GNU bug report logs - #9567
curious match bug (?)

Previous Next

To add a comment to this bug, you must first unarchive it, by sending
a message to control AT debbugs.gnu.org, with unarchive 9567 in the body.
You can then email your comments to 9567 AT debbugs.gnu.org in the normal way.

Toggle the display of automated, internal messages from the tracker.

Message #5 received at submit <at> debbugs.gnu.org (full text, mbox):

From: Andy Wingo <wingo <at> pobox.com>
To: bug-guile <bug-guile <at> gnu.org>
Subject: curious match bug (?)
Date: Wed, 21 Sep 2011 05:34:36 +0200

Hi,

Try this:

  (use-modules (language tree-il) (ice-9 match))
  (define foo (parse-tree-il '(let-values (apply (lambda () (lambda-case ((() #f #f #f () ()) (apply (primitive values) (const 1) (const 2)))))) (lambda-case (((a b) #f #f #f () (#{a 134390}# #{b 134391}#)) (apply (primitive list) (lexical a #{a 134390}#) (lexical b #{b 134391}#)))))))
  (match foo
    (($ <let-values> src exp
        ($ <lambda-case> src2 req #f #f #f () gensyms body #f))
     #t)
    (_
     #f))
  => #t
                         
  (match foo
    (($ <let-values> src foo ;; <- rename "exp" to "foo"
        ($ <lambda-case> src2 req #f #f #f () gensyms body #f))
     #t)
    (_
     #f))
  => #f

I tried to reduce this case a bit, but didn't succeed directly, and I
need to move on.  But what is the deal here?

Andy
-- 
http://wingolog.org/

Message #8 received at 9567 <at> debbugs.gnu.org (full text, mbox):

From: Stefan Israelsson Tampe <stefan.itampe <at> gmail.com>
To: 9567 <at> debbugs.gnu.org, Andy Wingo <wingo <at> pobox.com>
Subject: Re: bug#9567: curious match bug (?)
Date: Wed, 21 Sep 2011 14:15:53 +0200

[Message part 1 (text/plain, inline)]

This is an old bug I spotted before, it can be solved by inserting an extra
let
in the expansion aka

(macroexpand '(match a (par code  ...))
-> '(let ((arg a)) ....)

This is missing from atoms in match so I added that there and the
missbehavior dissapears. see the git diffed patch in this post.


On Wed, Sep 21, 2011 at 5:34 AM, Andy Wingo <wingo <at> pobox.com> wrote:

> Hi,
>
> Try this:
>
>  (use-modules (language tree-il) (ice-9 match))
>  (define foo (parse-tree-il '(let-values (apply (lambda () (lambda-case
> ((() #f #f #f () ()) (apply (primitive values) (const 1) (const 2))))))
> (lambda-case (((a b) #f #f #f () (#{a 134390}# #{b 134391}#)) (apply
> (primitive list) (lexical a #{a 134390}#) (lexical b #{b 134391}#)))))))
>  (match foo
>    (($ <let-values> src exp
>        ($ <lambda-case> src2 req #f #f #f () gensyms body #f))
>     #t)
>    (_
>     #f))
>  => #t
>
>  (match foo
>    (($ <let-values> src foo ;; <- rename "exp" to "foo"
>        ($ <lambda-case> src2 req #f #f #f () gensyms body #f))
>     #t)
>    (_
>     #f))
>  => #f
>
> I tried to reduce this case a bit, but didn't succeed directly, and I
> need to move on.  But what is the deal here?
>
> Andy
> --
> http://wingolog.org/
>
>
>
>

[Message part 2 (text/html, inline)]

[match-bug.patch (text/x-patch, attachment)]

Message #11 received at 9567 <at> debbugs.gnu.org (full text, mbox):

From: ludo <at> gnu.org (Ludovic Courtès)
To: Andy Wingo <wingo <at> pobox.com>
Cc: 9567 <at> debbugs.gnu.org
Subject: Re: bug#9567: curious match bug (?)
Date: Fri, 23 Sep 2011 16:45:43 +0200

Could it be a feature?

  (let ((pat '('a _ ...)))
    (match '(a b c) (pat #t)))
  => #t

Ludo'.

