[agj@alum.mit.edu: Re: ":" in Scheme names.]

[agj@alum.mit.edu: Re: ":" in Scheme names.]

[Aubrey Jaffer]

I have heard nothing, so I assume that "info.el" has not been fixed.

------- Start of forwarded message -------
From: Aubrey Jaffer <agj@alum.mit.edu>
To: karl@freefriends.org
Cc: eliz@is.elta.co.il, bug-texinfo@gnu.org
In-reply-to: <200211071419.gA7EJxX06324@f7.net> (karl@freefriends.org)
Subject: Re: ":" in Scheme names.
Date: Thu,  7 Nov 2002 21:16:12 -0500 (EST)

With this change, Emacs info-mode correctly indexes colon-containing
names in (existing) info files.  Does someone on bug-texinfo@gnu.org
maintain "info.el" or should I send this to bug-emacs@gnu.org?

cd /usr/share/emacs/21.2/lisp/
diff -c /usr/share/emacs/21.2/lisp/info.el\~ /usr/share/emacs/21.2/lisp/info.el
*** /usr/share/emacs/21.2/lisp/info.el~	Fri Dec 21 14:08:17 2001
- --- /usr/share/emacs/21.2/lisp/info.el	Wed Nov  6 22:30:47 2002
***************
*** 1830,1836 ****
    (interactive "sIndex topic: ")
    (let ((orignode Info-current-node)
  	(rnode nil)
! 	(pattern (format "\n\\* +\\([^\n:]*%s[^\n:]*\\):[ \t]*\\([^.\n]*\\)\\.[ \t]*\\([0-9]*\\)"
  			 (regexp-quote topic)))
  	node
  	(case-fold-search t))
- --- 1830,1836 ----
    (interactive "sIndex topic: ")
    (let ((orignode Info-current-node)
  	(rnode nil)
! 	(pattern (format "\n\\* +\\([^\n]*%s[^\n]*\\):[ \t]+\\([^.\n]*\\)\\.[ \t]*\\([0-9]*\\)"
  			 (regexp-quote topic)))
  	node
  	(case-fold-search t))

Diff finished at Wed Nov  6 22:32:26
------- End of forwarded message -------

Re: [agj@alum.mit.edu: Re: ":" in Scheme names.]

[Thien-Thi Nguyen]

Aubrey Jaffer <agj@alum.mit.edu> writes:

   I have heard nothing, so I assume that "info.el" has not been fixed.
   [patch]

i have applied the patch locally; testing (on /usr/share/info/slib.info.gz
no less) still misbehaves, however.  i see there are several additional
places in (the current cvs) info.el where ":" is taken to be end-of-tag.

thi

Re: [agj@alum.mit.edu: Re: ":" in Scheme names.]

[Thien-Thi Nguyen]

   From: Aubrey Jaffer <agj@alum.mit.edu>
   Date: Sun, 19 Jan 2003 16:59:45 -0500 (EST)

   If the patch improves some (but not all) behaviors, then it is still
   worth releasing.

yes.  which behaviors does it improve?  what goes in the change log?

   If you show me how to provoke the remaining misbehaviors, then I can
   look at fixing them.

here is an interactive test case, w/ expected output:

evaluate: (info "/usr/share/info/slib.tar.gz")

then, 
; type			; expect
  i RET			  buffer changes to show index
  C-s array-indexes RET   point moves to match
  m RET			  buffer changes to show array mapping funcs
  l			  back to index
  C-s batch: RET          point moves to match
  m RET			  buffer changes to show batch funcs
  l			  back to index

this sequence exercises several parsing contexts, to start; from here
you can try other `m' and `i' variants.

for me, not all of the expected behaviors take place.  some signal
end-of-file error.  (after applying the patch i started looking at all
functions that have ":" in their regexp strings to see if they make
similar assumptions (that this patch fixes for this function) about tag
charset; i'm not sure what i'm working with exactly anymore, ymmv. :-)

the constraint "must not contain colon followed by space" is easy for
tags comprising programming language terms to meet w/o special handling,
and it is more permissive than the current "must not contain colon", so
as a fix it is ok.  but i wonder if there is a better fix.

thi

Re: [agj@alum.mit.edu: Re: ":" in Scheme names.]

