Block comments and `read-hash-extend'

Block comments and `read-hash-extend'

[Ludovic Courtès]

Hi,

The patch below (1) fixes `#! ... !#' block comments and (2) allows to
override them with `read-hash-extend'.


(1)  This example makes Guile 1.6 and the current Guile 1.7 hang (for
     some reason which I did not track down):

       guile> #! this is a comment !# (+ 2 2)
       [wait forever]

     Once the patch is applied:

       guile> #! this is a comment !# (+ 2 2)
       4

     I guess block comments have not been widely used over the past
     years.  ;-)


(2)  It makes it possible to discard the block comment syntax with
     `read-hash-extend'.  The example below allows to recognize the
     DSSSL keyword syntax[*], at the expense of losing the block
     comment syntax:

       (read-hash-extend #\! (lambda (chr port)
                               (symbol->keyword (read port))))

     I find this particular example useful since, even if DSSSL is
     apparently not widely used, other Scheme implementations such as
     Bigloo do support this syntax, and allowing such extensions is what
     `read-hash-extend' is for.


Comments?

Thanks,
Ludovic.

[*] http://www.ibiblio.org/pub/sun-info/standards/dsssl/dssslo/do960816.htm


2005-06-16  Ludovic Courtès  <ludovic.courtes@laas.fr>

	* read.c (scm_flush_ws):  Do not handle `#! .. !#' block
	comments here so that the behavior of `#!' can be extended with
	`read-hash-extend'.
	(scm_lreadr):  Handle the `#!' case here instead of in
	`scm_flush_ws ()' after making sure that no hash reader
	extension exists for `!'.  Jump to `tryagain' after call to
	`skip_scsh_block_comment ()', which fixes a bug in block
	comments handling (where Guile would hang).



Index: read.c
===================================================================
RCS file: /cvsroot/guile/guile/guile-core/libguile/read.c,v
retrieving revision 1.117
diff -u -B -b -p -r1.117 read.c
--- read.c	23 May 2005 19:57:21 -0000	1.117
+++ read.c	16 Jun 2005 09:13:28 -0000
@@ -227,9 +227,6 @@ scm_flush_ws (SCM port, const char *eofe
 	  case EOF:
 	    eoferr = "read_sharp";
 	    goto goteof;
-	  case '!':
-	    skip_scsh_block_comment (port);
-	    break;
 	  default:
 	    scm_ungetc (c, port);
 	    return '#';
@@ -478,9 +475,10 @@ scm_lreadr (SCM *tok_buf, SCM port, SCM 
 	  goto num;
 
 	case '!':
-	  /* should never happen, #!...!# block comments are skipped
-	     over in scm_flush_ws. */
-	  abort ();
+	  /* Only handle `#! ... !#' block comments if no user extension was
+	     defined for `!' using `read-hash-extend'.  */
+	  skip_scsh_block_comment (port);
+	  goto tryagain;
 
 	case '*':
 	  j = scm_read_token (c, tok_buf, port, 0);



_______________________________________________
Guile-devel mailing list
Guile-devel@gnu.org
http://lists.gnu.org/mailman/listinfo/guile-devel

Re: Block comments and `read-hash-extend'

[Kevin Ryde]

ludovic.courtes@laas.fr (Ludovic Courtès) writes:
>
> (1)  This example makes Guile 1.6 and the current Guile 1.7 hang (for
>      some reason which I did not track down):
>
>        guile> #! this is a comment !# (+ 2 2)
>        [wait forever]

You mean it reads more input?

>        guile> #! this is a comment !# (+ 2 2)
>        4

The manual says the !# is supposed to appear on a line on its own.  Do
you think you need it joined on?


_______________________________________________
Guile-devel mailing list
Guile-devel@gnu.org
http://lists.gnu.org/mailman/listinfo/guile-devel

Re: Block comments and `read-hash-extend'

[Ludovic Courtès]

Hi,

Kevin Ryde <user42@zip.com.au> writes:

> The manual says the !# is supposed to appear on a line on its own.  Do
> you think you need it joined on?

I think I should have read the manual more carefully.  ;-)  Indeed, `!#'
needs to be on a line on its own, so the bug I described above is not an
actual bug, it's a feature.

