J".
>
> Henry Rich
>
> --------------------------------------------------------------------------------
> J Forum: for information about this list, see http://www.jsoftware.com/forum.htm
--------------------------------------------------------------------------------
J Forum: for information about this list, see http://www.jsoftware.com/forum.htm
From owner-jsoftware@lists.interlog.com Tue Apr 27 03:47:26 1999
Date: Tue, 27 Apr 1999 03:20:51 -0400
From: Fraser Jackson <Fraser_Jackson@compuserve.com>
Subject: Jforum: Mapped Files - a bug?
Content-Disposition: inline
Content-Transfer-Encoding: 8bit
X-MIME-Autoconverted: from quoted-printable to 8bit by plus.interlog.com id DAA22960
I have been using the mapped files facility to look at a large data set -
about 2.5 million records.
Suppose we have two mapped files Temp1 and Year1.
Year1 is a random sequence of integers from the set i.100
$~. Year1
100
Temp1 =. 95 - Year1
$~. Temp1
311
Where have the additional values in the nub of Temp1 come from?
When we look at the values a number are of the order of several thousand.
How could they get there?
Fraser
--------------------------------------------------------------------------------
J Forum: for information about this list, see http://www.jsoftware.com/forum.htm
From owner-jsoftware@lists.interlog.com Tue Apr 27 08:14:12 1999
Date: Tue, 27 Apr 1999 07:47:48 -0400
From: Eric Iverson <eiverson@interlog.com>
Organization: Iverson Software Inc.
Subject: Re: Jforum: Mapped Files - a bug?
References: <199904270320_MC2-738A-6E9D@compuserve.com>
to Fraser Jackson:
I'm puzzled by your Year1 and Temp1 problem. Could you give more
information.
What is the file format (jmf files with headers, or just data)?
What is the shape of the nouns?
What is the datatype of the nouns.
What is Temp1 (shape, datatype, contents) before the assignment?
What is the result of the following:
abc=: 95 - Year1
$~. abc
What is the result of the following:
abc=: 95 - Year1
$~. abc
Temp1=: abc
$~. Temp1
--------------------------------------------------------------------------------
J Forum: for information about this list, see http://www.jsoftware.com/forum.htm
From owner-jsoftware@lists.interlog.com Tue Apr 27 08:15:11 1999
Date: Tue, 27 Apr 1999 08:00:27 -0400
From: Eric Iverson <eiverson@interlog.com>
Organization: Iverson Software Inc.
Subject: Re: Jforum: Current directory on startup
References: <372470BA.31145619@math.umass.edu> <37248613.1BD6@interlog.com> <37250BF1.7D02AFB0@math.umass.edu>
to Murray Eisenberg:
The following is from win98 experiments done by Chris:
My win98 J file|open starts in the J directory.
However, if I delete all files in the main J directory except for
j.exe j.dll and j.lic, then file|open starts in My Documents (which
presumably Murray has remapped). But a directory listing 1!:0 '*.*'
shows the correct J directory.
If I copy a file such as system\main\dates.ijs into the main J
directory, file|open starts in J.
If instead, I copy a file such as jexe.tlb into the main J directory,
file|open starts in My Documents.
The system seems to be looking for files other than J system files in
the J directory. If it finds one, file|open starts in J, if not, it
starts in My Documents.
Win98 has changed the file open dialog. If a path is not specified (the
case when it is first used in J) and it doesn't find 'document like
files' in the current directory, it uses the 'document directory'
instead of current.
A workaround for now is to put a file that looks like it could be a
document file in the J directory so that the dialog will use the current
directory.
I'll fix this in the next release by initilizing the empty string to
explicitly reference the J directory.
--------------------------------------------------------------------------------
J Forum: for information about this list, see http://www.jsoftware.com/forum.htm
From owner-jsoftware@lists.interlog.com Tue Apr 27 08:24:54 1999
Date: Tue, 27 Apr 1999 08:07:37 -0400
From: Murray Eisenberg <murray@math.umass.edu>
Organization: Mathematics & Statistics, Univ. of Mass./Amherst
Subject: Re: Jforum: Current directory on startup
References: <372470BA.31145619@math.umass.edu> <37248613.1BD6@interlog.com> <37250BF1.7D02AFB0@math.umass.edu> <37252BF3.2B89@interlog.com>
Eric Iverson wrote:
>
> ... I wonder if this is a small
> change in the file open dialog in win98...
This is quite possible.