[Aubrey Jaffer]

 | From: Thien-Thi Nguyen <ttn@glug.org>
 | Date: Sun, 19 Jan 2003 21:44:03 -0500
 | 
 |    From: Aubrey Jaffer <agj@alum.mit.edu>
 |    Date: Sun, 19 Jan 2003 16:59:45 -0500 (EST)
 | 
 |    If the patch improves some (but not all) behaviors, then it is still
 |    worth releasing.
 | 
 | yes.  which behaviors does it improve?  what goes in the change log?

	(Info-index): Fixed for index names containing colons.

 |    If you show me how to provoke the remaining misbehaviors, then I can
 |    look at fixing them.
 | 
 | here is an interactive test case, w/ expected output:
 | 
 | evaluate: (info "/usr/share/info/slib.tar.gz")
 | 
 | then, 
 | ; type			; expect
 |   i RET			  buffer changes to show index
 |   C-s array-indexes RET   point moves to match
 |   m RET			  buffer changes to show array mapping funcs
 |   l			  back to index
 |   C-s batch: RET          point moves to match
 |   m RET			  buffer changes to show batch funcs

Appended is a patch to Info-menu so it grabs the correct default.
Also patched is Info-extract-menu-item so it goes to the ": ".  But I
still get an error from Info-find-node-2 with the wrong value for
nodename:

  let: No such anchor in tag table or node in tag table or file: modify-table: HTML Tables

Somewhere there is a forward search to ":"; but I lost the trail of
where this happens to nodename.

Here is an unrelated bug in Info-follow-nearest-node:
evaluate: (info "/usr/share/info/slib.tar.gz")
then, 
; type			; expect
  i RET			  buffer changes to show index
  C-s array-indexes RET   point moves to match
  RET			  buffer changes to show array mapping funcs
  l			  back to index
  RET			  buffer changes to Getopt !!

			      -=-=-=-=-

cd /usr/share/emacs/21.2/lisp/
diff -c /usr/share/emacs/21.2/lisp/info.el\~ /usr/share/emacs/21.2/lisp/info.el
*** /usr/share/emacs/21.2/lisp/info.el~	Sun Jan 19 17:31:53 2003
--- /usr/share/emacs/21.2/lisp/info.el	Tue Jan 21 00:33:40 2003
***************
*** 1477,1483 ****
  	    (save-excursion
  	      (goto-char p)
  	      (end-of-line)
! 	      (if (re-search-backward "\n\\* +\\([^:\t\n]*\\):" beg t)
  		  (setq default (match-string-no-properties 1))))))
       (let ((item nil))
         (while (null item)
--- 1477,1483 ----
  	    (save-excursion
  	      (goto-char p)
  	      (end-of-line)
! 	      (if (re-search-backward "\n\\* +\\([^\t\n]+\\): " beg t)
  		  (setq default (match-string-no-properties 1))))))
       (let ((item nil))
         (while (null item)
***************
*** 1509,1515 ****
  	(goto-char (point-min))
  	(or (search-forward "\n* menu:" nil t)
  	    (error "No menu in this node"))
! 	(or (re-search-forward (concat "\n\\* +" menu-item ":") nil t)
  	    (re-search-forward (concat "\n\\* +" menu-item) nil t)
  	    (error "No such item in menu"))
  	(beginning-of-line)
--- 1509,1515 ----
  	(goto-char (point-min))
  	(or (search-forward "\n* menu:" nil t)
  	    (error "No menu in this node"))
! 	(or (re-search-forward (concat "\n\\* +" menu-item ": ") nil t)
  	    (re-search-forward (concat "\n\\* +" menu-item) nil t)
  	    (error "No such item in menu"))
  	(beginning-of-line)

Diff finished at Tue Jan 21 00:33:44

Re: [agj@alum.mit.edu: Re: ":" in Scheme names.]

