info faces for strings and quotations

info faces for strings and quotations

[Drew Adams]

I find Info to be a lot more readable with strings ("blah blah toto titi")
and quoted names (`my-foobar-function') highlighted in different faces from
the rest of the text.

What about adding something simple like this to the end of
Info-fontify-node - just before (set-buffer-modified-p nil):

  (goto-char (point-min))
  (while (re-search-forward "`\\([^']+\\)'" nil t)
    (put-text-property (match-beginning 1) (match-end 1)
                       'face info-quoted-name-face))
  (goto-char (point-min))
  (while (re-search-forward "\"\\([^\"]+\\)\"" nil t)
    (put-text-property (match-beginning 1) (match-end 1)
                       'face info-string-face))

Granted, there are a few Info nodes that get thrown off by this simple
matching because of unbalanced double or single quotes. This happens when a
node discusses characters like \", \`, and \' or uses this, where the two
are nested: `""'.

Probably most of these problems could be eliminated by smartening up the
regexps. Most of the "rogue" occurrences of these characters are preceded by
backslashes: \".

Even with these simple regexps, however, 99% of the pages look better, and
the text is much more readable, IMO. In particular, it is very helpful that
quoted names stand out; the double-quoted strings are less important.

To see this at its _worst_, these are the nodes of the Emacs manual that
throw off the regexps to some extent:

 - Glossary
 - Key (Character) Index
 - Minor Modes
 - Init Syntax
 - Printing and MS-DOS
 - Cursor Position Information (at the bottom)
 - TeX Editing Commands
 - Options Controlling Comments
 - Etags Regexps
 - Acknowledgements

RE: info faces for strings and quotations

[Drew Adams]

Below is code that highlights "..." and `...' quotations in Info. The minor
display pbs I mentioned earlier (escaped quotes) have been eliminated, with
input from Lennart Borgman. The only Info nodes in the Emacs manual that
still display with slight problems are these:

 - Acknowledgements
 - Etags Regexps
 - Glossary (very minor)

These remaining display problems are because of isolated " characters. For
instance, in node Acknowledgements, this appears, which turns on
highlighting until the next " in the node:

  Torbjo"rn Einarsson contributed the...

And this appears in Glossary, which highlights only between the first ` and
the first ':

   ASCII printing character
     ASCII printing characters include letters, digits, space, and these
     punctuation characters: `!@#$%^& *()_-+=|\~` {}[]:;"' <>,.?/'.

The Etags Regexps pb is similar to the Glossary pb. Nothing to be done about
this. And as I say, I found only three such occurrences in the entire Emacs
manual.

I also went through the entire Elisp manual, and found only 7 nodes where
there were (similar) problems: Compilation Functions, Example Major Modes,
Auto Major Mode, Imenu, Regexp Example, Standard Errors, and Index.


So, I'm proposing these changes to `info.el':

1. Add these user options (or equivalent), for the quotation highlight faces
and for turning this highlighting on/off:

(defface info-quoted-name-face        ; For `...'
    '((((class color) (background light)) :foreground "red3")
      (((class color) (background dark)) :foreground "white")
      (t :weight bold :slant italic))
    "Face used for quoted names (`...') in `info'."
    :group 'info)

(defface info-string-face             ; For "..."
    '((((class color) (background light)) :foreground "green3")
      (((class color) (background dark)) :foreground "yellow")
      (t :slant italic))
    "Face used for strings (\"...\") in `info'."
    :group 'info)

(defcustom Info-fontify-quotations-flag t
  "*Non-nil means `info' fontifies text between quotes.
This applies to double-quote strings (\"...\") and text between
single-quotes (`...')."
  :type 'boolean
  :group 'info)

Note: Whatever colors are picked, I think `info-quoted-name-face' should
stand out more than `info-string-face'.


2. Add this function (or equivalent) to `info.el':

(defun info-fontify-quotations ()
  "Fontify double-quote strings (\"...\") and text between single-quotes
(`...')
For single-quotes, use `info-quoted-name-face'.
For double-quotes, use `info-string-face'."
  (goto-char (point-min))
  (let (;; double-quote strings: "...", "...\", "...\\", etc.; m1=\*
        ;; or single-quote strings: `...'
        (either-re "\"[^\"]*\\([\\\\]*\\)\"\\|`[^'\n]+'")
        ;; ", \", \\", \\\" etc.; m2=\*
        (dblquote-re "\\([\\\\]*\\)\"")
        m0 p0b p0e                      ; Whole match: `...' or "..."
        m1 p1b p1e                      ; \* subexp of "...\*" match
        m2 p2b p2e                      ; \* subexp of "...\*" match
        escaped-dblquote-p)
    (while (re-search-forward either-re nil t)
      (setq m0 (match-string 0)         ; Whole match string
            p0b (nth 0 (match-data))    ; Beginning of m0
            p0e (nth 1 (match-data))    ; End of m0
            m1 (match-string 1)         ; \* subexp of "...\*" match
            p1b (nth 2 (match-data))    ; Beginning of m1
            p1e (nth 3 (match-data)))   ; End of m2
      (when (equal (char-after p0b) ?\") ; double-quote string: "..."
        (when (> p1e p1b)               ; May be escaped: \ inside "...\*"
          (when (= (mod (- p1e p1b) 2) 1) ; Escaped (odd number of
backslashes: \", \\\",...)
            (setq escaped-dblquote-p t)
            (while escaped-dblquote-p
              (if (not (re-search-forward dblquote-re nil t)) ; Look for \*"
                  (setq escaped-dblquote-p nil) ; No \*"
                (setq m2 (match-string 0) ; \*"
                      p2b (nth 0 (match-data)) ; Beginning of \*": \ or "
                      p2e (nth 1 (match-data)) ; End of \*": "
                      p0e p2e)          ; Update pointer
                (if (= p2e p2b)
                    (setq escaped-dblquote-p nil) ; Not escaped - ", \\",
\\\\", etc.
                  (when (= (mod (- p2e p2b) 2) 1) (setq escaped-dblquote-p
nil))))))))
      (if (eq ?` (aref m0 0))
          (put-text-property (1+ p0b) (1- p0e) 'face 'info-quoted-name-face)
        (put-text-property p0b p0e 'face 'info-string-face)))))


3. Call `info-fontify-quotations' at the end of `Info-fontify-node' -- add
this just before the last line, (set-buffer-modified-p nil):

        (goto-char (point-min))
        (when Info-fontify-quotations-p (info-fontify-quotations))


Could others please try this and see if you think it should be added.

Thanks,

   Drew

Re: info faces for strings and quotations

[Miles Bader]

"Drew Adams" <drew.adams@oracle.com> writes:
> So, I'm proposing these changes to `info.el':
>
> 1. Add these user options (or equivalent), for the quotation highlight faces
> and for turning this highlighting on/off:

Please do not use the suffix `-face' for face names -- they have
their own namespace, so it is redundant.  Just use something like
(defface info-quoted-name ...) and (defface info-string ...).

Thanks,

-Miles
-- 
In New York, most people don't have cars, so if you want to kill a person, you
have to take the subway to their house.  And sometimes on the way, the train
is delayed and you get impatient, so you have to kill someone on the subway.
  [George Carlin]

RE: info faces for strings and quotations

[Drew Adams]

Sorry, the main regexp should be non-greedy. This:

  "\"[^\"]*?\\([\\\\]*\\)\"\\|`[^'\n]+'"

instead of this:

  "\"[^\"]*\\([\\\\]*\\)\"\\|`[^'\n]+'"

Thanks Lennart.