My bit of evidence: Right away when I replaced Win 95 with Win 98,
Wolfram Research's Mathematica ver. 3.0.1 (the current version) began
strange behavior (trouble printing to an HP PCL printer and causing Win
98 to hang upon exit from Mathematica). Wolfram supplied a patch, so
that it became feasible to use Mathematica under Win 98. But then a
open-file problem appeared: whereas one could open a file in the
still-closed Mathematica by double-clicking on the file in Explorer,
once Mathematica was open any subsequent attempt to open another file
the same way simply did nothing.
--
Murray Eisenberg murray@math.umass.edu
Mathematics & Statistics Dept. phone 413 549-1020 (H)
Univ. of Massachusetts 413 545-2859 (W)
Amherst, MA 01003-4515
--------------------------------------------------------------------------------
J Forum: for information about this list, see http://www.jsoftware.com/forum.htm
From owner-jsoftware@lists.interlog.com Tue Apr 27 10:43:16 1999
Date: Tue, 27 Apr 1999 10:31:54 -0400
From: Daniel Torres <torres@SAmerica.com>
Organization: Leo Burnett Worldwide, Latin America Headquarters
Subject: Re: Jforum: Mapped Files - a bug?
References: <199904270320_MC2-738A-6E9D@compuserve.com>
Content-Type: multipart/mixed;
boundary="------------D7CEAD563776E9313FF795C1"
This is a multi-part message in MIME format.
--------------D7CEAD563776E9313FF795C1
I am just beginning to use mapped files, so take the following with a *hill* of
salt:
I assume Temp1 had its size defined before the 95-Year1 assignment. If you did
not clear Temp1 before the assignment, my guess is that you are looking at the
100 nub values of Year1 AND the nub values of what was originally in Temp1
(excluding the first 100 values), when it was defined.
Daniel
Fraser Jackson wrote:
>
> I have been using the mapped files facility to look at a large data set -
> about 2.5 million records.
>
> Suppose we have two mapped files Temp1 and Year1.
>
> Year1 is a random sequence of integers from the set i.100
>
> $~. Year1
>
> 100
>
> Temp1 =. 95 - Year1
>
> $~. Temp1
>
> 311
>
> Where have the additional values in the nub of Temp1 come from?
>
> When we look at the values a number are of the order of several thousand.
> How could they get there?
>
> Fraser
--------------D7CEAD563776E9313FF795C1
name="torres.vcf"
Content-Description: Card for Daniel Torres
Content-Disposition: attachment;
filename="torres.vcf"
begin:vcard
n:Torres;Daniel
tel;fax:1-978.383-5817
tel;home:1-305.441-0369
tel;work:1-305.461-6829
x-mozilla-html:TRUE
adr:;;600 Biltmore Way APT 412;Coral Gables;FL;33134-7529;USA
version:2.1
email;internet:torres@SAmerica.com
fn:Daniel Torres
end:vcard
--------------D7CEAD563776E9313FF795C1--
--------------------------------------------------------------------------------
J Forum: for information about this list, see http://www.jsoftware.com/forum.htm
From owner-jsoftware@lists.interlog.com Tue Apr 27 14:57:40 1999
Date: Tue, 27 Apr 1999 14:36:12 -0400
From: Fraser Jackson <Fraser_Jackson@compuserve.com>
Subject: Re: Jforum: Mapped Files - a bug?
Content-Disposition: inline
Content-Transfer-Encoding: 8bit
X-MIME-Autoconverted: from quoted-printable to 8bit by plus.interlog.com id OAA14817
Eric wrote:
>>
I'm puzzled by your Year1 and Temp1 problem. Could you give more
information.
What is the result of the following:
abc=: 95 - Yeari
$~. abc
What is the result of the following:
abc=: 95 - Yeari
$~. abc
Temp1=: abc
$~. Temp1
>>
On starting J I used a script with map_jmf_ to map all the files. (They
were created earlier with commands of the form
create_jmf_ filename;byte_size. They were then mapped, and values
assigned - all in much earlier sessions. The files have 200 bytes more
than required for the data.)
I executed the above commands in Eric's note and they gave the expected
results.
I then executed the following script without any intervening commands
Temp1 =: 95 - Yeari
Temp2 =: Monthi <3
Temp3 =: (Monthi=3) *. Dayi<7
Temp4 =: Temp2 +. Temp3
Var1 =: Temp1 + Temp4
All variables are mapped files. I had separated the steps to try and
identify the problem.
After execution of the script Temp1 has a different nub inclusive of over
two hundred additional items. This gives a nub for Var1 which consists of
far more items than the nub of Yeari.
I get the following
e =: [:$~.
e Temp1
312
e Temp2
2
e Temp3
2
e Temp4
2
e Var1
1137
I can then re-execute the commands in Eric's note and again I get the
expected results.
It seems something strange is happening in the script execution.
