💾 Archived View for gemi.dev › gemini-mailing-list › 000597.gmi captured on 2024-03-21 at 17:54:13. Gemini links have been rewritten to link to archived content

View Raw

More Information

⬅️ Previous capture (2023-12-28)

-=-=-=-=-=-=-

[spec] Possible Tables Syntax

1. nothien (a) uber.space (nothien (a) uber.space)

There is a new discussion about table formats under the [users]
category, but I've thought of one possibility that roughly matches the
look of tables but keeps each cell on its own line, and I wanted to see
what others think of it in a more [spec]-relevant area.  Essentially,
indentation (tabs) is used to represent which column a cell is in.
Wrapping cell text is possible, so I hesitate to put it in a
preformatted block, but it's the best way to show that it's an
unfamiliar format.

> +--------------------------------+-------+
> | Food                           | Price |
> +--------------------------------+-------+
> | Eggs                           | $2    |
> | Eggs and spam                  | $4    |
> | Eggs, spam, eggs and spam      | $8    |
> | Spam spam baked beans and spam | $8    |
> | Just spam                      | $2    |
> +--------------------------------+-------+

 ```table
Food
	Price
Eggs
	$2
Eggs and spam
	$4
Eggs, spam, eggs and spam
	$8
Spam spam baked beans and spam
	$8
Just spam
	$2
 ```

It needs to be tried out a lot more to see how flexible it really is,
but it looks promising to me.  It looks quite structured visually, and
should be very easy to parse programmatically as well.  Of course, it
would be more difficult to visualize with more columns and longer cell
texts, but it's (much) better than nothing.

Each cell is on a different block, and the general expectation is that
the indentation level increases by one horizontal tab for every next
column, then resetting to no indentation for the next row.  We could
probably give specific interpretation to other possible things that
happen (e.g. 'jumps' in indentation level which are not resetting to
rows), or we could outright ban them, but I would prefer the former (as
then there is no room for extension).

I know that we're looking to finalize the spec, and that this will
probably not make it in (even if it gets accepted).  Tables are quite
uncommon, after all, so I don't there would be a big enough push to
warrant making it official.  But I argue that tables are a very
difficult thing to format to satisfy both humans and computers, so at
the minimum we would add this (or any agreed-upon table format) to the
best practices document.  The best-case scenario to me is that it gets
added to the gemtext spec as an optional type to parse, similar to how
one would parse other preformatted blocks (but with more state to keep
track of).

What do you think?

~aravk | ~nothien

Link to individual message.

2. John Cowan (cowan (a) ccil.org)

"User" editors usually don't generate HT when the user presses the TAB key.

On Sat, Jan 2, 2021 at 4:35 PM <nothien at uber.space> wrote:

> There is a new discussion about table formats under the [users]
> category, but I've thought of one possibility that roughly matches the
> look of tables but keeps each cell on its own line, and I wanted to see
> what others think of it in a more [spec]-relevant area.  Essentially,
> indentation (tabs) is used to represent which column a cell is in.
> Wrapping cell text is possible, so I hesitate to put it in a
> preformatted block, but it's the best way to show that it's an
> unfamiliar format.
>
> > +--------------------------------+-------+
> > | Food                           | Price |
> > +--------------------------------+-------+
> > | Eggs                           | $2    |
> > | Eggs and spam                  | $4    |
> > | Eggs, spam, eggs and spam      | $8    |
> > | Spam spam baked beans and spam | $8    |
> > | Just spam                      | $2    |
> > +--------------------------------+-------+
>
> ```table
> Food
>         Price
> Eggs
>         $2
> Eggs and spam
>         $4
> Eggs, spam, eggs and spam
>         $8
> Spam spam baked beans and spam
>         $8
> Just spam
>         $2
> ```
>
> It needs to be tried out a lot more to see how flexible it really is,
> but it looks promising to me.  It looks quite structured visually, and
> should be very easy to parse programmatically as well.  Of course, it
> would be more difficult to visualize with more columns and longer cell
> texts, but it's (much) better than nothing.
>
> Each cell is on a different block, and the general expectation is that
> the indentation level increases by one horizontal tab for every next
> column, then resetting to no indentation for the next row.  We could
> probably give specific interpretation to other possible things that
> happen (e.g. 'jumps' in indentation level which are not resetting to
> rows), or we could outright ban them, but I would prefer the former (as
> then there is no room for extension).
>
> I know that we're looking to finalize the spec, and that this will
> probably not make it in (even if it gets accepted).  Tables are quite
> uncommon, after all, so I don't there would be a big enough push to
> warrant making it official.  But I argue that tables are a very
> difficult thing to format to satisfy both humans and computers, so at
> the minimum we would add this (or any agreed-upon table format) to the
> best practices document.  The best-case scenario to me is that it gets
> added to the gemtext spec as an optional type to parse, similar to how
> one would parse other preformatted blocks (but with more state to keep
> track of).
>
> What do you think?
>
> ~aravk | ~nothien
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.orbitalfox.eu/archives/gemini/attachments/20210102/80f9
e9d3/attachment-0001.htm>

Link to individual message.