[Thien-Thi Nguyen]

there are some deeper changes necessary to properly encode the
":"-in-menu-item cases, which i am looking into.  for now, i have just
committed the patch below, which fixes "m" handling for slib.info.gz test
case:

; type                       ; expect
  i RET                        buffer changes to show index
  C-s array-indexes RET        point moves to match
  m RET                        buffer changes to show array mapping funcs
  l                            back to index
  C-s batch: RET               point moves to match
  m RET                        buffer changes to show batch funcs

there are still bugs in areas: completion for index, TAB command, font
locking.  i will get to them shortly.

thi


__________________________________________________
cd /home/ttn/build/GNU/emacs/lisp/
cvs diff -w -c -b info.el
Index: info.el
===================================================================
RCS file: /cvsroot/emacs/emacs/lisp/info.el,v
retrieving revision 1.331
diff -w -c -b -r1.331 info.el
*** info.el	22 Jan 2003 20:35:23 -0000	1.331
--- info.el	25 Jan 2003 00:33:45 -0000
***************
*** 27,32 ****
--- 27,36 ----
  ;;; Commentary:
  
  ;; Note that nowadays we expect info files to be made using makeinfo.
+ ;; In particular we make these assumptions:
+ ;;  - a menu item MAY contain colons but not colon-space ": "
+ ;;  - a menu item ending with ": " (but not ":: ") is an index entry
+ ;;  - a node name MAY NOT contain a colon
  
  ;;; Code:
  
***************
*** 2119,2126 ****
--- 2123,2140 ----
       ((setq node (Info-get-token (point) "\\*note[ \n]"
  				 "\\*note[ \n]\\([^:]*\\):"))
        (Info-follow-reference node))
+      ;; explicit node name
       ((setq node (Info-get-token (point) "\\* +" "\\* +\\([^:]*\\)::"))
        (Info-goto-node node))
+      ;; index entry (Another approach is to combine this w/ the following cond
+      ;;              branch "other menu item", but that also requires fixing
+      ;;              Info-extract-menu-node-name -- stay tuned. --ttn)
+      ((Info-get-token (point) "\\* +" "\\* +\\(.*\\):[ \t]")
+       (save-excursion
+         (re-search-forward ":[ \t]")
+         (setq node (Info-following-node-name "^.")))
+       (Info-goto-node node))
+      ;; other menu item
       ((Info-get-token (point) "\\* +" "\\* +\\([^:]*\\):")
        (beginning-of-line)
        (forward-char 2)

Compilation exited abnormally with code 1 at Fri Jan 24 16:32:01

Re: [agj@alum.mit.edu: Re: ":" in Scheme names.]

[Aubrey Jaffer]

 | From: Thien-Thi Nguyen <ttn@glug.org>
 | Date: 24 Jan 2003 20:03:06 -0500
 | 
 | there are some deeper changes necessary to properly encode the
 | ":"-in-menu-item cases, which i am looking into.  for now, i have
 | just committed the patch below, which fixes "m" handling for
 | slib.info.gz test case:

Thanks!

 | ; type                       ; expect
 |   i RET                        buffer changes to show index
 |   C-s array-indexes RET        point moves to match
 |   m RET                        buffer changes to show array mapping funcs
 |   l                            back to index
 |   C-s batch: RET               point moves to match
 |   m RET                        buffer changes to show batch funcs

With the patch, the last step gives me an error:

 let: No such anchor in tag table or node in tag table or file: call-with-output-script: Batch

But it is evident that Info-try-follow-nearest-node is fixed if the
last line is instead:

 |   RET                        buffer changes to show batch funcs

Re: [agj@alum.mit.edu: Re: ":" in Scheme names.]