Here is copy of a session which shows that execution of the commands from
a script gives a different result to execution of the commands from the
session manager:
abc =: 95-Temp1
e abc
100
Temp1 =: abc
e Temp1
100
load'c:\j402a\temp\eric.ijs'
e abc
100
e Temp1
307
The script eric.ijs is
NB. Eric's test script
abc =: 95 - Yeari
Temp1 =: abc
Any suggestions since I want to used mapped files in scripts.
Fraser
--------------------------------------------------------------------------------
J Forum: for information about this list, see http://www.jsoftware.com/forum.htm
From ljdickey Tue Apr 27 16:56:28 1999
Subject: Re: Jforum: Re: QR Decomposition (128!:0)
Did I send you a message the other day?
Sunday, I thought I shipped something, but today, I seem to have lost it.
From owner-jsoftware@lists.interlog.com Tue Apr 27 17:24:13 1999
From: "Bjorn G. Helgason" <gosi@centrum.is>
Subject: RE: Jforum: Current directory on startup
Date: Tue, 27 Apr 1999 13:21:23 +-100
Content-Type: multipart/mixed; boundary="---- =_NextPart_000_01BE90B2.6C92EFE0"
------ =_NextPart_000_01BE90B2.6C92EFE0
From: Murray Eisenberg[SMTP:murray@math.umass.edu]
> Wolfram supplied a patch, so
> that it became feasible to use Mathematica under Win 98.
You should look to see if you could get your programs to run in Linux.
Everyone I know is looking for an escaperoute out of Windows.
Several big companies are seriously considering alternatives in Linux.
/Gosi
------ =_NextPart_000_01BE90B2.6C92EFE0
Content-Type: application/ms-tnef
Content-Transfer-Encoding: base64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------ =_NextPart_000_01BE90B2.6C92EFE0--
--------------------------------------------------------------------------------
J Forum: for information about this list, see http://www.jsoftware.com/forum.htm
From piet.dejong@commerce.ubc.ca Tue Apr 27 17:49:57 1999
Date: 27 Apr 99 14:48:44 -0700
Subject: Re: Jforum: Re: QR Decomposition (128!:0)
From: "Piet de Jong" <piet.dejong@commerce.ubc.ca>
X-Fontfamily: Geneva
X-Fontsize: 12
Content-Transfer-Encoding: quoted-printable
No, I did not receive any message from you.
From owner-jsoftware@lists.interlog.com Tue Apr 27 18:53:55 1999
Date: Tue, 27 Apr 1999 18:40:42 -0400
From: Eric Iverson <eiverson@interlog.com>
Organization: Iverson Software Inc.
Subject: Re: Jforum: Mapped Files - a bug?
References: <199904271436_MC2-739F-C4C2@compuserve.com>
to Fraser Jackson:
Your mapped file problem is very strange. This is a real shot in the
dark. If you run a script with load, then those lines are local to the
verb load. In that case, any assignments done with =. are to a local
noun, not the global. Are you perhaps doing some assignments with =.
that should be done with =: ?
--------------------------------------------------------------------------------
J Forum: for information about this list, see http://www.jsoftware.com/forum.htm
From owner-jsoftware@lists.interlog.com Tue Apr 27 22:32:05 1999
Date: Tue, 27 Apr 1999 22:19:40 -0400
From: Fraser Jackson <Fraser_Jackson@compuserve.com>
Subject: Re: Jforum: Mapped Files - a bug?
Content-Disposition: inline
Content-Transfer-Encoding: 8bit
X-MIME-Autoconverted: from quoted-printable to 8bit by plus.interlog.com id WAA12756
Eric,
I agree it is very strange. The scripts in my last message were a direct
copy from the session manager and the script.
I have re run it again with some extra info. However the behaviour is
clearly reproducible. It is also very worrying. The number of items which
are added does not always seem to be the same, and it does not seem to
always start at the same location in the result. In the run reported below
the first location at which there was a difference was at location 1845563.
In some other cases they have begun much earlier.
The results are obtained on a Dell notebook running NT4. It has a 233 MHz
processor with 64MB. The files are on a SCSI drive attached via an
Adaptec Slim SCSI card.