Still, I don't understand the rationale behind this and I consider this
a limitation.  In fact, it's misleading compared to block comments in
other languages or in SRFI-30.

Thanks,
Ludovic.


_______________________________________________
Guile-devel mailing list
Guile-devel@gnu.org
http://lists.gnu.org/mailman/listinfo/guile-devel

Re: Block comments and `read-hash-extend'

[Rob Browning]

ludovic.courtes@laas.fr (Ludovic Courtès) writes:

> Still, I don't understand the rationale behind this and I consider this
> a limitation.  In fact, it's misleading compared to block comments in
> other languages or in SRFI-30.

Unless there's a strong counterargument, I'd like to see #| |# work
in-line, in accordance with SRFI-30 and Common Lisp, and perhaps #! !#
should just be a synonym.

-- 
Rob Browning
rlb @defaultvalue.org and @debian.org; previously @cs.utexas.edu
GPG starting 2002-11-03 = 14DD 432F AE39 534D B592  F9A0 25C8 D377 8C7E 73A4


_______________________________________________
Guile-devel mailing list
Guile-devel@gnu.org
http://lists.gnu.org/mailman/listinfo/guile-devel

Re: Block comments and `read-hash-extend'

[Neil Jerram]

Ludovic Courtès wrote:
> Hi,
> 
> Kevin Ryde <user42@zip.com.au> writes:
> 
> 
>>The manual says the !# is supposed to appear on a line on its own.  Do
>>you think you need it joined on?
> 
> 
> I think I should have read the manual more carefully.  ;-)  Indeed, `!#'
> needs to be on a line on its own, so the bug I described above is not an
> actual bug, it's a feature.
> 
> Still, I don't understand the rationale behind this and I consider this
> a limitation.  In fact, it's misleading compared to block comments in
> other languages or in SRFI-30.

I agree with you.  I think the rationale for the status quo is actually
just historical accident, reflecting that the support for #! ... !# was
initially designed only for the `#! /usr/local/bin/guile' line at the
beginning of a script file.  But it surely much more useful for #! ...
!# comments to work as you suggest, and I can't see any downside from
making them do so.

	Neil


_______________________________________________
Guile-devel mailing list
Guile-devel@gnu.org
http://lists.gnu.org/mailman/listinfo/guile-devel

Re: Block comments and `read-hash-extend'

[Ludovic Courtès]

The patch below (<87k6kumykq.fsf@laas.fr>) also failed to live through
the summer.  ;-)

Is it acceptable?

Thanks,
Ludovic.

ludovic.courtes@laas.fr (Ludovic Courtès) writes:

> Hi,
>
> The patch below (1) fixes `#! ... !#' block comments and (2) allows to
> override them with `read-hash-extend'.
>
>
> (1)  This example makes Guile 1.6 and the current Guile 1.7 hang (for
>      some reason which I did not track down):
>
>        guile> #! this is a comment !# (+ 2 2)
>        [wait forever]
>
>      Once the patch is applied:
>
>        guile> #! this is a comment !# (+ 2 2)
>        4
>
>      I guess block comments have not been widely used over the past
>      years.  ;-)
>
>
> (2)  It makes it possible to discard the block comment syntax with
>      `read-hash-extend'.  The example below allows to recognize the
>      DSSSL keyword syntax[*], at the expense of losing the block
>      comment syntax:
>
>        (read-hash-extend #\! (lambda (chr port)
>                                (symbol->keyword (read port))))
>
>      I find this particular example useful since, even if DSSSL is
>      apparently not widely used, other Scheme implementations such as
>      Bigloo do support this syntax, and allowing such extensions is what
>      `read-hash-extend' is for.
>
>
> Comments?
>
> Thanks,
> Ludovic.
>
> [*] http://www.ibiblio.org/pub/sun-info/standards/dsssl/dssslo/do960816.htm
>
>
> 2005-06-16  Ludovic Courtès  <ludovic.courtes@laas.fr>
>
> 	* read.c (scm_flush_ws):  Do not handle `#! .. !#' block
> 	comments here so that the behavior of `#!' can be extended with
> 	`read-hash-extend'.
> 	(scm_lreadr):  Handle the `#!' case here instead of in
> 	`scm_flush_ws ()' after making sure that no hash reader
> 	extension exists for `!'.  Jump to `tryagain' after call to
> 	`skip_scsh_block_comment ()', which fixes a bug in block
> 	comments handling (where Guile would hang).
>
>
>
> Index: read.c
> ===================================================================
> RCS file: /cvsroot/guile/guile/guile-core/libguile/read.c,v
> retrieving revision 1.117
> diff -u -B -b -p -r1.117 read.c
> --- read.c	23 May 2005 19:57:21 -0000	1.117
> +++ read.c	16 Jun 2005 09:13:28 -0000
> @@ -227,9 +227,6 @@ scm_flush_ws (SCM port, const char *eofe
>  	  case EOF:
>  	    eoferr = "read_sharp";
>  	    goto goteof;
> -	  case '!':
> -	    skip_scsh_block_comment (port);
> -	    break;
>  	  default:
>  	    scm_ungetc (c, port);
>  	    return '#';
> @@ -478,9 +475,10 @@ scm_lreadr (SCM *tok_buf, SCM port, SCM 
>  	  goto num;
>  
>  	case '!':
> -	  /* should never happen, #!...!# block comments are skipped
> -	     over in scm_flush_ws. */
> -	  abort ();
> +	  /* Only handle `#! ... !#' block comments if no user extension was
> +	     defined for `!' using `read-hash-extend'.  */
> +	  skip_scsh_block_comment (port);
> +	  goto tryagain;
>  
>  	case '*':
>  	  j = scm_read_token (c, tok_buf, port, 0);
>
>
>
> _______________________________________________
> Guile-devel mailing list
> Guile-devel@gnu.org
> http://lists.gnu.org/mailman/listinfo/guile-devel


