gmane.lisp.scheme.gauche

[ANN] GaUnit 0.1.6

Kouhei Sutou — 2008-10-22T13:06:27

Hi, I've released GaUnit 0.1.6: http://www.cozmixng.org/~rwiki/?cmd=view;name=GaUnit http://www.cozmixng.org/~kou/download/gaunit-0.1.6.tar.gz GaUnit is a xUnit based unit testing framework for Gauche. ! How to write? gauche.test: (test "add" 3 (lambda () (+ 1 2))) (test "sub" -1 (lambda () (- 1 2))) GaUnit: (define (test-add) (assert-equal 3 (+ 1 2))) (define (test-sub) (assert-equal -1 (- 1 2))) ! Output example A failure test derived from test/string.scm in the Gauche repository with a small change (bar -> baz): gauche.test: (test* "string-join" "foo::baz::baz" (string-join '("foo" "bar" "baz") "::")) GaUnit: (define (test-string-join) (assert-equal "foo::baz::baz" (string-join '("foo" "bar" "baz") "::"))) Here is output: gauche.test: test string-join: expects "foo::baz::baz" => got "foo::bar::baz" GaUnit: 1) Failure: test-string-join ./test-string.scm:4: (assert-equal "foo::baz::baz" (string-join '("foo" "bar" "baz") "::")) expected: <"foo::baz::baz"> but was: <"foo::bar::baz"> gauche.test's output is compact. GaUnit's output is verbose. It includes the followings: * stack trace with Emacs friendly format * aligned expected and actual value output for easy comparable * (diff if it is needed) Why is GaUnit's failure report verbose? It's for easy debugging. Thanks, -- kou ------------------------------------------------------------------------- This SF.Net email is sponsored by the Moblin Your Move Developer's challenge Build the coolest Linux based applications with Moblin SDK & win great prizes Grand prize is a trip for two to an Open Source event anywhere in the world http://moblin-contest.org/redirect.php?banner_id=100&url=/

Re: 0.8.14 FFI changes

Shiro Kawai — 2008-10-13T23:25:31

I'm not aware of any library-level API incompatibility in 0.8.14. The binary structure layout has been changed, as usual, so you need to recompile the stuff (the generated C files from 'stub' files have been changed, so you should build from clean state). But if you get an error even after that, could you describe some more concrete example where the incompatibility araises? I might have overlooked something. --shiro From: "Bill Schottstaedt" ccrma.Stanford.EDU> Subject: [Gauche-devel] 0.8.14 FFI changes Date: Mon, 13 Oct 2008 15:20:21 -0700 ------------------------------------------------------------------------- This SF.Net email is sponsored by the Moblin Your Move Developer's challenge Build the coolest Linux based applications with Moblin SDK & win great prizes Grand prize is a trip for two to an Open Source event anywhere in the world http://moblin-contest.org/redirect.php?banner_id=100&url=/

0.8.14 FFI changes

Bill Schottstaedt — 2008-10-13T22:20:21

I moved to 0.8.14 today, and all my stuff immediately segfaults. Is there a list somewhere of the changes in this version as relates to foreign code? For example, optional arguments to a foreign function trigger a "wrong number of arguments" error. ------------------------------------------------------------------------- This SF.Net email is sponsored by the Moblin Your Move Developer's challenge Build the coolest Linux based applications with Moblin SDK & win great prizes Grand prize is a trip for two to an Open Source event anywhere in the world http://moblin-contest.org/redirect.php?banner_id=100&url=/