The transcript below is from a rerun providing the additional infomentioned
in your earlier note.
load 'jmf'
load'c:\j402a\temp\96.ijs'
NB. the above script maps the file names
$~. Yeari
100
abc =: 95-Yeari
$abc
2411570
3!:0 abc
4
$Yeari
2411570
3!:0 Yeari
4
$~. Temp1
349
$Temp1
2411570
t =: 3!:0
t Temp1
4
Temp1 =: abc
$~. Temp1
100
$Temp1
2411570
t Temp1
4
load'c:\j402a\temp\eric.ijs'
$Temp1
2411570
$abc
2411570
$~.abc
100
$~. Temp1
268
t abc
4
t Temp1
4
NB. Eric's test script
abc =: 95 - Yeari
Temp1 =: abc
--------------------------------------------------------------------------------
J Forum: for information about this list, see http://www.jsoftware.com/forum.htm
From owner-jsoftware@lists.interlog.com Tue Apr 27 22:32:06 1999
Date: Tue, 27 Apr 1999 22:17:44 -0400
From: Murray Eisenberg <murray@math.umass.edu>
Organization: Mathematics & Statistics, Univ. of Mass./Amherst
Subject: Re: Jforum: Current directory on startup (off topic)
References: <01BE90B2.6C376260@geisli-339.centrum.is>
When I go to the doctor with a case of "flu", I hardly expect him to
recommend replacing my head.
I am not at all looking for an "escape route out of Windows" now. All
things considered, Windows machines are still MUCH easier to administer
than any flavor of Unix. Plus there's a MUCH larger number of devices
for which device drivers work, not to mention a huge number of
applications, compared to what's currently available in the Unix/Unix
clone world. And it's MUCH easier to buy a PC -- and to choose from a
variety of manufacturers -- with Windows installed than is the case for
Linux.
Anyway, my experience is that EVERY OS has its peculiarities. I've used
Unix long enough on a departmental server to know that.
"Bjorn G. Helgason" wrote:
>
> From: Murray Eisenberg[SMTP:murray@math.umass.edu]
> > Wolfram supplied a patch, so
> > that it became feasible to use Mathematica under Win 98.
>
> You should look to see if you could get your programs to run in Linux.
>
> Everyone I know is looking for an escaperoute out of Windows.
> Several big companies are seriously considering alternatives in Linux.
>
> /Gosi
>
> ------------------------------------------------------------------------
>
> Part 1.2 Type: application/ms-tnef
> Encoding: base64
--
Murray Eisenberg murray@math.umass.edu
Mathematics & Statistics Dept. phone 413 549-1020 (H)
Univ. of Massachusetts 413 545-2859 (W)
Amherst, MA 01003-4515
--------------------------------------------------------------------------------
J Forum: for information about this list, see http://www.jsoftware.com/forum.htm
From owner-jsoftware@lists.interlog.com Wed Apr 28 00:37:32 1999
Date: Wed, 28 Apr 1999 00:19:41 -0400
From: Murray Eisenberg <murray@math.umass.edu>
Organization: Mathematics & Statistics, Univ. of Mass./Amherst
Subject: Re: Jforum: Current directory on startup
References: <372470BA.31145619@math.umass.edu> <37248613.1BD6@interlog.com> <37250BF1.7D02AFB0@math.umass.edu> <3725A6DB.6752@interlog.com>
That's it! Although I had a bunch of other files in the main J
directory, none of them was an .ijs file. I put one of my own there,
and now File | Open starts there.
Indeed I had mapped My Documents to precisely the other directory where
file | Open had been starting.
So who's doing the dirty deed: J or Windows 98? If J, is that
desirable? (To my mind, File | Open should show the same current
directory as 1!:0 '*.*' does.)
Thanks for figuring out the problem so quickly. The easy fix will save
incredible amounts of mousing around (and perhaps nibbling away at
resources, too).
Eric Iverson wrote:
>
> to Murray Eisenberg:
>
> The following is from win98 experiments done by Chris:
>
> My win98 J file|open starts in the J directory.
>
> However, if I delete all files in the main J directory except for
> j.exe j.dll and j.lic, then file|open starts in My Documents (which
> presumably Murray has remapped). But a directory listing 1!:0 '*.*'
> shows the correct J directory.
>
> If I copy a file such as system\main\dates.ijs into the main J
> directory, file|open starts in J.
>
> If instead, I copy a file such as jexe.tlb into the main J directory,
> file|open starts in My Documents.
>
> The system seems to be looking for files other than J system files in
> the J directory. If it finds one, file|open starts in J, if not, it
> starts in My Documents.
>
> ****
>
> Win98 has changed the file open dialog. If a path is not specified (the
> case when it is first used in J) and it doesn't find 'document like
> files' in the current directory, it uses the 'document directory'
> instead of current.
>
> A workaround for now is to put a file that looks like it could be a
> document file in the J directory so that the dialog will use the current
> directory.
>
> I'll fix this in the next release by initilizing the empty string to
> explicitly reference the J directory.
