Development of CLISP

james anderson | 3 Feb 2004 18:31

Picon

Re: method combinations / cl-bench / ansi-tests


On Tuesday, Feb 3, 2004, at 17:30 Europe/Berlin, Sam Steingold wrote:

>> * james anderson <wnzrf.naqrefba <at> frgs.qr> [2004-02-03 16:28:48 +0100]:
>>
>> On Tuesday, Feb 3, 2004, at 15:06 Europe/Berlin, Paul F. Dietz wrote:
>>
>>> james anderson wrote:
>>>
>>>> define-method-combination-long-form.lsp
>>>>    a replacement for the respective ansi-tests file to exercise more
>>>
>>> Thanks!  I have been meaning to write long form tests, but hadn't
>>> gotten around to it yet.
>>>
>>> I assume this is licensable under GPL, so I can put it into
>>> gcl/ansi-tests?
>>
>> please. i didn't find any notices in the other test files proper, so i
>> added just a line with my mail address, and presumed that
>> COPYING.LIB-2.0 would apply.
>
> just a small nit pick: actually, CLISP is under GPL2, so, by default,
> everything that comes with it is GPL2.  Since you are the author, you,
> of course, are free to contribute it to CLISP under GPL2 and to GCL (of
> which ansi-tests are a part) under LGPL2.

the COPYING.LIB-2.0 reference was in response to p.dietz's above  
enquiry about the define-method-combination-long-form.lisp file, as  
that should end up among ansi-tests, which appear(s) to be under GCL's

(Continue reading)

Sam Steingold | 3 Feb 2004 21:36

Picon

Re: method combinations / cl-bench / ansi-tests

> * james anderson <wnzrf.naqrefba <at> frgs.qr> [2004-02-03 18:31:59 +0100]:
>
> ? if the one i noted in the message from this morning
>
> http://home.arcor.de/james_anderson/setf/clisp/20040203/clos-20040203.diff

I am sorry, I did not see this e-mail.

> is not what you need, please tell me the diff command which will
> generate something which you can use. this one was generated
> differently than the clos-20040202.diff which i mailed yesterday. this
> time, by
>
>    diff -u clos-20040202.lisp clos.lisp > clos-20040203.diff
>
> where the clos-20040202.lisp file is from yesterday's cvs.

this is _very_ good, but, unfortunately, it includes whitespace
changes, so it is too big (and thus hard to read).

could you please do

$ cvs diff -ub clos.lisp

thanks!

--

-- 
Sam Steingold (http://www.podval.org/~sds) running w2k
<http://www.camera.org> <http://www.iris.org.il> <http://www.memri.org/>
<http://www.mideasttruth.com/> <http://www.honestreporting.com>

(Continue reading)

james anderson | 3 Feb 2004 21:54

Picon

Re: method combinations


On Tuesday, Feb 3, 2004, at 21:36 Europe/Berlin, Sam Steingold wrote:

>> ...
>
> could you please do
>
> $ cvs diff -ub clos.lisp
>

http://home.arcor.de/james_anderson/setf/clisp/20040203/clos- 
20040203b.diff

...

-------------------------------------------------------
The SF.Net email is sponsored by EclipseCon 2004
Premiere Conference on Open Tools Development and Integration
See the breadth of Eclipse activity. February 3-5 in Anaheim, CA.
http://www.eclipsecon.org/osdn

Sam Steingold | 4 Feb 2004 00:04

Picon

Re: method combinations

> * james anderson <wnzrf.naqrefba <at> frgs.qr> [2004-02-03 21:54:57 +0100]:
>> $ cvs diff -ub clos.lisp
> http://home.arcor.de/james_anderson/setf/clisp/20040203/clos- 
> 20040203b.diff

OK, thanks, now we are in business.

I applied this patch (with some changes, mostly formatting).