ANN: dyncomp 0.2.0

KOGURO, Naoki — 2008-10-11T12:19:57

Hi all. dyncomp 0.2.0 is released. dycomp is a module that enables to create C function with gauche.cgen.cise (C in S expression). This function is compiled to native code with Tiny C Compiler (http://bellard.org/tcc/), so you can create a fast function. The change from the previous version is: - Gauche module has bindings of C symbols, so different gauche modules can associate different binding to the same C symbol. - Added some macros (define-cfn, dyncomp-reset!). See http://homepage.mac.com/naoki.koguro/prog/dyncomp/index.html for more details. -- KOGURO, Naoki koguro.net> ------------------------------------------------------------------------- This SF.Net email is sponsored by the Moblin Your Move Developer's challenge Build the coolest Linux based applications with Moblin SDK & win great prizes Grand prize is a trip for two to an Open Source event anywhere in the world http://moblin-contest.org/redirect.php?banner_id=100&url=/

Release 0.8.14

Shiro Kawai — 2008-10-07T14:35:34

Gauche 0.8.14 is available. http://practical-scheme.net/gauche/ This release is to flush out cumulative bug fixes and small improvements. Nothing particularly shiny, but the performance is also slightly improved, so you might want to upgrade. In SVN trunk there's a module parser.peg, a parser combinator library for parsing expression grammar. It is NOT included in this release. I want it be the default choice of the parser of Gauche, and not satisfied by the current implementation. You may try it from svn, but be aware that the API will be changed a lot in future. Some suggested changes are not incorporated due to the lack of the time of testing, including Tomas Stanek's suggestion of using USE_MMAP in GC. It's been over 4 years since Gauche reached 0.8. It's not that the development is slowed down; in fact a lot of code has been added or rewritten since then. These days I get more and more opportunities to use Gauche in real work, and keep adding stuff that are required in the production environment. During the course, I started to feel to adjust the original plan a little bit, though. Originally I wanted to fix C ABI by 0.9, so that after 1.0 you don't need to recompile every extensions for each new release of Gauche. However the goal seems further than I thought, mainly because improvements in various parts (including VM) tend to require API change. I also start to think that it won't be that bad to recompile extensions, for the job can be largely automated by clever package management system. On the other hand, from the production application I feel the need to work more on the domain of debugging, performance tuning, and better deployment machinery. So I decided to put off C ABI fix after 1.0; I aim at releasing 0.9 as the wrap up of all those extensions of 0.8.x series, then start concentrating more on the 'production friendly' region. Well it's just a plan so I might adjust it again, but this is what I have in mind currently. --shiro ------------------------------------------------------------------------- This SF.Net email is sponsored by the Moblin Your Move Developer's challenge Build the coolest Linux based applications with Moblin SDK & win great prizes Grand prize is a trip for two to an Open Source event anywhere in the world http://moblin-contest.org/redirect.php?banner_id=100&url=/

Re: garbage collector issues

Tomas Stanek — 2008-09-04T11:54:30

Is this branch already available, or will this technique be used in future releases of gauche? I would especially be interested if there would be some relatively cheap technique (similar to integer shifting) which would distinguish those doubles from potential pointers. Yes, it's the latter. It is bounded, but in my case, the bounds heavily depend on those larger blocks rather than structures like queues etc. I've tried playing with those intermediate structures with virtualy no effect on the problem. Also, when I looked on GC_Dump, almost all used memory was in a large amount of free lists. It seems to me, that the problem is sort of still there even when -DUSE_M[UN]MAP is used, because GC still keeps those free lists. The corresponding pages are however ramapped as PROT_NONE (and potentialy mapped back later), so the overhead of "fragmented free heap" stays mainly in GC's internal structures and OS memory maps, which is relatively low. Best regards, Tomas Stanek ------------------------------------------------------------------------- This SF.Net email is sponsored by the Moblin Your Move Developer's challenge Build the coolest Linux based applications with Moblin SDK & win great prizes Grand prize is a trip for two to an Open Source event anywhere in the world http://moblin-contest.org/redirect.php?banner_id=100&url=/