>
> --------------------------------------------------------------------------------
> J Forum: for information about this list, see http://www.jsoftware.com/forum.htm
--
Murray Eisenberg murray@math.umass.edu
Mathematics & Statistics Dept. phone 413 549-1020 (H)
Univ. of Massachusetts 413 545-2859 (W)
Amherst, MA 01003-4515
--------------------------------------------------------------------------------
J Forum: for information about this list, see http://www.jsoftware.com/forum.htm
From owner-jsoftware@lists.interlog.com Wed Apr 28 03:54:14 1999
Date: 28 Apr 99 00:32:17 -0700
Subject: Re: Jforum: Re: QR Decomposition (128!:0)
From: "Piet de Jong" <piet.dejong@commerce.ubc.ca>
X-Fontfamily: Geneva
X-Fontsize: 12
Content-Transfer-Encoding: quoted-printable
Roger Hui wrote:
> In the "standard" usage, matrix inverse applies only
>to square matrices, and invertible matrices are non-singular.
>In the J (also APL) usage, matrix inverse has been extended
>to apply to tall matrices (as well as applying to square matrices),
>and invertible matrices are, I contend, non-singular:
and
>The specification for 128!:0 is that the argument must be
>non-singular. If you want something different you are
>welcome to figure it out.
My J manual uses "nonsingular" as implicitly implying "square" .
Generalized inverses are standard in the literature including those
for "tall" matrices. All matrices are invertible in some sense and
most "inverses" I am aware of reduce to the usual inverse in the
nonsingular (and hence square) case.
One of the reasons I took to J is that it contained things like QR
decompositions via 128!:0. But the manual describes it in a
nonstandard way ("Hermitian", "non-singular") making for difficulties.
Now it turns out it only applies to full column rank matrices.
(One reason why the general QR decomposition is so useful is that you
can determine rank (in the non J sense) in a numerically safe way.)
"Figuring out" what 128!:0 does in non full column rank cases without
having the computer code is silly and time consuming. Is this the
spirit of J? Those who have the relevant code might helpfully
reveal whether the output in these cases is junk, usefull or unknown.
I can live with being told any of these things. But I don't want to
waste time trying to figure out imprecise descriptions or trying to
figure out, via trial and error, what an operator is actually doing.
-- Piet
--------------------------------------------------------------------------------
J Forum: for information about this list, see http://www.jsoftware.com/forum.htm
From owner-jsoftware@lists.interlog.com Wed Apr 28 09:44:09 1999
Date: Wed, 28 Apr 1999 09:04:51 -0400
From: Eric Iverson <eiverson@interlog.com>
Organization: Iverson Software Inc.
Subject: Re: Jforum: Mapped Files - a bug?
References: <199904272219_MC2-73A6-B742@compuserve.com>
to Fraser Jackson:
I don't have complete information in the transcript you provide so it is
difficult to see what is happening. I suggest the following:
Build a single, as simple as possible, script that demonstrates the
problem. The script should include the creation of the files, the
mapping, initialization of the data and the calculations that are bad.
The script should look something like:
- *****************************************
mapbug.ijs
load'jmf'
createjmf_jmf_ 'year.jmf';size
createjmf_jmf_ 'temp.jmf';size
JINT map 'year';...
JINT map 'temp';...
year=: .... NB. initialize year
$~. year
temp=: 95-year
$~. temp NB. wrong answer!
- ******************************************
You may have to add more steps in order to recreate the problem, but
keep it as simple as possible.
--------------------------------------------------------------------------------
J Forum: for information about this list, see http://www.jsoftware.com/forum.htm
From ljdickey Wed Apr 28 13:12:59 1999
Subject: Re: Jforum: Re: QR Decomposition (128!:0)
Hmmm.... lost to the great bit bucket in the sky.
> No, I did not receive any message from you.
From ljdickey@math.uwaterloo.ca Wed Apr 28 13:13:00 1999
From: Lee Dickey <ljdickey@math.uwaterloo.ca>
Date: Wed, 28 Apr 1999 13:12:59 -0400 (EDT)
Subject: Re: Jforum: Re: QR Decomposition (128!:0)
Hmmm.... lost to the great bit bucket in the sky.
> No, I did not receive any message from you.
From ljdickey Wed Apr 28 14:23:30 1999
Subject: RE: Jforum: Current directory on startup
There is a lot of interest in Linux in the Math Faculty.
At present it is not one of the officially supported operating
systems, but I think it will be.
> You should look to see if you could get your programs to run in Linux.
>
> Everyone I know is looking for an escaperoute out of Windows.
> Several big companies are seriously considering alternatives in Linux.