Notes:

1. you do _NOT_ need MAKE-LOAD-FORM because METHOD-COMBINATION is a
   struct and will be saved in the #S() format (faster)

2. this test still fails:

Form:
(progn
  (defun positive-integer-qualifier-p (method-qualifiers)
    (and (= (length method-qualifiers) 1)
         (typep (first method-qualifiers) '(integer 0 *))))
  (define-method-combination example-method-combination ()
    ((method-list positive-integer-qualifier-p))
    `(progn , <at> (mapcar #'(lambda (method) `(call-method ,method))
                      (stable-sort method-list #'<
                                   :key #'(lambda (method)
                                            (first (method-qualifiers
                                                    method)))))))
  (defgeneric mc-test-piq (p1 p2 s)
    (:method-combination example-method-combination)

(Continue reading)

james anderson | 4 Feb 2004 18:27

Picon

Re: method combinations

>
> 4. COMPUTE-EFFECTIVE-METHOD-FUNCTION: is it possible to use backquote
>    instead of (list 'function (list 'lambda ...))
>

in theory yes. the changes you had made were similar to my first 
attempt to code that passage. i find it very hard to code, or to read, 
or to debug double-back-quoted code. some people have the hang of it. i 
don't. your changes caused problems similar to my also 
double-back-quoted initial version: some of the symbols don't get 
evaluated in the intended context. i put back in the version i can read.

...

-------------------------------------------------------
The SF.Net email is sponsored by EclipseCon 2004
Premiere Conference on Open Tools Development and Integration
See the breadth of Eclipse activity. February 3-5 in Anaheim, CA.
http://www.eclipsecon.org/osdn

james anderson | 4 Feb 2004 18:22

Picon

Re: method combinations


On Wednesday, Feb 4, 2004, at 00:04 Europe/Berlin, Sam Steingold wrote:
>
> 2. this test still fails:
>
> Form:
> (progn
>   (defun positive-integer-qualifier-p (method-qualifiers)
>     (and (= (length method-qualifiers) 1)
>          (typep (first method-qualifiers) '(integer 0 *))))
>   (define-method-combination example-method-combination ()
>     ((method-list positive-integer-qualifier-p))
>     `(progn , <at> (mapcar #'(lambda (method) `(call-method ,method))
>                       (stable-sort method-list #'<
>                                    :key #'(lambda (method)
>                                             (first (method-qualifiers
>                                                     method)))))))
>   (defgeneric mc-test-piq (p1 p2 s)
>     (:method-combination example-method-combination)
>     (:method 1 ((p1 t) (p2 t) s) (vector-push-extend (list 1 p1 p2) s))
>     (:method 2 ((p1 t) (p2 t) s) (vector-push-extend (list 2 p1 p2) s))
>     (:method 3 ((p1 t) (p2 t) s) (vector-push-extend (list 3 p1 p2) 
> s)))
>   (let ((s (make-array 10 :adjustable t :fill-pointer 0)))
>     (mc-test-piq 1 2 s)
>     s))
> CORRECT: #((1 1 2) (2 1 2) (3 1 2))
> CLISP  : ERROR
> GET: 1 is not a SYMBOL

(Continue reading)

Sam Steingold | 5 Feb 2004 17:17

Picon

Re: method combinations

> * james anderson <wnzrf.naqrefba <at> frgs.qr> [2004-02-04 18:22:09 +0100]:
>
> --- clos-20040203.lisp	Tue Feb  3 11:52:27 2004
> +++ clos.lisp	Wed Feb  4 18:02:55 2004
>  <at>  <at>  -1633,10 +1633,11  <at>  <at> 
>                  `(, <at> (subseq lambda-list 0 index) &ALLOW-OTHER-KEYS
>                    , <at> (subseq lambda-list index)))))
>            (let* ((self (gensym))
> -                 (wants-next-method-p
> +                 #|(wants-next-method-p
>                    (or (method-combination-object (car qualifiers)
>                                                   :if-does-not-exist nil)
> -                      (equal qualifiers '()) (equal qualifiers '(:around))))
> +                      (equal qualifiers '()) (equal qualifiers '(:around))))|#
> +                 (wants-next-method-p t)
>                   (compile nil)
>                   (lambdabody
>                     (multiple-value-bind (body-rest declarations docstring)

with this patch, the number of failed regression tests goes through the
roof.