 - Drew



-----Original Message-----
Below is code that highlights "..." and `...' quotations in Info.
...
(defun info-fontify-quotations ()
  "..."
  (goto-char (point-min))
  (let ((either-re "\"[^\"]*\\([\\\\]*\\)\"\\|`[^'\n]+'")
      ...))

Re: info faces for strings and quotations

[Stefan]

>   "\"[^\"]*?\\([\\\\]*\\)\"\\|`[^'\n]+'"
                 ^^^^^^

This is the regexp [\\] (once you remove the quoting related to strings).
In a char-range, \ is not special, so [\\] is just like [aa] (expect for
the char \ instead of the char a), so it's equivalent to [\] which is
equivalent to \\ except it's slower.  Just use

   "\"[^\"]*?\\(\\\\*\\)\"\\|`[^'\n]+'"

but I must say that I do not understand this \\(\\\\*\\) business.


        Stefan

Re: info faces for strings and quotations

[Stefan]

> (defun info-fontify-quotations ()
>   "Fontify double-quote strings (\"...\") and text between single-quotes
> (`...')
> For single-quotes, use `info-quoted-name-face'.
> For double-quotes, use `info-string-face'."
>   (goto-char (point-min))
>   (let (;; double-quote strings: "...", "...\", "...\\", etc.; m1=\*
>         ;; or single-quote strings: `...'
>         (either-re "\"[^\"]*\\([\\\\]*\\)\"\\|`[^'\n]+'")
>         ;; ", \", \\", \\\" etc.; m2=\*
>         (dblquote-re "\\([\\\\]*\\)\"")
>         m0 p0b p0e                      ; Whole match: `...' or "..."
>         m1 p1b p1e                      ; \* subexp of "...\*" match
>         m2 p2b p2e                      ; \* subexp of "...\*" match
>         escaped-dblquote-p)
>     (while (re-search-forward either-re nil t)
>       (setq m0 (match-string 0)         ; Whole match string
>             p0b (nth 0 (match-data))    ; Beginning of m0

Never do (nth x (match-data)).  Always use (match-beginning N) or
(match-end N) instead.  Much more readable (and efficient).

>             p0e (nth 1 (match-data))    ; End of m0
>             m1 (match-string 1)         ; \* subexp of "...\*" match

The code never seems to use `m1' so you can spare this string-allocation.

>             p1b (nth 2 (match-data))    ; Beginning of m1
>             p1e (nth 3 (match-data)))   ; End of m2
>       (when (equal (char-after p0b) ?\") ; double-quote string: "..."
>         (when (> p1e p1b)               ; May be escaped: \ inside "...\*"
>           (when (= (mod (- p1e p1b) 2) 1) ; Escaped (odd number of
> backslashes: \", \\\",...)
>             (setq escaped-dblquote-p t)
>             (while escaped-dblquote-p
>               (if (not (re-search-forward dblquote-re nil t)) ; Look for \*"
>                   (setq escaped-dblquote-p nil) ; No \*"
>                 (setq m2 (match-string 0) ; \*"

Similary `m2' is never used.

>                       p2b (nth 0 (match-data)) ; Beginning of \*": \ or "
>                       p2e (nth 1 (match-data)) ; End of \*": "
>                       p0e p2e)          ; Update pointer
>                 (if (= p2e p2b)
>                     (setq escaped-dblquote-p nil) ; Not escaped - ", \\",
> \\\\", etc.
>                   (when (= (mod (- p2e p2b) 2) 1) (setq escaped-dblquote-p
> nil))))))))

Why not use (either-re "\"\\([^\\\"]\\|\\\\[\\\"]\\)*\"\\|`[^'\n]+'")
and get rid of all this code (i.e. the regexp-matching does the
escape-counting for you)?

>       (if (eq ?` (aref m0 0))

Use (eq ?` (char-after p0b)) and you can get rid of `m0'.

>           (put-text-property (1+ p0b) (1- p0e) 'face 'info-quoted-name-face)
>         (put-text-property p0b p0e 'face 'info-string-face)))))

Why fontify the interior of `...' but fontify both the interior and the
quotes for "..." ?  I.e. why not use

    (put-text-property p0b p0e 'face
     (if (eq ?` (char-after p0b)) 'info-quoted-name-face 'info-string-face))
or
    (put-text-property (1+ p0b) (1- p0e) 'face
     (if (eq ?` (char-after p0b)) 'info-quoted-name-face 'info-string-face))

Is it just because your quoted face is bold and you don't like to see the `
and ' in bold, or is there a deeper reason?


        Stefan

RE: info faces for strings and quotations

[Drew Adams]

Thanks for lots of good input, Stephan. You made several good points about
the implementation, which I'll put into practice immediately.

As you know, in my first message, I used just these (separate) simple regexp
searches, which didn't take care of escaped "s but worked 99% of the time:

  (re-search-forward "`\\([^']+\\)'" nil t)
  (re-search-forward "\"\\([^\"]+\\)\"" nil t)

I mentioned the fact that the regexps could be improved, and Lennart was
kind enough to do some coding and testing. He came up with something that
solved the escaped "s pb, which I tried and sent along to the list (though I
mistakenly left off the non-greedy "?"). I sent the updated code because it
improved the functionality, taking care of some corner cases.

I'll make the implementation changes you suggest. Besides the implementation
issues, you asked this question about the _functionality_:

  Why fontify the interior of `...' but fontify both the interior and the
quotes for "..." ?   Is it just because your quoted face is bold and you
don't like to see the `
and ' in bold, or is there a deeper reason?

Good question. That was by design. In most Info files, `...' is used to
identify commands, key sequences, filenames and the like; "..." is used
(usually) to identify strings in code. I found that not highlighting the
single-quotes and highlighting the double-quotes improved readability. Try
different combinations of highlighting the quotes (or not), yourself, and
see what you think. My second choice would be to not highlight either kind
of delimiter.

BTW, I didn't test this with a terminal Emacs; I just threw the bold-italic
and italic face attributes in there at the last minute to have something for
the non-(class color) case. Do you have a suggestion on what to use in that
case? (In fact, just before I sent the email, the version I use had face
variables with defvar, which is why I used the names *-face -- a no-no
nowadays, as Miles pointed out.)

The code will be cleaner & leaner thanks to input from Lennart, Miles, and
you.
So, what do you and others think of the _desirability_ of (optionally)
highlighting quoted expressions in Info?

 - Drew

Re: info faces for strings and quotations

[LENNART BORGMAN]

----- Original Message -----
From: Stefan <monnier@iro.umontreal.ca>
Date: Tuesday, October 5, 2004 1:56 pm
Subject: Re: info faces for strings and quotations


> Why not use (either-re "\"\\([^\\\"]\\|\\\\[\\\"]\\)*\"\\|`[^'\n]+'")
> and get rid of all this code (i.e. the regexp-matching does the
> escape-counting for you)?

I have never heard of either-re, is that something in the CVS-version?

I did not notice match-beginning/end, thanks!