Re: garbage collector issues

Shiro Kawai — 2008-09-04T10:58:03

Thanks, Tomas. From: Tomas Stanek psi.cz> Subject: [Gauche-devel] garbage collector issues Date: Thu, 4 Sep 2008 12:05:44 +0200 I think you're correct. It doesn't need to be SCM_NEW (technically ScmFlonum has a pointer to the class, but it is allocated staticallly and will never be GCed, so we don't need to consider it.) In a branch of the source tree I implemented a different treatment of flonums, in which ScmFlonum is only 8 bytes (it's just double). It also uses a sort of specialized generational GC to reduce overhead of ephemeral flonums. (The detail is in my DLS2008 paper which will be available on ACM site sometime in future). Good suggestion. We need to watch two numbers w.r.t memory consumption: (1) Does the value grows over time without bound? Then it is a leak, and one likely cause would be false pointers. Especially you need to watch out non weak-GC-robust data structures such as queues. (2) If it isn't a leak, still the absolute amount of memory could be a problem, if there exists unnecessary large overhead. A likely cause would be fragmentation of relatively large blocks. I assume what you're observing is the latter. You may be able to get some clue by setting GC_PRINT_STATS environment variable. Anyways, using mmap option seems worth looking. Thanks again for reporting this. --shiro ------------------------------------------------------------------------- This SF.Net email is sponsored by the Moblin Your Move Developer's challenge Build the coolest Linux based applications with Moblin SDK & win great prizes Grand prize is a trip for two to an Open Source event anywhere in the world http://moblin-contest.org/redirect.php?banner_id=100&url=/

garbage collector issues

Tomas Stanek — 2008-09-04T10:05:44

Hello, I've encountered quite large memory consumption of gauche under some circumstances. My program reads an xml document, converts it into a tree with lots of flonums (sometimes about 20000), do some work with it and continues to another document. The program also allocates few large (up to a megabyte) blocks and lot of pairs/flonums/etc during each iteration. Problem is, that it's memory consuption grows up to 200 - 300MB and never goes down. I've noticed two things... First, Scm_MakeFlonum uses SCM_NEW (not SCM_NEW_ATOMIC) for allocation of the ScmFlonum, which is pointer free. I expectd this to be a "typo". Changing SCM_NEW to SCM_NEW_ATOMIC improves the situation a bit, but not as much. Second, Boehm Weiser GC is by default configured not to use -DUSE_MMAP and -DUSE_MUNMAP flags. This way, GC doesn't ever return allocated memory to the operating system. By using -DUSE_MMAP and -DUSE_MUNMAP, memory consumption keeps around 20 - 30MB. I'm not sure why -DUSE_M[UN]MAP makes a such difference. My speculation is that GC's freelists are poluted with large blocks which are not very usable for future allocations and GC tends to grow heap. My second point is just a remark and maybe suggestion that -DUSE_M[UN]MAP could be useful as gauche's compile-time option. Best regards, Tomas Stanek ------------------------------------------------------------------------- This SF.Net email is sponsored by the Moblin Your Move Developer's challenge Build the coolest Linux based applications with Moblin SDK & win great prizes Grand prize is a trip for two to an Open Source event anywhere in the world http://moblin-contest.org/redirect.php?banner_id=100&url=/