_______________________________________________
Guile-devel mailing list
Guile-devel@gnu.org
http://lists.gnu.org/mailman/listinfo/guile-devel

Re: Block comments and `read-hash-extend'

[Marius Vollmer]

Rob Browning <rlb@defaultvalue.org> writes:

> ludovic.courtes@laas.fr (Ludovic Courtès) writes:
>
>> Still, I don't understand the rationale behind this and I consider this
>> a limitation.  In fact, it's misleading compared to block comments in
>> other languages or in SRFI-30.
>
> Unless there's a strong counterargument, I'd like to see #| |# work
> in-line, in accordance with SRFI-30 and Common Lisp, and perhaps #! !#
> should just be a synonym.

Right now, HEAD already copes with in-line #! !# comments.  We could
copy it over to 1.6... Do you think this would be OK?

And yes, we should probably support #| |# as the primary block comment
syntax and make #! !# an option so that #! can be used as a read-hash
extension.  Patches welcome! :-)

-- 
GPG: D5D4E405 - 2F9B BCCC 8527 692A 04E3  331E FAF8 226A D5D4 E405


_______________________________________________
Guile-devel mailing list
Guile-devel@gnu.org
http://lists.gnu.org/mailman/listinfo/guile-devel

Re: Block comments and `read-hash-extend'

[Marius Vollmer]

ludovic.courtes@laas.fr (Ludovic Courtès) writes:

> Comments?

Your patch has the problem that it breaks things like

    (+ 2 2 #! 2 !#)

We can still support the #! read-hash extension, of course, which I
think is a good idea.

-- 
GPG: D5D4E405 - 2F9B BCCC 8527 692A 04E3  331E FAF8 226A D5D4 E405


_______________________________________________
Guile-devel mailing list
Guile-devel@gnu.org
http://lists.gnu.org/mailman/listinfo/guile-devel

Re: Block comments and `read-hash-extend'

[Rob Browning]

Marius Vollmer <mvo@zagadka.de> writes:

> Right now, HEAD already copes with in-line #! !# comments.  We could
> copy it over to 1.6... Do you think this would be OK?

So the only change to 1.6 would be that #! would work in-line (where
now they only work at the start of a line)?  If so, then I think that
would probably be OK.  I'd be surprised if anyone knew about and
depended on the current in-line behavior.

-- 
Rob Browning
rlb @defaultvalue.org and @debian.org; previously @cs.utexas.edu
GPG starting 2002-11-03 = 14DD 432F AE39 534D B592  F9A0 25C8 D377 8C7E 73A4


_______________________________________________
Guile-devel mailing list
Guile-devel@gnu.org
http://lists.gnu.org/mailman/listinfo/guile-devel