- Lennart

RE: info faces for strings and quotations

[Drew Adams]

`either-re' was just my renamed version of your variable `rr'.  I was trying
to make the name say what the regexp represented. - Drew

-----Original Message-----
From: LENNART BORGMAN [mailto:lennart.borgman.073@student.lu.se]

> Why not use (either-re "\"\\([^\\\"]\\|\\\\[\\\"]\\)*\"\\|`[^'\n]+'")
> and get rid of all this code (i.e. the regexp-matching does the
> escape-counting for you)?

I have never heard of either-re, is that something in the CVS-version?

Re: info faces for strings and quotations

[Luc Teirlinck]

Drew Adams wrote:

   So, what do you and others think of the _desirability_ of (optionally)
   highlighting quoted expressions in Info?

Info files are essentially plain text with some minimal markup that is
intended for navigation, not highlighting.  So I believe that there is
no way that one can do what you are trying to do completely reliably
any more than that it is possible to implement
Info-hide-note-references completely reliably.  To do that, one would
need to base the Emacs Info reader on the XML output format instead of
on the Info output format.  (Maybe somebody might actually do that
some day.)

So I would suggest that if any feature like this is added to the Emacs
Info reader, that it should be done as a user option, that is off by
default.  Also, the docstring should explicitly mention that it
occasionally can lead to anomalous highlighting.

Sincerely,

Luc.

RE: info faces for strings and quotations

[Drew Adams]

Info highlighting off by default. I can live with that. What do others
think?

BTW, have you -- have others -- _tried_ the code? Your argument sounds,
well, hypothetical. In _practice_, there are only *4* nodes of the Elisp
Manual that exhibit any highlighting problem at all, and those pbs are
minor. This is a big manual.

Yes, it's obvious that there will always be some examples of code in the
Info doc (on different programming languages, for instance) that use
expressions (e.g. regexp expressions or byte-code examples) that won't
highlight 100% correctly. Info highlighting works perfectly 99% of the time,
however, including lots of crazy regexp expressions in the Elisp manual
(yes, it's surprising).

And, the same rare problems can be seen with font-lock highlighting of Emacs
Lisp code -- which is presumably one reason we turn off font-lock by
default. (Do the doc strings of font-lock functions mention that they
"occasionally can lead to anomalous highlighting"?  Are we worried about
being sued, here, or what?)

Some of the rare Info highlighting problems point out things that should
really be changed in the Info files themselves, such as ``region'' instead
of "region" at node (emacs)Top (highlighted from first ` to first '), or
this unnecessary single-quoting of a Lisp sexp at node (elisp)Auto Major
Mode: `("\\.gz\\'" FUNCTION t)'.

You say that Info markup is "intended for navigation, not highlighting". The
`...' convention is explicitly intended to _highlight_ (demarcate, delimit,
bring out, set off, draw attention to) commands, key sequences, and perhaps
file names. It has absolutely nothing to do with navigating. That's the
markup convention I want to exploit here -- that and "..." for strings. This
proposal has nothing to do with other, hidden info-file markup; it is all
about user-visible "highlighting".

Perhaps what you meant was that `...' was not _originally designed_ with an
eye to being highlighted. Agreed. But it lends itself to highlighting
remarkably well, with no changes needed.

Again, I invite you to try it, to see what it does in _practice_. Here is
the code, with the latest regexp, which allows escaped chars and newlines
inside `...' and "...".  (Some Info nodes, like (emacs)Visiting, split a key
sequence or command name over two lines.)

(defun info-fontify-quotations ()
    "Fontify double-quote strings (\"...\") and text between single-quotes
(`...')
For single-quotes, use face `info-quoted-name'.
For double-quotes, use face `info-string'."
    (while
        (re-search-forward

"\"\\(?:[^\\\"]\\|\\\\\\(?:.\\|[\n]\\)\\)*\"\\|`\\(?:[^'\\]\\|\\\\\\(?:.\\|[
\n]\\)\\)+'"
         nil t)
      (if (eq ?` (aref (match-string 0) 0))
          (put-text-property (1+ (match-beginning 0)) (1- (match-end 0))
                             'face 'info-quoted-name)
        (put-text-property (match-beginning 0) (match-end 0)
                           'face 'info-string))))

Again, to use this, just call `info-fontify-quotations' at the end of
`Info-fontify-node' -- add this just before the last line, which is
(set-buffer-modified-p nil):
        (goto-char (point-min))
        (when Info-fontify-quotations-p (info-fontify-quotations))

Hoping to hear from more people who have tried the code, regardless of which
way your opinion leans.

 - Drew

P.S. FYI, the 4 Elisp-manual nodes that do _not_ highlight 100% correctly
are these:

 - Example Major Modes -- an isolated " (?\")
 - Index -- an isolated "  (" entry in the index)
 - Auto Major Modes -- the single-quoted Lisp sexp mentioned above
 - Standard Error -- ' inside "..." inside `...' (`"Symbol's chain of
function indirections contains a loop"') -- this too should probably be
corrected in the doc itself

If the latter two were corrected in the Info files, there would be only *2*
nodes of the entire Elisp manual that were not 100% perfect. I don't think
we'll be sued.


-----Original Message-----
From: Luc Teirlinck [mailto:teirllm@dms.auburn.edu]

   So, what do you and others think of the _desirability_ of (optionally)
   highlighting quoted expressions in Info?

Info files are essentially plain text with some minimal markup that is
intended for navigation, not highlighting.  So I believe that there is
no way that one can do what you are trying to do completely reliably
any more than that it is possible to implement
Info-hide-note-references completely reliably.  To do that, one would
need to base the Emacs Info reader on the XML output format instead of
on the Info output format.  (Maybe somebody might actually do that
some day.)

So I would suggest that if any feature like this is added to the Emacs
Info reader, that it should be done as a user option, that is off by
default.  Also, the docstring should explicitly mention that it
occasionally can lead to anomalous highlighting.

Re: info faces for strings and quotations

[Miles Bader]

"Drew Adams" <drew.adams@oracle.com> writes:
> BTW, have you -- have others -- _tried_ the code?

Could you send it as a single patch file?  That's a good idea any time
you submit code for consideration (easier to test, and often less
ambiguous).

Thanks,

-Miles
-- 
`Cars give people wonderful freedom and increase their opportunities.
 But they also destroy the environment, to an extent so drastic that
 they kill all social life' (from _A Pattern Language_)

Re: info faces for strings and quotations

[Stefan Monnier]

> And, the same rare problems can be seen with font-lock highlighting of Emacs

It's actually slightly different: the nature of programming languages
implies that there exist an automatic way to unambiguously define what is
a correct parse.  I.e., font-lock highlighting has bugs, but they can
clearly be recognized as such (and often fixed if it matters enough).
Also font-lock bugs generally can be worked around by changing the
source code because such buffers are generally writable.

In Info files, the "..." and `...' syntax is not unambiguous (it is intended
to be parsed by humans only), so there is no automatic system that works
100% percent: any fix for a particular case will introduce problems
elsewhere.  And Info buffers are normally read-only and most Info users will
have a hard time finding the Info or Texinfo files to edit them (and they
often can't edit them because they don't have the required access rights).

Your regexps work great for Info pages describing Elisp code.  For Info
pages relating to some other language where the convention for escape
sequences in strings is significantly different, they may work
very poorly.  But that's what the config var is for, anyway.