3. Philip Linde (linde.philip (a) gmail.com)

On Sat, 02 Jan 2021 22:35:36 +0100
nothien at uber.space wrote:

> What do you think?

I think that using preformatted blocks as an extension vector for table
data is a bad idea. Preformatted then no longer always means
preformatted.

I also have qualms about the suggested format itself because there are
already formats for describing tabular data with a lot of tooling
support. If the goal is machine readability, choose a format which is
already supported by numerous software.

In particular I think it's a bad idea to introduce something like this
as an optional recommended best practice. Either go all in, or what
looks like a table in one client will look nothing like a table in
another.

My suggested solution: if you really need an in-line human readable
table, lay it out in a preformatted block such that it makes sense to a
user with a client that interprets and displays preformatted blocks as
preformatted blocks. If you need the same data to be available machine
readable, link to a corresponding CSV/TSV/*SV below.

-- 
Philip
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 488 bytes
Desc: not available
URL: <https://lists.orbitalfox.eu/archives/gemini/attachments/20210103/4355
5c13/attachment.sig>

Link to individual message.

4. nothien (a) uber.space (nothien (a) uber.space)

Philip Linde <linde.philip at gmail.com> wrote:
> I think that using preformatted blocks as an extension vector for
> table data is a bad idea. Preformatted then no longer always means
> preformatted.

Yes, I agree, the container format (```table```) is bad.  I was more
interested in the actual format - but we can definitely change the
containing layout.  Do you have any suggestions?

> I also have qualms about the suggested format itself because there are
> already formats for describing tabular data with a lot of tooling
> support. If the goal is machine readability, choose a format which is
> already supported by numerous software.

The issue is that none of the other table formats I've seen are both
human readable and easily machine parsable.  This one is.

> In particular I think it's a bad idea to introduce something like this
> as an optional recommended best practice. Either go all in, or what
> looks like a table in one client will look nothing like a table in
> another.

There are two issues with going 'all in' (which I understand as making
it a proper part of the gemtext part of the spec): firstly, that clients
will treat it as a must-have, which it isn't (it's pretty readable
without any fancy rendering), and secondly that the spec is almost
finalized right now, and I don't see it making it in (because of how
rarely it comes up).  I did mention this as a best case scenario, but
your last point is why I've introduced this idea at all - to prevent the
aggregation of multiple different supported table formats.

> My suggested solution: if you really need an in-line human readable
> table, lay it out in a preformatted block such that it makes sense to
> a user with a client that interprets and displays preformatted blocks
> as preformatted blocks. If you need the same data to be available
> machine readable, link to a corresponding CSV/TSV/*SV below.

Well, with my format, you can have both.  Of course, even slighly large
tables can and should be kept in CSV or similar, but this works well for
small tables which may be modified often.  The issue with a fully laid
out table in a preformatted block is 1) that it's getting into
presentation (different users will create different looking tables,
adding style as they see fit, but that's against the general gemtext
philosophy), and 2) that it's difficult to modify cells without having
to adjust the entire table.  My format is affected by neither, because
all the table control is done via whitespace (and one can't style HTs
and LFs), and because cells are on individual lines which are much
easier to move around or to add to.

~aravk | ~nothien

Link to individual message.

5. nothien (a) uber.space (nothien (a) uber.space)

John Cowan <cowan at ccil.org> wrote:
> "User" editors usually don't generate HT when the user presses the TAB
> key.

Sounds like an opinion, but then so is mine.  I'm using (Neo)Vi(m),
where whether HTs or spaces are inserted depends upon the format, but is
easily changeable (:set [no]expandtab or :set [no]et).  I don't actually
know about many other editors.  Do you have any data to back this up?
I'll go looking myself as well, but if it turns out that HTs are
uncommon then we may be able to replace them with 4 spaces (of course,
that opens up the question of what happens when the number of spaces on
a table line is not a multiple of 4, but we can figure that out).