From owner-jsoftware@lists.interlog.com Wed Apr 28 15:55:21 1999
Date: Wed, 28 Apr 1999 15:34:57 -0400
From: Fraser Jackson <Fraser_Jackson@compuserve.com>
Subject: Jforum: Re: QR Decomposition (128!:0)
Content-Disposition: inline
Content-Transfer-Encoding: 8bit
X-MIME-Autoconverted: from quoted-printable to 8bit by plus.interlog.com id PAA14684
Roger has specified that the present QR decomposition function is defined
over non-singular matrices. That has not always been the case, though it
may have been the intent. As Piet de Jong has noted at some stage the
behaviour of this function was changed and while it had included singular
matrices it no longer did so after the change.
It is clearly not profitable for us as users to go away and try and work
out exactly what the old function did when there is code around which could
describe that precisely. Some of us find that the new situation is not as
satisfactory as the old one. However going back to the past and that code
is not the best thing either.
Roger asked that we try and define what we want 128!:0 to do.
Its primary use is almost certainly going to be in finding a solution to
the least squares problem. As Piet has asserted there are well known
generalised inverses which provide a solution when the data is singular.
The question therefore is how are we to calculate the solution given by one
of these generalised inverses.
Those familiar with the area can point to a number of solution methods,
with differing computational demands. If we want to determine the rank of
the system and be confident of getting correct results even in
'pathological' cases then it is necessary to turn to the Singular Value
Decomposition. However if we are prepared to settle for a method which
works very well in most cases then we might settle for the QR with Column
pivoting. The reason one might want to do this is computational time.
In solving least squares problems we can use the normal equations, a method
like the QR decomposition, or a method based on the SVD. You pay an
increasing computational cost as you move from the poor numerical
propereties of the normal equations to the detailed information about the
structure of the data and numerical estimates obtained in the SVD. Faced
with a question which has to be decided in a given time, the analyst will
make a decision about method based on the importance of a numerically
precise result, and the time(and cost) required to reach it. I have found
that depending on the application I may need the speed of the normal
equations, the improved accuracy of the QR, or the complete analysis of the
SVD.
The time costs can make the difference between feasibility and
infeasibility for large data sets or large numbers of repetitions of the
calculation. The SVD is therefore not necessarily the appropriate tool for
all cases.
Because singular systems arise very frequently in practise, it would
clearly be preferable to have a function which enabled the user to deal
with them. The LAPACK manual provides some alternatives. which will have
much lower costs than the SVD. However before any changes are made we need
to carefully determine which ones provide the information essential to
cover the majority of applications. That is a non trivial task, and some
careful analysis followed by experimentation using the LAPACK interface
seems to me the way to go.
Methods which satisfy the basic requirements of many potential users will
probably have three or four elements in the boxed result. The additional
elements being a permutation matrix and a matrix specifying some orthogonal
transforms.
Fraser
--------------------------------------------------------------------------------
J Forum: for information about this list, see http://www.jsoftware.com/forum.htm
From owner-jsoftware@lists.interlog.com Fri Apr 30 00:11:39 1999
Date: Thu, 29 Apr 1999 22:47:59 -0500
From: Brian Bambrough <b.bambrough@worldnet.att.net>
Subject: Jforum: Cutting edge marketing techniques
We're about to enter a new millenium. J is the language of the
future, but there's fierce competition. Visual Basic, C++, Java,
Delphi, you name it.
So, if we're going to succeed we can't use our fathers' marketing
methods. We need new, hot, methods. We must network. Get our
message out onto the Web. And, above all, be prepared with PowerPoint
presentations. Right?
Well, not necessarily.
Last year I updated and enhanced a couple of APL systems for an
insurance company. Made them Y2K compliant too. They didn't like the
systems much. Mainly because they were written in version 4 of
Manugistics APL*PLUS II/386. Entering DOS mode and using a command
line interface was out of the Dark Ages as far as they were concerned.
They know Windows, they love Windows. All they ever want to work in
is Windows.
At the end of the job I said to my contact "You know, the guy who
created APL, Ken Iverson, has come out with a new language, J. It's
even better than APL and it's totally Windows. I could convert these
systems to J and you'd never have to work in DOS again."
Well, he talked to his boss. His boss said, "Ok." And now I'm
happily working on my first commercial J project.
--------------------------------------------------------------------------------
J Forum: for information about this list, see http://www.jsoftware.com/forum.htm
From owner-jsoftware@lists.interlog.com Fri Apr 30 03:13:59 1999
From: <paul_gauthier@dynasys.tm.fr>
Subject: RE: Jforum: Cutting edge marketing techniques
Date: Fri, 30 Apr 1999 09:02:53 +0200
Content-Transfer-Encoding: 8bit
X-MIME-Autoconverted: from quoted-printable to 8bit by plus.interlog.com id DAA19587
>-----Message d'origine-----
>De: Brian Bambrough [SMTP:b.bambrough@worldnet.att.net]
>Date: vendredi 30 avril 1999 05:48
>�: jforum
>Objet: Jforum: Cutting edge marketing techniques
>
>
>... At the end of the job I said to my contact "You know, the guy who
>created APL, Ken Iverson, has come out with a new language, J. It's
>even better than APL and it's totally Windows. I could convert these
>systems to J and you'd never have to work in DOS again."