I would probably turn it on ;-)


        Stefan

RE: info faces for strings and quotations

[Drew Adams]

OK, sure, sure. Programming languages are formal languages, so your point is
of course correct, technically. And, if existing font-lock code doesn't do a
better job in practice than Info quote highlighting, at least it has the
theoretical advantage that it _could_ be made to parse the language as
needed. My point was a practical one, about the current state of font-lock
code. It is good enough; we live with it; we probably don't want it to spend
much more time trying to be more precise.

I'd much rather hear about results from practicing with this than arguments
of this sort. Or, lets say, in addition to arguments of this sort.

I only have the usual info files delivered with Emacs and Cygwin, and, yes,
none of them are devoted to examples of APL code...sigh. But, based on your
argument about not necessarily being appropriate to other languages than
Lisp, I did take a look at some of the 33 standard Info manuals I do have.
In fact, I checked them all -- all nodes (the highlighting helps to check
them quickly).

Many are of course about Emacs and Emacs Lisp packages. Even in those,
however, there are examples of code in other languages. I did find the
following highlighting problems in all the 33 manuals I checked. I didn't
find any pbs in the manuals not mentioned.

------------

CC MODE -- node Text Filling and Line Breaking: `\' (this case isn't treated
yet, but perhaps should be)

CL -- node Old CL Compatibility: middle of title "The `cl-compat'package"
got highlighted as a `...'

IDLWAVE -- node Getting Started: same pb as CC Mode (`\')
        -- node Code Templates: same

MH-E -- node Moving Your Mail Around: `"lpr -J '%s'"' (first ' ends `...';
second " starts "...")

MESSAGE -- node: Message Actions: lisp quoted sexp --
'(set-window-configuration...

EMACS FAQ -- several nodes have `...' or "..." in node titles. These are
highlighted.

WIDGET -- several nodes have `...' in titles

REFTEX -- node fancyref (LaTeX package): `fancyref' in title highlighted

VIP -- node New Commands: same pb as CC Mode (`\')
    -- node Important Keys: same

VIPER -- node Emacs Preliminaries: `C-\' (' is considered escaped, so
doesn't turn off highlighting)
      -- node Vi State: same
      -- node Insert State: same
      -- node Emacs Related Commands: same
      -- node New Commands: same pb as CC Mode (`\')
      -- node Key Index: isolated " -- * "<a-z1-9>P:

CALC -- node Arithmetic Tutorial: same pb as CC Mode (`\')
     -- node Types Tutorial: isolated " from hours, minutes, seconds entries
in table, such as 2@ 30' 0"
     -- node Programming Tutorial: this comment with isolated `: # Save
local values (Z `)
     -- node List Answer 9: same pb as CC Mode (`\')
     -- node List Tutorial Exercise 13: isolated " -- "Testing, 1, 2, 3
<RET>
     -- node Types Tutorial Exercise 4: more hours, minutes, seconds entries
     -- node Programming Tutorial Exercise 8, 9, and 10: isolated ` -- C-x
( Z `
     -- node HMS Forms: more hours, minutes, seconds as @ ' and "
     -- node HMS Formats: more hours, minutes, seconds as @ ' and "
     -- node Business Days: more hours, minutes, seconds as @ ' and "
     -- node Polynomials: key sequence `a \' (the ' is considered escaped by
the highlighting)
     -- node More About Embedded Mode: same pb as CC Mode (`\')
     -- node Customizing Embedded Mode: regexp `"\\'\\|\n$\\|\\$\\$?"'
(first ' ends `...' and second " starts "...")
     -- Calc Summary: isolated ` and " in table of key sequences
     -- Key Index: same as Calc Summary

     Note: the Calc manual is very large, has zillions of symbolic and
formulaic examples, and deals with the syntaxes of several other languages,
including Eqn, Mathematica, and Tex. It includes tabular material and ASCII
art. And yes, this manual, in particular, _benefits_ greatly from
quoted-name highlighting. I find it very difficult to parse (read) without
highlighting. Here are two typical paragraphs of _text_:

     Quotients of negative-looking values are simplified according to `(-a)
/ (-b)' to `a / b', `(-a) / (b - c)' to `a / (c - b)', and `(a - b) / (-c)'
to `(b - a) / c'.

     Generic identity matrices (*note Matrix Mode::) are simplified by the
following rules:  `idn(a) + b' to `a + b' if `b' is provably scalar, or
expanded out if `b' is a matrix; `idn(a) + idn(b)' to `idn(a + b)';
`-idn(a)' to `idn(-a)'; `a idn(b)' to `idn(a b)' if `a' is provably
scalar, or to `a b' if `a' is provably non-scalar; `idn(a) idn(b)' to
`idn(a b)'; analogous simplifications for quotients involving `idn';
and `idn(a)^n' to `idn(a^n)' where `n' is an integer.

EMACS LISP INTRO -- a few nodes, such as Top: `...' in titles
                 -- a few nodes, such as Height of label: sexp with ' inside
`...': `(apply 'max numbers-list)'
                 -- node beginning-of-buffer complete: same pb as CC Mode
(`\')
                 -- node sentence-end: same
                 -- node the-the: unbalanced "s in text: _I duplicate "the'_

TEXINFO -- several nodes, such as The `subsub' Commands, with `...' in title
        -- node First Line: same pb as CC Mode (`\')
        -- node Command Syntax: same
        -- node math: same
        -- node verb: isolated ' --  @verb{+@'e?`!`{}\+} this?
        -- node Inserting Accents: @`o
        -- node Defining Macros: `\PARAM1\' -- the ' is considered escaped

 - Drew




-----Original Message-----
From: emacs-devel-bounces+drew.adams=oracle.com@gnu.org
[mailto:emacs-devel-bounces+drew.adams=oracle.com@gnu.org]On Behalf Of
Stefan Monnier
Sent: Tuesday, October 05, 2004 9:53 PM
To: Drew Adams
Cc: lennart.borgman.073@student.lu.se; Luc Teirlinck;
emacs-devel@gnu.org
Subject: Re: info faces for strings and quotations


> And, the same rare problems can be seen with font-lock highlighting of
Emacs

It's actually slightly different: the nature of programming languages
implies that there exist an automatic way to unambiguously define what is
a correct parse.  I.e., font-lock highlighting has bugs, but they can
clearly be recognized as such (and often fixed if it matters enough).
Also font-lock bugs generally can be worked around by changing the
source code because such buffers are generally writable.

In Info files, the "..." and `...' syntax is not unambiguous (it is intended
to be parsed by humans only), so there is no automatic system that works
100% percent: any fix for a particular case will introduce problems
elsewhere.  And Info buffers are normally read-only and most Info users will
have a hard time finding the Info or Texinfo files to edit them (and they
often can't edit them because they don't have the required access rights).

Your regexps work great for Info pages describing Elisp code.  For Info
pages relating to some other language where the convention for escape
sequences in strings is significantly different, they may work
very poorly.  But that's what the config var is for, anyway.

I would probably turn it on ;-)


        Stefan


_______________________________________________
Emacs-devel mailing list
Emacs-devel@gnu.org
http://lists.gnu.org/mailman/listinfo/emacs-devel

RE: info faces for strings and quotations

[Drew Adams]

[-- Attachment #1: Type: text/plain, Size: 496 bytes --]

Here's the patch to info.el that highlights `...' and "...".