Message #14 received at 9567 <at> debbugs.gnu.org (full text, mbox):

From: Stefan Israelsson Tampe <stefan.itampe <at> gmail.com>
To: Ludovic Courtès <ludo <at> gnu.org>, 9567 <at> debbugs.gnu.org, 
	wingo <at> pobox.com
Subject: Re: bug#9567: curious match bug (?)
Date: Fri, 23 Sep 2011 18:24:30 +0200

[Message part 1 (text/plain, inline)]

On the other hand

(let ((pat '(a _ ...)))
   (match '(x y z) (pat #t)))
$1 = #t

as well,

a more canonical example of this problem (where I saw it first)
are when using 'and' e.g.

(define a '(1 2))

(match a ((and (a 2) (1 b)) (+ a b)) (_ #f))
-> #f

But with my fix you wil get the correct 3.

Now the reson are that for 'and' as well as for '$' another location a is
under the command
of an implicit (let ((a car)) ...). One could perhaps argue that this is a
bug in syntax handling but I would not think so. Anyway maybe we should
consider moving this fix upstream.

/Stefan
2011/9/23 Ludovic Courtès <ludo <at> gnu.org>

> Could it be a feature?
>
>  (let ((pat '('a _ ...)))
>    (match '(a b c) (pat #t)))
>  => #t
>
> Ludo'.
>
>
>
>

[Message part 2 (text/html, inline)]

Message #17 received at 9567 <at> debbugs.gnu.org (full text, mbox):

From: Andy Wingo <wingo <at> pobox.com>
To: ludo <at> gnu.org (Ludovic Courtès)
Cc: 9567 <at> debbugs.gnu.org
Subject: Re: bug#9567: curious match bug (?)
Date: Sat, 24 Sep 2011 16:51:00 +0200

On Fri 23 Sep 2011 16:45, ludo <at> gnu.org (Ludovic Courtès) writes:

> Could it be a feature?
>
>   (let ((pat '('a _ ...)))
>     (match '(a b c) (pat #t)))
>   => #t

`match' compiles patterns to Scheme code at macro expansion time.  It
doesn't do runtime matching.  Are you thinking that `pat' is being used
as a pattern?  It is not:

  (let ((pat '('b _ ...)))
    (match '(a b c) (pat #t)))
  => #t
  
If it were a pattern, it would not match.  It is a bug IMO.

Andy
-- 
http://wingolog.org/

Message #20 received at 9567 <at> debbugs.gnu.org (full text, mbox):

From: Andy Wingo <wingo <at> pobox.com>
To: Alex Shinn <alexshinn <at> gmail.com>
Cc: 9567 <at> debbugs.gnu.org
Subject: `match' bug ?
Date: Sat, 24 Sep 2011 17:01:37 +0200

Hi Alex,

We are getting the following bug in Guile:

    > (use-modules (ice-9 match))

This imports your `match' library.

    > (define a '(1))
    > (match a (((and a 1)) a) (_ #f))
    1

Here we destructured the first element from '(1).

    > (define a '(1 2))
    > (match a ((and (a 2) (1 b)) (+ a b)) (_ #f))
    #f

Now we are trying to destructure the first two elements from '(1 2).
But it doesn't work!  OTOH it does work if we rename the pattern vars:

    > (match a ((and (x 2) (1 y)) (+ x y)) (_ #f))
    3

Can you reproduce this bug on other Schemes?  Could it be that the code
to extract vars to be bound is erroneously propagating the var bound to
the input value?  Or is it a bug in Guile?

Stefan Israelsson Tampe reports that the following redefinition of
`match' fixes the problem for him:

  (define-syntax match
    (syntax-rules ()
      ((match)
       (match-syntax-error "missing match expression"))
      ((match atom)
       (match-syntax-error "no match clauses"))
      ((match (app ...) (pat . body) ...)
       (let ((v (app ...)))
         (match-next v ((app ...) (set! (app ...))) (pat . body) ...)))
      ((match #(vec ...) (pat . body) ...)
       (let ((v #(vec ...)))
         (match-next v (v (set! v)) (pat . body) ...)))
      ((match atom (pat . body) ...)
       (let ((v atom))
         (match-next v (atom (set! atom)) (pat . body) ...)))))

As you see the last case introduces a `let'.