~aravk | ~nothien

Link to individual message.

6. zwatotem (a) gmail.com (zwatotem (a) gmail.com)

First of all, **Hello**, as it's my first message here.
I like the original idea to some extent. It bothers me, that what is meant
to be a row doesn't actually lay in one line. I immediately wanted to
propose usage of TSVs as they are pretty human readable and have a good
support in many text parsers. I don't want explicit support for tables, but
rather making TSV inside a preformatted block a *best practice* for
presenting data of tabular form. Fancy clients could have support for a
technique of HT rendering known as [elastic
tabstops](https://nickgravgaard.com/elastic-tabstops/) (tl;dr render tabs in
blocks of neighboring lines to the same stop point, widening ones that are
too short, if needed). But then simple terminal-based clients aiming to
benefit from one-scan-to-render feature of text/gemini would suffer with all
the usual problems we know from attempts to display TSV in the terminal.

Example data (pure TSV):
 ```
Name	Age	Address
Paul	23	1115 W Franklin
Bessy the Cow	5	Big Farm Way
Zeke	45	W Main St
 ```

Elastically rendered (using more tabs here, for illustration):
 ```
Name			Age	Address
Paul			23	1115 W Franklin
Bessy the Cow		5	Big Farm Way
Zeke			45	W Main St
 ```

Solution could be to implement elastic tabstops in a greedy manner (align to
the widest cell so-far). I think that greedy renderer is a good midpoint
between naive aligning to nearest multiples of 4 (or 8 or whatever) and full
elastic tabstops rendering. With this we could encourage users to make a
table header containing cells of maximum needed length (for example by
padding the remaining width with spaces). Obvious flaw of this solution is
requiring text author to know the maximum widths of their columns (which can
potentially be big raw data impossible to look up). But then TSV will look
like table no matter what.

Greedy/semi elastic render (without padded header) (`Name` column made wider
after encountering Bessy the Cow cell):
 ```
Name	Age	Address
Paul	23	1115 W Franklin
Bessy the Cow		5	Big Farm Way
Zeke			45	W Main St
 ```
Now that I think about it, this has little to nothing to do with spec (maybe
I should send it to [tech] as well, or instead) as all of this can be put in
best practices or even emerge as non-written standard for
formulating/rendering documents. However, I think this is an interesting use
of an uncommon text rendering technique to make plaintext capable of
presenting tabular data. Let me know, what you think about that.

Grzegorz Przybylski aka Zwatotem

Link to individual message.

7. ew.gemini (ew.gemini (a) nassur.net)

Hello,


nothien at uber.space writes:

> There is a new discussion about table formats under the [users]
> category, but I've thought of one possibility that roughly matches the
> look of tables but keeps each cell on its own line, and I wanted to see
> what others think of it in a more [spec]-relevant area.  Essentially,
> indentation (tabs) is used to represent which column a cell is in.
> Wrapping cell text is possible, so I hesitate to put it in a
> preformatted block, but it's the best way to show that it's an
> unfamiliar format.
>
>> +--------------------------------+-------+
>> | Food                           | Price |
>> +--------------------------------+-------+
>> | Eggs                           | $2    |
>> | Eggs and spam                  | $4    |
>> | Eggs, spam, eggs and spam      | $8    |
>> | Spam spam baked beans and spam | $8    |
>> | Just spam                      | $2    |
>> +--------------------------------+-------+
>
> ```table
> Food
>       Price
> Eggs
>       $2
> Eggs and spam
>       $4
> Eggs, spam, eggs and spam
>       $8
> Spam spam baked beans and spam
>       $8
> Just spam
>       $2
> ```

Well, I quite strongly oppose this idea.
If your table like layout cannot be made with toggling
preformatted lines useing ```, what's the point then?
If you really need tables, maybe large tables, who is stopping
the author from using xml, http, csv, or whatnot, and make it a
downloadable file?

>snip<
>
> What do you think?

Don't do it.
~ew

-- 
Keep it simple!
-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 861 bytes
Desc: not available
URL: <https://lists.orbitalfox.eu/archives/gemini/attachments/20210103/a647
3247/attachment.sig>

Link to individual message.

8. nothien (a) uber.space (nothien (a) uber.space)

"ew.gemini" <ew.gemini at nassur.net> wrote:
> If your table like layout cannot be made with toggling preformatted
> lines useing ```, what's the point then?

I'm sorry, I don't understand what you mean by this.  Also, I actually
don't like the ```...``` layout I've used here, and I'm open to
suggestions for replacing it.  The primary concern to me is the actual
table syntax.