Sending it as an attachment because my mailer apparently wraps the lines
otherwise.

 - Drew


-----Original Message-----
From: Miles Bader [mailto:miles@lsi.nec.co.jp]

"Drew Adams" <drew.adams@oracle.com> writes:
> BTW, have you -- have others -- _tried_ the code?

Could you send it as a single patch file?  That's a good idea any time
you submit code for consideration (easier to test, and often less
ambiguous).

Thanks,

-Miles

[-- Attachment #2: diff-info-highlight-quotations.txt --]
[-- Type: text/plain, Size: 4007 bytes --]

diff -c "c:/emacs-21.3.50/lisp/info.el" "c:/drews-lisp-20/info-w-quotes.el"
*** c:/emacs-21.3.50/lisp/info.el	Mon Jul 26 09:42:06 2004
--- c:/drews-lisp-20/info-w-quotes.el	Tue Oct  5 21:45:12 2004
***************
*** 65,70 ****
--- 65,91 ----
  The Lisp code is executed when the node is selected.")
  (put 'Info-enable-active-nodes 'risky-local-variable t)
  
+ (defcustom Info-fontify-quotations-flag t
+   "*Non-nil means `info' fontifies text between quotes.
+ This applies to double-quote strings (\"...\") and text between
+ single-quotes (`...')."
+   :type 'boolean
+   :group 'info)
+ 
+ (defface info-quoted-name
+   '((((class color) (background light)) :foreground "red3")
+     (((class color) (background dark)) :foreground "white")
+     (t :weight bold :slant italic))
+   "Face used for quoted names (`...') in `info'."
+   :group 'info)
+ 
+ (defface info-string
+   '((((class color) (background light)) :foreground "green4")
+     (((class color) (background dark)) :foreground "yellow")
+     (t :slant italic))
+   "Face used for strings (\"...\") in `info'."
+   :group 'info)
+ 
  (defface info-node
    '((((class color) (background light)) :foreground "brown" :weight bold :slant italic)
      (((class color) (background dark)) :foreground "white" :weight bold :slant italic)
***************
*** 3669,3676 ****
                                 '(font-lock-face info-xref
                                                  mouse-face highlight
                                                  help-echo "mouse-2: go to this URL"))))
! 
        (set-buffer-modified-p nil))))
  \f
  
  ;; When an Info buffer is killed, make sure the associated tags buffer
--- 3690,3736 ----
                                 '(font-lock-face info-xref
                                                  mouse-face highlight
                                                  help-echo "mouse-2: go to this URL"))))
!       (goto-char (point-min))
!       (when Info-fontify-quotations-flag (info-fontify-quotations))
        (set-buffer-modified-p nil))))
+ 
+ 
+ ;; The regexp has these parts: double-quoted string or single-quoted stuff.
+ ;;
+ ;; String has, inside "...", zero or more of these characters:
+ ;;   - any character except \ and "
+ ;;   - \ followed by any character
+ ;;
+ ;; Single-quoted stuff has, inside `...', one or more of these characters:
+ ;;   - any character except \ and '
+ ;;   - \ followed by any character
+ ;;
+ (if (< emacs-major-version 21)
+     (defun info-fontify-quotations ()
+       "Fontify double-quote strings (\"...\") and text between single-quotes (`...')
+ For single-quotes, use face `info-quoted-name'.
+ For double-quotes, use face `info-string'."
+       (while (re-search-forward
+               "\"\\([^\\\"]\\|\\\\\\(.\\|[\n]\\)\\)*\"\\|`\\([^'\\]\\|\\\\\\(.\\|[\n]\\)\\)+'"
+               nil t)
+         (if (eq ?` (aref (match-string 0) 0))
+             (put-text-property (1+ (match-beginning 0)) (1- (match-end 0))
+                                'face info-quoted-name)
+           (put-text-property (match-beginning 0) (match-end 0)
+                              'face info-string))))
+   (defun info-fontify-quotations ()
+     "Fontify double-quote strings (\"...\") and text between single-quotes (`...')
+ For single-quotes, use face `info-quoted-name'.
+ For double-quotes, use face `info-string'."
+     (while
+         (re-search-forward
+          "\"\\(?:[^\\\"]\\|\\\\\\(?:.\\|[\n]\\)\\)*\"\\|`\\(?:[^'\\]\\|\\\\\\(?:.\\|[\n]\\)\\)+'"
+          nil t)
+       (if (eq ?` (aref (match-string 0) 0))
+           (put-text-property (1+ (match-beginning 0)) (1- (match-end 0))
+                              'face 'info-quoted-name)
+         (put-text-property (match-beginning 0) (match-end 0)
+                            'face 'info-string)))))
  \f
  
  ;; When an Info buffer is killed, make sure the associated tags buffer

[-- Attachment #3: Type: text/plain, Size: 142 bytes --]

_______________________________________________
Emacs-devel mailing list
Emacs-devel@gnu.org
http://lists.gnu.org/mailman/listinfo/emacs-devel

Re: info faces for strings and quotations

[Matt Hodges]

[-- Attachment #1: Type: text/plain, Size: 512 bytes --]

>>>>> Miles Bader writes:

 > Please do not use the suffix `-face' for face names -- they have
 > their own namespace, so it is redundant. Just use something like
 > (defface info-quoted-name ...) and (defface info-string ...).

I was going to suggest that an apropos-face command might be useful,
but I see that customize-apropos-faces already exists. This is quite
slow, at least on my machine, so how about allowing a filter for
list-faces-display?

Something along the lines of the attached.

Thanks,

Matt


[-- Attachment #2: Patch for faces.el. --]
[-- Type: text/plain, Size: 3452 bytes --]

*** faces.el	15 Sep 2004 11:48:16 +0100	1.290
--- faces.el	06 Oct 2004 09:22:57 +0100	
***************
*** 1136,1150 ****
  ;; conflict with Lucid, which uses that name differently.
  
  (defvar help-xref-stack)
! (defun list-faces-display ()
    "List all faces, using the same sample text in each.
  The sample text is a string that comes from the variable
! `list-faces-sample-text'."
!   (interactive)
    (let ((faces (sort (face-list) #'string-lessp))
- 	(face nil)
  	(frame (selected-frame))
  	disp-frame window face-name)
      (with-output-to-temp-buffer "*Faces*"
        (save-excursion
  	(set-buffer standard-output)
--- 1136,1157 ----
  ;; conflict with Lucid, which uses that name differently.
  