>
>Well, he talked to his boss. His boss said, "Ok." And now I'm
>happily working on my first commercial J project.
>
>[ Paul GAUTHIER]
>Personally, I find your suggestion of J questionable in the way you
>formulated it but rather efficient to achieve the goal of having your
>first J project in this particular case.
>
>The users where I work down here are a tad more informed and they
>already are using APL2000, know all about Dyalog but prefered to
>develop their Windows interfaces with C++.
>
>I am surrounded by engineers and PhD's in OR so they wont be fooled
>that easy. They also know about products like GAMS, ASML, CPLEX
>suite and MATHLAB so they wont even be impressed by the math
>capabilities of J. I also find it annoying to avoid saying things in order
>to get my way... I guess I have seen enough manipulations to try my
>best to avoid it if possible.
>
>What I would like to see is some informal/formal members of a J team
>develop competing product to MATHLAB in order to at least keep our
>edge in the market J seems built for. LAPACK is a good move in that
>direction.
>
>I am no expert in OR and I regret it but if any team is formed or being
>formed with such a purpose in mind, please contact me since I am
>willing to give every bit of spare energy I can come up with.
>
>Regards/Paul
--------------------------------------------------------------------------------
J Forum: for information about this list, see http://www.jsoftware.com/forum.htm
From owner-jsoftware@lists.interlog.com Fri Apr 30 15:01:09 1999
Date: Fri, 30 Apr 1999 09:24:43 -0400
From: Eric Iverson <eiverson@interlog.com>
Organization: Iverson Software Inc.
Subject: Jforum: QR
References: <Pine.GSU.4.05.9904290754580.5231-100000@panix3.panix.com>
128!:0 was added in an early version of J primarily to demonstrate one
way of extending J. It was not really intended for the kind of use
recently discussed in the forum.
Lapack provides many more services and options at much better
performance in this area than we could ever hope to provide directly in
J. That is why we provide the Lapack Addon. Anyone seriously interested
should use the Lapack AddOn.
Unfortunately, Lapack is not available in J for the Mac. It is available
for all other platforms. We currently have no plans to provide Lapack on
the Mac.
128!:0 is an interesting bit of J history. If it does what you want,
great. But if not, it is not productive to worry about it as we have no
intention of making any changes in this area. Except perhaps to decommit
it in future releases.
--------------------------------------------------------------------------------
J Forum: for information about this list, see http://www.jsoftware.com/forum.htm
From owner-jsoftware@lists.interlog.com Fri Apr 30 16:36:46 1999
From: "John Tsolometes" <johnsolo@csi.com>
Subject: Re: Jforum: QR
Date: Fri, 30 Apr 1999 16:33:42 -0400
charset="iso-8859-1"
Except perhaps to decommit
>it in future releases.
>
Why take a feature out ?
It's hard enough to get the stupid machine to work as it is.
It should be clearly identified a "wierd".
Leave me
-----Original Message-----
From: Eric Iverson <eiverson@interlog.com>
Date: Friday, April 30, 1999 3:06 PM
Subject: Jforum: QR
>128!:0 was added in an early version of J primarily to demonstrate one
>way of extending J. It was not really intended for the kind of use
>recently discussed in the forum.
>
>Lapack provides many more services and options at much better
>performance in this area than we could ever hope to provide directly in
>J. That is why we provide the Lapack Addon. Anyone seriously interested
>should use the Lapack AddOn.
>
>Unfortunately, Lapack is not available in J for the Mac. It is available
>for all other platforms. We currently have no plans to provide Lapack on
>the Mac.
>
>128!:0 is an interesting bit of J history. If it does what you want,
>great. But if not, it is not productive to worry about it as we have no
>intention of making any changes in this area. Except perhaps to decommit
>it in future releases.
>
>
>---------------------------------------------------------------------------
-----
>J Forum: for information about this list, see
http://www.jsoftware.com/forum.htm
>
--------------------------------------------------------------------------------
J Forum: for information about this list, see http://www.jsoftware.com/forum.htm
From owner-jsoftware@lists.interlog.com Fri Apr 30 16:38:07 1999
Date: Fri, 30 Apr 1999 15:16:58 -0400 (EDT)
From: "J. Patrick Harrington" <jph@astro.umd.edu>
Subject: Jforum: Re: QR
Eric Iverson:
>Unfortunately, Lapack is not available in J for the Mac. It is available
>for all other platforms. We currently have no plans to provide Lapack on
>the Mac.