--

-- 
Sam Steingold (http://www.podval.org/~sds) running w2k
<http://www.camera.org> <http://www.iris.org.il> <http://www.memri.org/>
<http://www.mideasttruth.com/> <http://www.honestreporting.com>
In every non-trivial program there is at least one bug.

-------------------------------------------------------
The SF.Net email is sponsored by EclipseCon 2004

(Continue reading)

james anderson | 9 Feb 2004 21:03

Picon

Re: method combinations

nb. re today's patch. if one applies it in a running image, one must 
also reevaluate the next expression in clos.lisp:

(setf (method-combination-expander (method-combination-object 
'standard))
       #'standard-method-combination-expander)

in order for it to take effect.

...

-------------------------------------------------------
The SF.Net email is sponsored by EclipseCon 2004
Premiere Conference on Open Tools Development and Integration
See the breadth of Eclipse activity. February 3-5 in Anaheim, CA.
http://www.eclipsecon.org/osdn

james anderson | 9 Feb 2004 20:44

Picon

Re: method combinations

On Thursday, Feb 5, 2004, at 17:17 Europe/Berlin, Sam Steingold wrote:

>> * james anderson <wnzrf.naqrefba <at> frgs.qr> [2004-02-04 18:22:09 +0100]:
>>
>> --- clos-20040203.lisp	Tue Feb  3 11:52:27 2004
>> +++ clos.lisp	Wed Feb  4 18:02:55 2004
>> ...
>
> with this patch, the number of failed regression tests goes through the
> roof.

there are two clauses in the standard combination effective method 
computation which had hard-wired the next-method handling. this changes 
them to look at the method. at the moment a next-method context is 
always required, but if one leaves the test in despite that, one could 
eventually allow defgeneric to optimize for a case when it defines the 
entire "known" function.

Attachment (clos-20040204,09.diff): application/octet-stream, 1867 bytes

james anderson | 4 Feb 2004 17:32

Picon

Re: method combinations


On Wednesday, Feb 4, 2004, at 00:04 Europe/Berlin, Sam Steingold wrote:

>> * james anderson <wnzrf.naqrefba <at> frgs.qr> [2004-02-03 21:54:57 +0100]:
>>> $ cvs diff -ub clos.lisp
>> http://home.arcor.de/james_anderson/setf/clisp/20040203/clos-
>> 20040203b.diff
>
> OK, thanks, now we are in business.
>
> I applied this patch (with some changes, mostly formatting).
>
> Notes:
>
> 1. you do _NOT_ need MAKE-LOAD-FORM because METHOD-COMBINATION is a
>    struct and will be saved in the #S() format (faster)

i would have thought so. until i observed that i an error was being 
signaled when i tried to compile a file which had a literal 
method-combination structure instance as part of some other value which 
the file compiler was working on.
- i didn't try to get to the bottom of the circumstance. it was easier 
to just provide the load form method.
- i recall that i had the entire clisp bullt at that point, so i don't 
think it was an issue of an incomplete runtime at that moment.
- the method-combination structure didn't really have any business 
being there and i think i've fixed that since.
- i never did try compiling them explicitly as there's no real reason 
to.

(Continue reading)

Return

Return to gmane.lisp.clisp.devel.

Project Web Page

Development of CLISP

Search Archive

Language

Change language

Options

Current view: Threads only / Showing only 20 lines / Not hiding cited text.
Change to All messages, whole messages, or hide cited text.

Post a message
NNTP Newsgroup
Classic Gmane web interface
XML

XML

RSS Feed
List Information

About Gmane