  (defvar help-xref-stack)
! (defun list-faces-display (&optional regexp)
    "List all faces, using the same sample text in each.
  The sample text is a string that comes from the variable
! `list-faces-sample-text'. With a prefix arg, limit faces to those
! matching REGEXP."
!   (interactive (list (and current-prefix-arg
!                           (read-string "Filter by regexp: "))))
    (let ((faces (sort (face-list) #'string-lessp))
  	(frame (selected-frame))
  	disp-frame window face-name)
+     ;; Perhaps filter by regexp
+     (when regexp
+       (setq faces (delq nil (mapcar (lambda (f)
+                                       (when (string-match regexp
+                                                           (symbol-name f)) f))
+                                     faces))))
      (with-output-to-temp-buffer "*Faces*"
        (save-excursion
  	(set-buffer standard-output)
***************
*** 1157,1165 ****
  	   "\\[help-follow] on a face name to customize it\n"
  	   "or on its sample text for a description of the face.\n\n")))
  	(setq help-xref-stack nil)
! 	(while faces
! 	  (setq face (car faces))
! 	  (setq faces (cdr faces))
  	  (setq face-name (symbol-name face))
  	  (insert (format "%25s " face-name))
  	  ;; Hyperlink to a customization buffer for the face.  Using
--- 1164,1170 ----
  	   "\\[help-follow] on a face name to customize it\n"
  	   "or on its sample text for a description of the face.\n\n")))
  	(setq help-xref-stack nil)
! 	(mapcar (lambda (face)
  	  (setq face-name (symbol-name face))
  	  (insert (format "%25s " face-name))
  	  ;; Hyperlink to a customization buffer for the face.  Using
***************
*** 1187,1192 ****
--- 1192,1198 ----
  	    (while (not (eobp))
  	      (insert "                          ")
  	      (forward-line 1))))
+                  faces)
  	(goto-char (point-min)))
        (print-help-return-message))
      ;; If the *Faces* buffer appears in a different frame,
***************
*** 1196,1205 ****
      (setq disp-frame (if window (window-frame window)
  		       (car (frame-list))))
      (or (eq frame disp-frame)
! 	(let ((faces (face-list)))
! 	  (while faces
! 	    (copy-face (car faces) (car faces) frame disp-frame)
! 	    (setq faces (cdr faces)))))))
  
  (defun describe-face (face &optional frame)
    "Display the properties of face FACE on FRAME.
--- 1202,1210 ----
      (setq disp-frame (if window (window-frame window)
  		       (car (frame-list))))
      (or (eq frame disp-frame)
!         (while faces
!           (copy-face (car faces) (car faces) frame disp-frame)
!           (setq faces (cdr faces))))))
  
  (defun describe-face (face &optional frame)
    "Display the properties of face FACE on FRAME.

[-- Attachment #3: Type: text/plain, Size: 142 bytes --]

_______________________________________________
Emacs-devel mailing list
Emacs-devel@gnu.org
http://lists.gnu.org/mailman/listinfo/emacs-devel

Re: info faces for strings and quotations

[Oliver Scholz]

Luc Teirlinck <teirllm@dms.auburn.edu> writes:
[...]
> Info files are essentially plain text with some minimal markup that is
> intended for navigation, not highlighting.  So I believe that there is
> no way that one can do what you are trying to do completely reliably
> any more than that it is possible to implement
> Info-hide-note-references completely reliably.  To do that, one would
> need to base the Emacs Info reader on the XML output format instead of
> on the Info output format.

Indeed.  You need more information in the markup to do this in a
reliable way.  (A side effect of this would be that you can also use
font-lock for code examples.)

> (Maybe somebody might actually do that some day.)

There was another longish thread on this issue here about a year ago,
in which I participated extensively.  Well, this is about rendering
XML ... 

I regard it as part of the word processing framework. The
half-finished XML parser with which I am working (an implementation of
SSAX in Elisp) is written with rendering info XML in mind in that it
allows for partial on-demand parsing (which is necessary for files as
large as the Emacs manual).

    Oliver
-- 
Oliver Scholz               15 Vendémiaire an 213 de la Révolution
Ostendstr. 61               Liberté, Egalité, Fraternité!
60314 Frankfurt a. M.       

Re: info faces for strings and quotations

[Robert J. Chassell]

Today's GNU Emacs CVS snapshot, Wed, 2004 Oct  6  16:05 UTC
GNU Emacs 21.3.50.6 (i686-pc-linux-gnu, GTK+ Version 2.4.10)

in which info.el is patched from Drew's message of Wed, 6 Oct 2004
00:40:07 -0700

started with

    /usr/local/src/emacs/src/emacs -Q

The Info buffer fails to highlight properly when the Info file is
produced from a correct Texinfo source.

"Drew Adams" <drew.adams@oracle.com> wrote:

    I did find the following highlighting problems in all the 33
    manuals I checked.

Regarding the `Introduction to Programming in Emacs Lisp', 
you mentioned

        -- a few nodes, such as Height of label: sexp with ' inside
        `...': `(apply 'max numbers-list)'

The regexp failed even though the Texinfo is correct.  In Info, we see
this:

    expression is `(apply 'max numbers-list)'.  This returns the precise

Using the patch, only `apply' is highlighted.  The Texinfo source is
...emacs/lispintro/emacs-lisp-intro.texi and is:

    expression is @code{(apply 'max numbers-list)}.  This returns the

In addition to the three cases of @code{(apply 'max numbers-list)}, I
found nine other case where @code is used with quoted Emacs Lisp
expressions.

There are also problems with @samp, which also puts single quotes
around its arguments, as in

    The error message can be understood: `Symbol's function definition
    is void: this'. 

Only `Symbol' is highlighted.  In Texinfo, this is written:

    The error message can be understood: @samp{Symbol's function
    definition is void:@: this}.  

There are six other @samp problems that I found.

-- 
    Robert J. Chassell                         
    bob@rattlesnake.com                         GnuPG Key ID: 004B4AC8
    http://www.rattlesnake.com                  http://www.teak.cc

RE: info faces for strings and quotations

[Drew Adams]

[-- Attachment #1: Type: text/plain, Size: 451 bytes --]

Here's an updated patch, which replaces the previous one.

It does everything the previous version did, but also correctly treats some
of the common pbs I saw yesterday, such as `\' and `C-\'. In fact, it just
allows _any_ character (except ') inside `...'.

Sorry about sending two patches. I don't expect to change this again.

  - Drew



From: Drew Adams [mailto:drew.adams@oracle.com]
Here's the patch to info.el that highlights `...' and "...".

[-- Attachment #2: diff-info-highlight-quotations-second.txt --]
[-- Type: text/plain, Size: 3237 bytes --]

cd c:/drews-lisp-20/
diff -c "c:/emacs-21.3.50/lisp/info.el" "c:/drews-lisp-20/info-w-quotes-second.el"
*** c:/emacs-21.3.50/lisp/info.el	Mon Jul 26 09:42:06 2004
--- c:/drews-lisp-20/info-w-quotes-second.el	Wed Oct  6 13:51:18 2004
***************
*** 65,70 ****
--- 65,91 ----
  The Lisp code is executed when the node is selected.")
  (put 'Info-enable-active-nodes 'risky-local-variable t)
  
+ (defcustom Info-fontify-quotations-flag t
+   "*Non-nil means `info' fontifies text between quotes.
+ This applies to double-quote strings (\"...\") and text between
+ single-quotes (`...')."
+   :type 'boolean
+   :group 'info)
+ 
+ (defface info-quoted-name
+   '((((class color) (background light)) :foreground "red3")
+     (((class color) (background dark)) :foreground "white")
+     (t :weight bold :slant italic))
+   "Face used for quoted names (`...') in `info'."
+   :group 'info)
+ 
+ (defface info-string
+   '((((class color) (background light)) :foreground "green4")
+     (((class color) (background dark)) :foreground "yellow")
+     (t :slant italic))
+   "Face used for strings (\"...\") in `info'."
+   :group 'info)
+ 
  (defface info-node
    '((((class color) (background light)) :foreground "brown" :weight bold :slant italic)
      (((class color) (background dark)) :foreground "white" :weight bold :slant italic)
***************
*** 3669,3676 ****
                                 '(font-lock-face info-xref
                                                  mouse-face highlight
                                                  help-echo "mouse-2: go to this URL"))))
! 
        (set-buffer-modified-p nil))))
  \f
  