Re: Connecting 2 processes by a pipe

Shiro Kawai — 2008-08-28T22:19:13

From: "Stas Boukarev" gmail.com> Subject: [Gauche-devel] Connecting 2 processes by a pipe Date: Fri, 29 Aug 2008 00:55:22 +0400 The general way is a bit cumbersome in the current Gauche. If you're looking for a quick way, delegating the piping work to the shell is the easiest. For example: (process-output->string "cat file | wc -l") If you don't want shell to be involved, you have to do the job what shells are doing: Create pipes and connect the processes with them. The high-level process interface doesn't help much, unfortunately. This is a quick sample I typed in my REPL: gosh> (receive (i0 o0) (sys-pipe) (receive (i1 o1) (sys-pipe) (let ((p1 (sys-fork-and-exec "cat" '("cat" "configure") :iomap `((1 . ,o0) (2 . 2)))) (p2 (sys-fork-and-exec "wc" '("wc" "-l") :iomap `((0 . ,i0) (1 . ,o1) (2 . 2))))) (close-output-port o0) (close-output-port o1) (read-line i1)))) "2625" Basically, what it does is to connect processes using two pipes: cat -> o0==pipe==i0 -> wc -> o1==pipe==i1 -> read-line There are a few caveats. Don't forget to close o0, otherwise 'wc' will keep waiting for input even 'cat' is done its job (because, if you keep o0 open, it effectively becomes a 'Y' pipe with two entrance.) Also, strictly speaking, you have to call sys-waitpid on p1 and p2 to collect exit status, otherwise they become zombies. (In practice, if your script is not long-running, you can leave them alone and the system will collect them after your script exits). Scsh has a nice high-level interface that automates these under-the-hood wirings, and I have a plan to port it to Gauche. --shiro ------------------------------------------------------------------------- This SF.Net email is sponsored by the Moblin Your Move Developer's challenge Build the coolest Linux based applications with Moblin SDK & win great prizes Grand prize is a trip for two to an Open Source event anywhere in the world http://moblin-contest.org/redirect.php?banner_id=100&url=/

Connecting 2 processes by a pipe

Stas Boukarev — 2008-08-28T20:55:22

How one can connect two processes by a pipe like "cat file | wc -l"? I read "9.16 gauche.process" section of the manual but still can't get it.

Re: expt hangs

Shiro Kawai — 2008-08-09T21:15:35

From: "Bill Schottstaedt" ccrma.Stanford.EDU> Subject: [Gauche-devel] expt hangs Date: Sat, 9 Aug 2008 12:52:15 -0700 Not exactly. It just takes long to calculate. gosh> (time (begin (expt 1/500029 362880/3) #t)) ;(time (expt 1/500029 120960)) ; real 9.364 ; user 9.340 ; sys 0.000 #t But more importantly, it takes way too long to *print* it. Note that you're asking exact calculation. The result will be something like: 1/........huge_number......... The denominator will be about 690000 digits long. Converting bignum binary to decimal involves lots of divisions of bignums. Exact numeric arithmetics is not in Gauche's emphasis; it uses n^2 algorithm for bignum multiplication and division, so the performance deteriorates quickly if the number gets huge. The question is, do you need exact calculation here? --shiro ------------------------------------------------------------------------- This SF.Net email is sponsored by the Moblin Your Move Developer's challenge Build the coolest Linux based applications with Moblin SDK & win great prizes Grand prize is a trip for two to an Open Source event anywhere in the world http://moblin-contest.org/redirect.php?banner_id=100&url=/

expt hangs

Bill Schottstaedt — 2008-08-09T19:52:15

I think the expression (expt 1/500029 362880/3) causes Gauche to hang. ------------------------------------------------------------------------- This SF.Net email is sponsored by the Moblin Your Move Developer's challenge Build the coolest Linux based applications with Moblin SDK & win great prizes Grand prize is a trip for two to an Open Source event anywhere in the world http://moblin-contest.org/redirect.php?banner_id=100&url=/

Re: write-object from write vs. from display

Shiro Kawai — 2008-08-02T03:36:36

From: "Leonardo Boiko" gmail.com> Subject: Re: [Gauche-devel] write-object from write vs. from display Date: Tue, 29 Jul 2008 19:57:25 -0300 That was the original plan. The problem is this: It is quite common to call 'format' from write-object, but there's no way to pass the mode info to 'format'. (define-method write-object ((obj ) port mode) (format port "...." ...)) We can add a new function, say 'format/mode' and 'format/ss/mode', that takes extra mode argument. But we can't save the existing write-object methods that are using standard 'format'---it means for any recursive structure, we can't rely on the mode argument to propagate down. With the parameter approach, we don't need to worry about propagation part. --shiro ------------------------------------------------------------------------- This SF.Net email is sponsored by the Moblin Your Move Developer's challenge Build the coolest Linux based applications with Moblin SDK & win great prizes Grand prize is a trip for two to an Open Source event anywhere in the world http://moblin-contest.org/redirect.php?banner_id=100&url=/