[Thien-Thi Nguyen]

   From: Aubrey Jaffer <agj@alum.mit.edu>
   Date: Fri, 24 Jan 2003 21:37:04 -0500 (EST)

    let: No such anchor in tag table or node in tag table or file:
    call-with-output-script: Batch

   But it is evident that Info-try-follow-nearest-node is fixed if the
   last line is instead:

    |   RET                        buffer changes to show batch funcs

oops, yes this is a mistake on my part in reporting the consequences of
the fix.  the good news is that the "m RET" case is now fixed (revision
1.333 in cvs), as well.

w/ 1.333 i ran around slib.info and dired-x.info using "RET", "m", "l"
and various "i" (default/exact/inexact input) commands w/o problems.
please let me know if you find any new (and interesting) bugs, otherwise
i'm inclined to look at other things now.  M-x viva-scheme!

thi

Re: [agj@alum.mit.edu: Re: ":" in Scheme names.]

[Kim F. Storm]

Thien-Thi Nguyen <ttn@giblet.glug.org> writes:

>    From: Aubrey Jaffer <agj@alum.mit.edu>
>    Date: Fri, 24 Jan 2003 21:37:04 -0500 (EST)
> 
>     let: No such anchor in tag table or node in tag table or file:
>     call-with-output-script: Batch
> 
>    But it is evident that Info-try-follow-nearest-node is fixed if the
>    last line is instead:
> 
>     |   RET                        buffer changes to show batch funcs
> 
> oops, yes this is a mistake on my part in reporting the consequences of
> the fix.  the good news is that the "m RET" case is now fixed (revision
> 1.333 in cvs), as well.
> 
> w/ 1.333 i ran around slib.info and dired-x.info using "RET", "m", "l"
> and various "i" (default/exact/inexact input) commands w/o problems.
> please let me know if you find any new (and interesting) bugs, otherwise
> i'm inclined to look at other things now.  M-x viva-scheme!

Your change broke a lot of *note .... :: links where the :: is not
followed by a space, like these:

        (*note some note::)
        some text, *note some note::, more text

Try C-h i m elisp RET i signal RET TAB RET

I've installed a fix to CVS to stop searching for the end of the
reference when a comma, period, semicolon, or right parenthesis
follows the last colon.  

However, I'm not sure this the right fix.  It would probably be better
to specify the characters which shall cause a continued search if they
follow a colon, but I'm not sure what characters to include...

-- 
Kim F. Storm  http://www.cua.dk

Re: [agj@alum.mit.edu: Re: ":" in Scheme names.]

[Thien-Thi Nguyen]

   From: no-spam@cua.dk (Kim F. Storm)
   Date: 26 Jan 2003 20:51:59 +0100

   Your change broke a lot of *note .... :: links where the :: is not
   followed by a space, like these:

	   (*note some note::)
	   some text, *note some note::, more text

   Try C-h i m elisp RET i signal RET TAB RET

   I've installed a fix to CVS to stop searching for the end of the
   reference when a comma, period, semicolon, or right parenthesis
   follows the last colon.  

thanks for catching this.

   However, I'm not sure this the right fix.  It would probably be better
   to specify the characters which shall cause a continued search if they
   follow a colon, but I'm not sure what characters to include...

if i understand correctly, you mean to replace while-not, w/ simply
while.  initially i tried something like:

 (while (looking-at ":*\\sw") ...)

but testing against elisp index fails for "&rest" and anything else w/
chars not in the word-constituent syntax class.  unless there is a way
to create a syntax class that is \sw plus \s- plus \s_ minus ":", i
think we are stuck w/ explicitly enumerating trailing context.  the
fixed info.el appears to be enough for now; we just need to track
makeinfo (x)ref rendering (not a big deal -- i don't think that is
changing very much).

thi