Regards,

Andy
-- 
http://wingolog.org/

Message #23 received at 9567 <at> debbugs.gnu.org (full text, mbox):

From: Alex Shinn <alexshinn <at> gmail.com>
To: Andy Wingo <wingo <at> pobox.com>
Cc: 9567 <at> debbugs.gnu.org
Subject: Re: `match' bug ?
Date: Sun, 25 Sep 2011 15:59:17 +0900

Hi Andy,

On Sun, Sep 25, 2011 at 12:01 AM, Andy Wingo <wingo <at> pobox.com> wrote:
>
> We are getting the following bug in Guile:
> [...]
>
>    > (define a '(1 2))
>    > (match a ((and (a 2) (1 b)) (+ a b)) (_ #f))
>    #f

Yes, that's a bug - confirmed in Chibi.  The
diagnosis is also correct.  I've applied the
patch and updated the synthcode match.scm
and the Chibi repo.

Thanks Stefan!

-- 
Alex

Message #26 received at 9567 <at> debbugs.gnu.org (full text, mbox):

From: ludo <at> gnu.org (Ludovic Courtès)
To: Alex Shinn <alexshinn <at> gmail.com>
Cc: Andy Wingo <wingo <at> pobox.com>, 9567 <at> debbugs.gnu.org
Subject: Re: bug#9567: `match' bug ?
Date: Sat, 15 Oct 2011 16:07:20 +0200

Hi Alex,

Alex Shinn <alexshinn <at> gmail.com> skribis:

> On Sun, Sep 25, 2011 at 12:01 AM, Andy Wingo <wingo <at> pobox.com> wrote:
>>
>> We are getting the following bug in Guile:
>> [...]
>>
>>    > (define a '(1 2))
>>    > (match a ((and (a 2) (1 b)) (+ a b)) (_ #f))
>>    #f
>
> Yes, that's a bug - confirmed in Chibi.  The
> diagnosis is also correct.  I've applied the
> patch and updated the synthcode match.scm
> and the Chibi repo.

I’m trying to update Guile’s copy from Chibi but changeset
876:528cdab3f818 in the default branch doesn’t seem to contain the fix.

What am I missing?

Thanks,
Ludo’.

Message #29 received at 9567 <at> debbugs.gnu.org (full text, mbox):

From: Alex Shinn <alexshinn <at> gmail.com>
To: Ludovic Courtès <ludo <at> gnu.org>
Cc: Andy Wingo <wingo <at> pobox.com>, 9567 <at> debbugs.gnu.org
Subject: Re: bug#9567: `match' bug ?
Date: Sun, 16 Oct 2011 16:28:30 +0900

Hi Ludovic,

2011/10/15 Ludovic Courtès <ludo <at> gnu.org>:
>
> I’m trying to update Guile’s copy from Chibi but changeset
> 876:528cdab3f818 in the default branch doesn’t seem to contain the fix.
>
> What am I missing?

"hg export 851" will show you the changes, including
the new test case.  Does it not work in Guile?

On the off chance you're unfamiliar with mercurial,
did you remember to "hg update" after fetching the
changes?

-- 
Alex

Message #34 received at 9567-done <at> debbugs.gnu.org (full text, mbox):

From: ludo <at> gnu.org (Ludovic Courtès)
To: Alex Shinn <alexshinn <at> gmail.com>
Cc: Andy Wingo <wingo <at> pobox.com>, 9567-done <at> debbugs.gnu.org
Subject: Re: bug#9567: `match' bug ?
Date: Sun, 16 Oct 2011 18:38:52 +0200

Hi Alex,

Alex Shinn <alexshinn <at> gmail.com> skribis:

> 2011/10/15 Ludovic Courtès <ludo <at> gnu.org>:
>>
>> I’m trying to update Guile’s copy from Chibi but changeset
>> 876:528cdab3f818 in the default branch doesn’t seem to contain the fix.
>>
>> What am I missing?
>
> "hg export 851" will show you the changes, including
> the new test case.  Does it not work in Guile?

Yes.

> On the off chance you're unfamiliar with mercurial,
> did you remember to "hg update" after fetching the
> changes?

Oh indeed, I had run ‘hg pull’ (I think) but not ‘hg update’.

Thanks!

Ludo’.

Previous Next

GNU bug tracking system
Copyright (C) 1999 Darren O. Benham, 1997,2003 nCipher Corporation Ltd, 1994-97 Ian Jackson.