Re: write-object from write vs. from display

Leonardo Boiko — 2008-07-29T22:57:25

Thanks for the attention Shiro, I'll wait more then to see how it will turn out then. I'm serializing the objects of my game's world with write-object plus srfi10 read-time constructors (via a simple generic printer/constructor for arbitrary gauche classes; dunno if someone tried this already?) — so far I didn't notice any problems, but then I'm not using shared/nested structures. 2008/7/29 Shiro Kawai lava.net>: I still don't get it; will the user ever need to set the write-mode? If not, wouldn't it be simpler to just make write-object receive two arguments, the object and the write-mode? ------------------------------------------------------------------------- This SF.Net email is sponsored by the Moblin Your Move Developer's challenge Build the coolest Linux based applications with Moblin SDK & win great prizes Grand prize is a trip for two to an Open Source event anywhere in the world http://moblin-contest.org/redirect.php?banner_id=100&url=/ _______________________________________________ Gauche-devel mailing list Gauche-devel< at >lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/gauche-devel

Re: write-object from write vs. from display

Shiro Kawai — 2008-07-29T22:02:48

From: "Leonardo Boiko" gmail.com> Subject: Re: [Gauche-devel] write-object from write vs. from display Date: Tue, 29 Jul 2008 18:16:24 -0300 The bottom problem is that the current write-object API is broken; it's not only about write/display distinction. There's another special circumstances when write-object is called, one of which is the 'walk' stage of write/ss and format/ss (I omit lengthy explanation here, but the point is that currently I use nasty duct- taping to hide this case, which may break in some edge cases). Realizing generic pretty-printer also requires the context information. I've been thinking how to support this context problem without changing the current API drastically. I think the best way, from the user's point of view, is using dynamic parameters. Since you brought up the topic, it may be a good time to hammer out the concrete implementation. How about this. In your write-object method you can call write-mode procedure that returns a symbol indicating which mode it is called. (write-mode) => a symbol The symbol may be one of the followings: write - called from write, or ~s directive of format display - called from display, or ~a directive of format walk - a special 'walk' pass from write/ss or format/ss. In the 'walk' mode usually you can do the same thing as you do in 'write' mode, but there are certain things you may want to avoid. I'll elaborate it when I write down the reference. There are other context information to be passed around, but I think this API won't conflict with future development. --shiro ------------------------------------------------------------------------- This SF.Net email is sponsored by the Moblin Your Move Developer's challenge Build the coolest Linux based applications with Moblin SDK & win great prizes Grand prize is a trip for two to an Open Source event anywhere in the world http://moblin-contest.org/redirect.php?banner_id=100&url=/_______________________________________________ Gauche-devel mailing list Gauche-devel< at >lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/gauche-devel

Re: write-object from write vs. from display

Leonardo Boiko — 2008-07-29T21:16:24

That's a bit over my level… If I just patched in a display-object method (defaulting to write-object) and made display use it, would you accept? =) ------------------------------------------------------------------------- This SF.Net email is sponsored by the Moblin Your Move Developer's challenge Build the coolest Linux based applications with Moblin SDK & win great prizes Grand prize is a trip for two to an Open Source event anywhere in the world http://moblin-contest.org/redirect.php?banner_id=100&url=/ _______________________________________________ Gauche-devel mailing list Gauche-devel< at >lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/gauche-devel