  ;; When an Info buffer is killed, make sure the associated tags buffer
--- 3690,3721 ----
                                 '(font-lock-face info-xref
                                                  mouse-face highlight
                                                  help-echo "mouse-2: go to this URL"))))
!       (goto-char (point-min))
!       (when Info-fontify-quotations-flag (info-fontify-quotations))
        (set-buffer-modified-p nil))))
+ 
+ 
+ ;; The regexp has these parts: double-quoted string or single-quoted stuff.
+ ;;
+ ;; String has, inside "...", zero or more of these characters:
+ ;;   - any character except \ and "
+ ;;   - \ followed by any character
+ ;;
+ ;; Single-quoted stuff has, inside `...': any character except '
+ ;;
+ (defun info-fontify-quotations ()
+     "Fontify double-quote strings (\"...\") and text between single-quotes (`...')
+ For single-quotes, use face `info-quoted-name'.
+ For double-quotes, use face `info-string'."
+     (while
+         (re-search-forward
+          "\"\\(?:[^\\\"]\\|\\\\\\(?:.\\|[\n]\\)\\)*\"\\|`[^']+'"
+          nil t)
+       (if (eq ?` (aref (match-string 0) 0))
+           (put-text-property (1+ (match-beginning 0)) (1- (match-end 0))
+                              'face 'info-quoted-name)
+         (put-text-property (match-beginning 0) (match-end 0)
+                            'face 'info-string))))
  \f
  
  ;; When an Info buffer is killed, make sure the associated tags buffer

Diff finished at Wed Oct 06 13:51:40

[-- Attachment #3: Type: text/plain, Size: 142 bytes --]

_______________________________________________
Emacs-devel mailing list
Emacs-devel@gnu.org
http://lists.gnu.org/mailman/listinfo/emacs-devel

RE: info faces for strings and quotations

[Drew Adams]

[-- Attachment #1.1: Type: text/plain, Size: 3038 bytes --]

Thanks, Bob. I'm glad someone played with the code and reported on what it
does. The incidents you mention where highlighting was not correct are the
kinds of thing I found too. I missed the last incident you mentioned. I did
run quickly through all the nodes of 33 manuals late last night, and I
probably missed a few others, as well. I wasn't hoping to be exhaustive in
my report, but to give a good idea of 1) the relative frequency of problems
(small) and 2) the kinds of problems encountered.

The problems you mention fall into this class: `...' highlighting does not
allow for unescaped enclosed ' marks, as in `foobar 'baz'. Highlighting here
will be only around the word (and space) "foobar ". So, in particular, it
will fail on single-quoted Lisp sexps with quotes in them, such as `(foobar
'baz)'. I think this is something we can live with.
  The Info buffer fails to highlight properly when the Info file is produced
from a correct Texinfo source... `(apply 'max numbers-list)'  Using the
patch, only `apply' is highlighted.
That's what I reported too. I guess you are confirming this and also
reporting that the Info text was correctly generated from Texinfo source.
Yes, there is nothing wrong with the Info text -- the highlighting fails on
it; that's all.
  There are also problems with @samp, which also puts single quotes around
its arguments, as in `Symbol's function definition is void: this'. Only
`Symbol' is highlighted.
Yes, for the same reason. I guess you are also saying something about the
mapping from Texinfo @samp. Again, there is nothing wrong with the Info
text.

The highlighting is not very smart, but it works most of the time. Note too
that, in this case, it would in fact be more proper (from a doc point of
view), to use double-quotes around the English expression that is being
discussed -- and in that case the highlighting works as it should. I'm not
saying that using @samp is wrong here; I'm just saying that in technical
documentation double-quotes would normally be used here.

FYI, here is the paragraph you cited, with double-quotes instead of
single-quotes, as it would appear highlighted:
  The error message can be understood: "Symbol's function definition is
void: this".  The symbol (that is, the word `this') lacks instructions for
the computer to carry out.
Don't get me wrong -- I'm not suggesting that this Info text needs to be
changed, or that any text should be changed just to fit the highlighting
code!  I'm simply saying: 1) yes, the highlighting code fails on text like
this, 2) this particular text would perhaps be better using double-quotes,
anyway (for reasons unrelated to highlighting), and 3) if the text used
double-quotes there would be no highlighting problem here.

Another common case that I reported on last night was backslashes between
single-quotes, where the backslashes were not escaping anything. I've fixed
the regexp to treat this, and have sent a new patch to the list
(separately).

Thanks for spending time on this and reporting carefully.

 - Drew

[-- Attachment #1.2: Type: text/html, Size: 5318 bytes --]

[-- Attachment #2: Type: text/plain, Size: 142 bytes --]

_______________________________________________
Emacs-devel mailing list
Emacs-devel@gnu.org
http://lists.gnu.org/mailman/listinfo/emacs-devel

Re: info faces for strings and quotations

[Juri Linkov]

Apart from being a very useful feature, this also helps to find errors
visually in Info manuals.  For example, I noticed

1. one backquote missing in emacs-lisp-intro.texi:

  day alternative for the decision `if it is warm and sunny, then go to
  the beach, else read a book!''.

which should be:

  day alternative for the decision ``if it is warm and sunny, then go to
  the beach, else read a book!''.

2. closing paren missing in calc.texi:

-to a function are somehow of the wrong type (@cite{@t{tan}([2,3,4])},
+to a function are somehow of the wrong type (@cite{@t{tan}([2,3,4])}),

3. extra backslash in gnus.texi:

-      (any "mypackage@@somewhere\" - "bugs-mypackage" "mypkg.list")
+      (any "mypackage@@somewhere" - "bugs-mypackage" "mypkg.list")

But OTOH, there are some places where it fails on legitimate constructs.
Most serious cases I found where this feature completely fails are

1. (info "(eintr)else") with

  `"It's a tiger!"'; but when you evaluate `(type-of-animal 'zebra)',
  you will see `"It's not fierce!"'.

2. (info "(gnus)Group Parameters") where after the first quote
on the line

            if address \"sender\" \"sieve-admin@extundo.com\" {

the rest of the node is fontified in string face.

While I think this feature is useful enough to be ON by default,
expect a lot of bug reports unless these problems are fixed.

-- 
Juri Linkov
http://www.jurta.org/emacs/

Re: info faces for strings and quotations

[Juri Linkov]

"Drew Adams" <drew.adams@oracle.com> writes:
> In most Info files, `...' is used to identify commands, key
> sequences, filenames and the like; "..." is used (usually) to
> identify strings in code. I found that not highlighting the
> single-quotes and highlighting the double-quotes improved
> readability.

I guess this is the same reason why in programming modes like Emacs
Lisp mode single-quotes are not highlighted while double-quotes are.