Where do I find Lapack for the Linux or Solaris versions of J?
- Patrick Harrington
--------------------------------------------------------------------------------
J Forum: for information about this list, see http://www.jsoftware.com/forum.htm
From owner-jsoftware@lists.interlog.com Fri Apr 30 16:44:05 1999
Date: Fri, 30 Apr 1999 16:35:52 -0400
From: David Ness <DNess@home.com>
Subject: Re: Jforum: QR
References: <000901be9348$c08c4860$80a3c0d8@johntsol>
There are answers to your general question, although I have no
opinion of whether they should apply in this case to `J'.
I can think of at least two:
(1) All features present in a system consume `support' (Space,
debugging, documentation, ...)
(2) All features present occupy some memory, even if you don't use
them.
So sometimes it is a `good thing' to decommit code and not burden the
future supporting _all_ of the past. But, it probably is best for most
of us if these situations are kept rare...
John Tsolometes wrote:
>
> Except perhaps to decommit
> >it in future releases.
> >
> Why take a feature out ?
> It's hard enough to get the stupid machine to work as it is.
> It should be clearly identified a "wierd".
> Leave me
>
--------------------------------------------------------------------------------
J Forum: for information about this list, see http://www.jsoftware.com/forum.htm
From owner-jsoftware@lists.interlog.com Fri Apr 30 18:21:17 1999
Date: 30 Apr 99 14:13:32 -0800
Subject: Re: Jforum: Re: QR Decomposition (128!:0)
From: "Piet de Jong" <piet.dejong@commerce.ubc.ca>
X-Fontfamily: Geneva
X-Fontsize: 12
Content-Transfer-Encoding: quoted-printable
You wrote in one of your previous messages:
>Roger asked that we try and define what we want 128!:0 to do.
I never saw this request. Can you email it to me? And who is "we".
Thanks for your help.
--------------------------------------------------------------------------------
J Forum: for information about this list, see http://www.jsoftware.com/forum.htm
From owner-jsoftware@lists.interlog.com Fri Apr 30 18:22:06 1999
Date: 30 Apr 99 14:09:15 -0800
Subject: Re: Jforum: QR
From: "Piet de Jong" <piet.dejong@commerce.ubc.ca>
X-Fontfamily: Geneva
X-Fontsize: 12
Content-Transfer-Encoding: quoted-printable
Where can one find documentation on the LAPACK add on in J?
--------------------------------------------------------------------------------
J Forum: for information about this list, see http://www.jsoftware.com/forum.htm
From owner-jsoftware@lists.interlog.com Fri Apr 30 18:23:23 1999
From: "Bjorn G. Helgason" <gosi@centrum.is>
Subject: RE: Jforum: Current directory on startup
Date: Thu, 29 Apr 1999 16:05:49 +-100
Content-Type: multipart/mixed; boundary="---- =_NextPart_000_01BE9358.EE028420"
------ =_NextPart_000_01BE9358.EE028420
From: Murray Eisenberg[SMTP:murray@math.umass.edu]
> So who's doing the dirty deed: J or Windows 98?
Windows 98 of course. MS has replaced the head again.
/Gosi
------ =_NextPart_000_01BE9358.EE028420
Content-Type: application/ms-tnef
Content-Transfer-Encoding: base64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------ =_NextPart_000_01BE9358.EE028420--
--------------------------------------------------------------------------------
J Forum: for information about this list, see http://www.jsoftware.com/forum.htm
From owner-jsoftware@lists.interlog.com Fri Apr 30 19:39:07 1999
Date: Fri, 30 Apr 1999 19:27:18 -0400
From: Eric Iverson <eiverson@interlog.com>
Organization: Iverson Software Inc.
Subject: Re: Jforum: Re: QR Decomposition (128!:0)
References: <B34F6B0E-AD1D2D@137.82.66.177>
Piet de Jong wrote:
> You (???) wrote in one of your previous messages:
> >Roger asked that we try and define what we want 128!:0 to do.
> I never saw this request. Can you email it to me? And who is "we".
This is water under the bridge. With the availability of Lapack for J on
most platforms, we (Iverson Software Inc.) are not interested in any
work on 128!:0 . It is not that it is without interest or merit, but
there is just too much else to do that is clearly more important.
There is information about Lapack at www.jsoftware.com. You can download
the Lapack AddOn from there, or get it from the J cdrom. The AddOn
includes the complete Lapack documentation of all its routines and also
includes a lab that shows how to use Lapack from J.
--------------------------------------------------------------------------------
J Forum: for information about this list, see http://www.jsoftware.com/forum.htm