Re: write-object from write vs. from display

Shiro Kawai — 2008-07-29T21:10:50

Right now there's no way to distinguish them. I've been planning to replace the object write protocol to passing around context information (including write/display distinction). --shiro From: "Leonardo Boiko" gmail.com> Subject: [Gauche-devel] write-object from write vs. from display Date: Tue, 29 Jul 2008 12:33:04 -0300 ------------------------------------------------------------------------- This SF.Net email is sponsored by the Moblin Your Move Developer's challenge Build the coolest Linux based applications with Moblin SDK & win great prizes Grand prize is a trip for two to an Open Source event anywhere in the world http://moblin-contest.org/redirect.php?banner_id=100&url=/

write-object from write vs. from display

Leonardo Boiko — 2008-07-29T15:33:04

Hi. According to gauche manual, When `write' and `display' encounter an object of a user-defined class, they call the generic function `write-object'. The same generic is called for both methods? How do I make my objects react differently to 'write' and 'display'? ------------------------------------------------------------------------- This SF.Net email is sponsored by the Moblin Your Move Developer's challenge Build the coolest Linux based applications with Moblin SDK & win great prizes Grand prize is a trip for two to an Open Source event anywhere in the world http://moblin-contest.org/redirect.php?banner_id=100&url=/

Re: c-wrapper doesn't play with the module system?

Leonardo Boiko — 2008-07-29T13:12:55

2008/7/29 KOGURO, Naoki koguro.net>: Aaah I see now. Thanks :) ------------------------------------------------------------------------- This SF.Net email is sponsored by the Moblin Your Move Developer's challenge Build the coolest Linux based applications with Moblin SDK & win great prizes Grand prize is a trip for two to an Open Source event anywhere in the world http://moblin-contest.org/redirect.php?banner_id=100&url=/

Re: c-wrapper doesn't play with the module system?

KOGURO, Naoki — 2008-07-29T12:59:25

Hi Leonardo, On 2008/07/29, at 11:18, Leonardo Boiko wrote: Without c-wrapper, this happens. gosh> (define (filter) #f) filter gosh> (use gauche.collection) # gosh> filter # gosh> (filter (lambda (i) (lisp? i)) '(a b c)) *** ERROR: wrong number of arguments for # (required 0, got 2) It seems to me that Gauche searches a binding in a current module first, then searches a binding in imported modules. c-wrapper defines symbols in a current module, so ncursesw's filter defined by c-wrapper is used, I think. If you avoid this case, define a module for ncursesw. ;;; the module for ncursesw (define-module my-curses (use c-wrapper) (export ncursesw-filter << add symbols you need >>) ) (select-module my-curses) (c-load-library "libncursesw") (c-include "ncursesw.h") (define ncursesw-filter filter) (provide "my-curses") ;;; main program (use my-curses) (use gauche.collection) (filter (lambda (i) (lisp? i)) '(scheme ruby cl haskell lolcat)))) (ncursesw-filter) -- KOGURO, Naoki koguro.net> ------------------------------------------------------------------------- This SF.Net email is sponsored by the Moblin Your Move Developer's challenge Build the coolest Linux based applications with Moblin SDK & win great prizes Grand prize is a trip for two to an Open Source event anywhere in the world http://moblin-contest.org/redirect.php?banner_id=100&url=/

c-wrapper doesn't play with the module system?

Leonardo Boiko — 2008-07-29T02:18:04

Hi. Sorry if this is a frequent question, but I couldn't find info. I'm using ncursesw with c-wrapper like this: (use c-wrapper) (c-load-library "libncursesw") (c-include "ncursesw/curses.h") […] (use gauche.collection) From the way the module system works, I'd expect a symbol exported by both the C libs and the "collection" module to be bound to the module in the last "use" clause, but these aren't the results I get: gosh> filter # gosh> (with-module gauche.collection filter) # gosh> (filter (lambda (i) (lisp? i)) '(scheme ruby cl haskell lolcat)))) *** ERROR: wrong number of arguments: filter requires 0, but got 2 ------------------------------------------------------------------------- This SF.Net email is sponsored by the Moblin Your Move Developer's challenge Build the coolest Linux based applications with Moblin SDK & win great prizes Grand prize is a trip for two to an Open Source event anywhere in the world http://moblin-contest.org/redirect.php?banner_id=100&url=/ _______________________________________________ Gauche-devel mailing list Gauche-devel< at >lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/gauche-devel