💾 Archived View for heavysquare.com › unix › 9996-multiconnection-download-scsh.txt captured on 2022-01-08 at 14:15:07.
⬅️ Previous capture (2021-12-03)
-=-=-=-=-=-=-
There's a server (of a radio) I download audio files from. The thing is, that
the bandwidth for a connection is limited on their side to ~24Kb/sec nowdays
(several years ago there wasn't any limit). By getting the file with multiple
parallel connections solves the problem somewhat. Unfortunately the
number of connections from a given IP address is also limited to ~15. Anyway,
let's say ~240Kb/sec (when using 10 connections) is much more than 24Kb/sec.
Parts of a file can be obtained by [Curl](http://en.wikipedia.org/wiki/CURL). I
decided to use [The Scheme Shell](http://www.scsh.net/) to implement my idea
due to its thread support and strong relationship with command line tools (as
being a shell).
The solution is a fast hack. Let's see..
$ ls
getItFast.scm
$ ./getItFast.scm http://someserver/2200.mp3
$ ls
2200.mp3.00 2200.mp3.08 2200.mp3.16 2200.mp3.24 2200.mp3.32 2200.mp3.40
2200.mp3.01 2200.mp3.09 2200.mp3.17 2200.mp3.25 2200.mp3.33 2200.mp3.41
2200.mp3.02 2200.mp3.10 2200.mp3.18 2200.mp3.26 2200.mp3.34 2200.mp3.42
2200.mp3.03 2200.mp3.11 2200.mp3.19 2200.mp3.27 2200.mp3.35 getItFast.scm
2200.mp3.04 2200.mp3.12 2200.mp3.20 2200.mp3.28 2200.mp3.36
2200.mp3.05 2200.mp3.13 2200.mp3.21 2200.mp3.29 2200.mp3.37
2200.mp3.06 2200.mp3.14 2200.mp3.22 2200.mp3.30 2200.mp3.38
2200.mp3.07 2200.mp3.15 2200.mp3.23 2200.mp3.31 2200.mp3.39
$ cat 2200.mp3.* > 2200.mp3
$ rm 2200.mp3.*
$ ls
2200.mp3 getItFast.scm
$ cat getItFast.scm
#!/usr/bin/scsh
-o placeholders -o threads -o locks -s
!#
; this many thread will be started,
; each of'em represents a connection
(define POOL-SIZE 10)
; the length of a chunk in bytes
; (downloaded with one connection)
(define STEP 1000000)
(define URL (argv 1))
(define FNAME (file-name-nondirectory URL))
(define url-content-length
(lambda (url)
(string->number
(cadr ((infix-splitter (rx (+ white)))
(run/string
(| (curl -s -S -I ,url)
(grep "Content-Length"))))))))
(define LENGTH (url-content-length URL))
(define make-queue
(lambda (data-list)
(let ((lock (make-lock)))
(lambda ()
(let ((re '()))
(obtain-lock lock)
(if (null? data-list)
(set! re '())
(begin
(set! re (car data-list))
(set! data-list (cdr data-list))))
(release-lock lock)
re)))))
(define range-string
(lambda (beg end)
(let ((begs (number->string beg))
(ends (number->string end)))
(string-append begs "-" ends))))
(define get-part
(lambda (beg end fn)
(run (curl -o ,fn -s -S -r
(range-string beg end) ,URL))))
; this long is the number field
; in the filenames of parts
(define PADLEN
(string-length
(number->string
(ceiling
(/ LENGTH STEP)))))
(define file-counter-string
(lambda (i)
(let loop ((s (number->string i)))
(if (<= PADLEN (string-length s))
s
(loop (string-append "0" s))))))
(define counted-file-name
(lambda (i)
(string-append FNAME
"."
(file-counter-string i))))
; this contains the works to do
; (work ~ download a specific chunk)
; e.g. ((0 999999 "foo.mp3.00") (1000000 1999999 "foo.mp3.01") ... )
(define QUEUE
(make-queue
(let loop ((work-list '()) (low 0) (upp (- STEP 1)) (counter 0))
(if (> low LENGTH)
work-list
(loop (cons (list low upp (counted-file-name counter)) work-list)
(+ upp 1)
(min LENGTH (+ upp STEP))
(+ counter 1))))))
(define signal-thread-finish
(lambda (waiter)
(placeholder-set! waiter #f)))
(define start-worker
(lambda ()
(let ((waiter (make-placeholder)))
(spawn
(lambda ()
(let loop ()
(let ((work (QUEUE)))
(if (null? work)
(signal-thread-finish waiter)
(begin
(apply get-part work)
(loop)))))))
waiter)))
(let loop ((i POOL-SIZE) (waiters '()))
(if (= i 0)
(map placeholder-value waiters)
(loop (- i 1) (cons (start-worker) waiters))))
$
Useful links:
- [scsh concurrent programming](http://www.scsh.net/docu/html/man-Z-H-10.html)
- [scsh process notation](http://www.scsh.net/docu/html/man-Z-H-3.html)
- [r6rs](http://www.r6rs.org/final/html/r6rs/r6rs.html)