> If you really need tables, maybe large tables, who is stopping the
> author from using xml, http, csv, or whatnot, and make it a
> downloadable file?

Well, yes, I agree.  In fact, I said the same thing in my initial mail
on this topic.  But what about small tables?  Yes, they can be done in a
separate file, but that's often too much work for them - it's much
easier to just do it inline.  Especially when there are many of them.
Having a separate file also makes it harder on users, particularly for
those who don't have clients that do table inlining (e.g. gmnlm).

~aravk | ~nothien

Link to individual message.

9. Petite Abeille (petite.abeille (a) gmail.com)



> On Jan 3, 2021, at 21:09, nothien at uber.space wrote:
> 
> But what about small tables?  Yes, they can be done in a
> separate file, but that's often too much work for them - it's much
> easier to just do it inline.  

Use embedded data links:

=> data:text/csv;1997%2CFord%2CE350 Table Caption

Ask you favorite client to render them inline.

? ???

Link to individual message.

10. Karmanyaah Malhotra (karmanyaahm (a) gmail.com)

I think having some, any, sort of table syntax that can be adjusted by
the client is useful.


Perhaps a balance between pre-formatted and attached could be something like
 ```parse:table
+--------------------------------+-------+
| Food                           | Price |
+--------------------------------+-------+
| Eggs                           | $2    |
| Eggs and spam                  | $4    |
| Just spam                      | $2    |
+--------------------------------+-------+
 ```
This way, clients don't *have* to parse it, but if supported, they can.


Perhaps easier to parse would be.
 ```parse:table:/link/to/file.csv
+--------------------------------+-------+
| Food                           | Price |
+--------------------------------+-------+
| Eggs                           | $2    |
| Eggs and spam                  | $4    |
| Just spam                      | $2    |
+--------------------------------+-------+
 ```
This could again be optional and would allow for choice on whether
content authors want their tables rendered and if clients support it. If
the client doesn't support it, the table would show up as pre-formatted.

The second approach would require more work by authors but the
preformatted tables can be easily generated from a CSV (or any other
format) by servers or static site generators.
-- 
Karmanyaah Malhotra
https://karmanyaah.malhotra.cc/contact/


-------------- next part --------------
A non-text attachment was scrubbed...
Name: OpenPGP_signature
Type: application/pgp-signature
Size: 488 bytes
Desc: OpenPGP digital signature
URL: <https://lists.orbitalfox.eu/archives/gemini/attachments/20210103/d333
56a6/attachment.sig>

Link to individual message.

11. Katarina Eriksson (gmym (a) coopdot.com)

Karmanyaah Malhotra <karmanyaahm at gmail.com> wrote:

> I think having some, any, sort of table syntax that can be adjusted by
> the client is useful.
>

Plenty of things would be useful. The question isn't "is it useful?" The
question is "can we do without?"

 ```parse:table:/link/to/file.csv
> +--------------------------------+-------+
> | Food                           | Price |
> +--------------------------------+-------+
> | Eggs                           | $2    |
> | Eggs and spam                  | $4    |
> | Just spam                      | $2    |
> +--------------------------------+-------+
> ```
>

Ah, yes. Abusing the alt text. I've done that [1] and Solderpunk was not
exactly amused. [2]

Sure, use of alt text is at the client's discretion and the spec mentiones
using it to indicate a programming language. Do keep in mind that it will
be read out to visually impaired users, so please go easy on the syntax.

For links, I recommend link lines. That way, everyone can get the file.

Here is another example, with added line numbers to show intentional line
breaks:

00]
01] ``` The table listing Food and Price says: Eggs are $2, Eggs and spam
are $4, Just spam are $2.
02] +--------------------------------+-------+
03] | Food                           | Price |
04] +--------------------------------+-------+
05] | Eggs                           | $2    |
06] | Eggs and spam                  | $4    |
07] | Just spam                      | $2    |
08] +--------------------------------+-------+
09] ```
10] => /link/to/file.csv parse:table
11]

>
I put the link line on line 10 but you could put it on line 0 if you want.

[1] gemini://gemi.dev/gemini-mailing-list/messages/000930.gmi
[2] gemini://gemi.dev/gemini-mailing-list/messages/000933.gmi

-- 
Katarina
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.orbitalfox.eu/archives/gemini/attachments/20210104/b323
cd64/attachment.htm>

Link to individual message.

---

Previous Thread: [announce] Gemini in 2021

Next Thread: [users] Lightweight markup language