With the same reason to be similar to quotes in programming modes
I suggest to inherit new Info faces from corresponding standard faces
by default:

(defface info-quoted-name               ; For `...'
  '((t (:inherit font-lock-constant-face)))
  "Face used for quoted names (`...') in `info'."
  :group 'info)

(defface info-string                    ; For "..."
  '((t (:inherit font-lock-string-face)))
  "Face used for strings (\"...\") in `info'."
  :group 'info)

-- 
Juri Linkov
http://www.jurta.org/emacs/

Re: info faces for strings and quotations

[Juri Linkov]

"Drew Adams" <drew.adams@oracle.com> writes:
> EMACS FAQ -- several nodes have `...' or "..." in node titles. These are
> highlighted.
>
> WIDGET -- several nodes have `...' in titles
>
> REFTEX -- node fancyref (LaTeX package): `fancyref' in title highlighted

Quotes highlighted in titles are not a problem.  Just place the code
for quote fontification above the title fontification to override the
`font-lock-face' property with quote faces by title faces.

-- 
Juri Linkov
http://www.jurta.org/emacs/

RE: info faces for strings and quotations

[Drew Adams]

Regarding the two kinds of pb you mention:

1. I think the (sometimes) incorrect treatment of one or more single-quotes
inside `...' is something we should not bother to try to fix. Attempts to do
so will make for other pbs elsewhere. How can you know which is the quoted
name to be highlighted here: xxx`yyy'zzz'www -- yyy'zzz or just yyy?

Consider:

 - `foo' is George's plan...         -- we should highlight foo
 - `(funcall 'foo)' is what I think  -- we should highlight (funcall 'foo)

Note: font-lock highlighting of even something as simple as `C-x 2' fails,
because of the space. (Or, let's say it doesn't fail; it was designed to not
highlight this.)


2. The pb of not interpreting " to begin a string if it is escaped (\") is
no doubt fixable, at the risk of complicating the regexp. Suggested fixes
welcome.

Thanks,

  Drew


-----Original Message-----From: Juri Linkov

OTOH, there are some places where it fails on legitimate constructs.

1. `"It's a tiger!"'; but when you evaluate `(type-of-animal 'zebra)',
  you will see `"It's not fierce!"'.

2. if address \"sender\" \"sieve-admin@extundo.com\" {

the rest of the node is fontified in string face.

While I think this feature is useful enough to be ON by default,
expect a lot of bug reports unless these problems are fixed.

RE: info faces for strings and quotations

[Drew Adams]

[-- Attachment #1: Type: text/plain, Size: 813 bytes --]

Great ideas. Patch attached, in case you want to try it.

 - Drew

-----Original Message-----
From: Juri Linkov [mailto:juri@jurta.org]

1. With the same reason to be similar to quotes in programming modes
I suggest to inherit new Info faces from corresponding standard faces
by default:

(defface info-quoted-name               ; For `...'
  '((t (:inherit font-lock-constant-face)))
  "Face used for quoted names (`...') in `info'."
  :group 'info)

(defface info-string                    ; For "..."
  '((t (:inherit font-lock-string-face)))
  "Face used for strings (\"...\") in `info'."
  :group 'info)


2. Quotes highlighted in titles are not a problem.  Just place the code
for quote fontification above the title fontification to override the
`font-lock-face' property with quote faces by title faces.


[-- Attachment #2: diff-info-highlight-quotations-third.txt --]
[-- Type: text/plain, Size: 3377 bytes --]

diff -c "c:/emacs-21.3.50/lisp/info.el" "c:/drews-lisp-20/info-w-quotes-third.el"
*** c:/emacs-21.3.50/lisp/info.el	Mon Jul 26 09:42:06 2004
--- c:/drews-lisp-20/info-w-quotes-third.el	Thu Oct  7 00:10:04 2004
***************
*** 65,70 ****
--- 65,87 ----
  The Lisp code is executed when the node is selected.")
  (put 'Info-enable-active-nodes 'risky-local-variable t)
  
+ (defcustom Info-fontify-quotations-flag t
+   "*Non-nil means `info' fontifies text between quotes.
+ This applies to double-quote strings (\"...\") and text between
+ single-quotes (`...')."
+   :type 'boolean
+   :group 'info)
+ 
+ (defface info-quoted-name               ; For `...'
+   '((t (:inherit font-lock-constant-face)))
+   "Face used for quoted names (`...') in `info'."
+   :group 'info)
+ 
+ (defface info-string                    ; For "..."
+   '((t (:inherit font-lock-string-face)))
+   "Face used for strings (\"...\") in `info'."
+   :group 'info)
+ 
  (defface info-node
    '((((class color) (background light)) :foreground "brown" :weight bold :slant italic)
      (((class color) (background dark)) :foreground "white" :weight bold :slant italic)
***************
*** 3416,3421 ****
--- 3433,3442 ----
                (skip-chars-backward " \t,")
                (put-text-property (point) header-end 'invisible t)))))
  
+       ;; Fontify `...' and "..."
+       (goto-char (point-min))
+       (when Info-fontify-quotations-flag (info-fontify-quotations))
+ 
        ;; Fontify titles
        (goto-char (point-min))
        (when not-fontified-p
***************
*** 3669,3676 ****
                                 '(font-lock-face info-xref
                                                  mouse-face highlight
                                                  help-echo "mouse-2: go to this URL"))))
- 
        (set-buffer-modified-p nil))))
  \f
  
  ;; When an Info buffer is killed, make sure the associated tags buffer
--- 3690,3719 ----
                                 '(font-lock-face info-xref
                                                  mouse-face highlight
                                                  help-echo "mouse-2: go to this URL"))))
        (set-buffer-modified-p nil))))
+ 
+ 
+ ;; The regexp has these parts: double-quoted string or single-quoted stuff.
+ ;;
+ ;; String has, inside "...", zero or more of these characters:
+ ;;   - any character except \ and "
+ ;;   - \ followed by any character
+ ;;
+ ;; Single-quoted stuff has, inside `...': any character except '
+ ;;
+ (defun info-fontify-quotations ()
+     "Fontify double-quote strings (\"...\") and text between single-quotes (`...')
+ For single-quotes, use face `info-quoted-name'.
+ For double-quotes, use face `info-string'."
+     (while
+         (re-search-forward
+          "\"\\(?:[^\\\"]\\|\\\\\\(?:.\\|[\n]\\)\\)*\"\\|`[^']+'"
+          nil t)
+       (if (eq ?` (aref (match-string 0) 0))
+           (put-text-property (1+ (match-beginning 0)) (1- (match-end 0))
+                              'font-lock-face 'info-quoted-name)
+         (put-text-property (match-beginning 0) (match-end 0)
+                            'font-lock-face 'info-string))))
  \f
  
  ;; When an Info buffer is killed, make sure the associated tags buffer

Diff finished.  Thu Oct 07 00:12:15 2004

[-- Attachment #3: Type: text/plain, Size: 142 bytes --]

_______________________________________________
Emacs-devel mailing list
Emacs-devel@gnu.org
http://lists.gnu.org/mailman/listinfo/emacs-devel

Re: info faces for strings and quotations

[Stefan Monnier]

> 2. The pb of not interpreting " to begin a string if it is escaped (\") is
> no doubt fixable, at the risk of complicating the regexp. Suggested fixes
> welcome.

Prefix the regexp with "\\(^\\|[^\\]\\)\\(\\\\\\\\\\)*".  The (match-end) of
this regexp is guaranteed to not be preceded by an odd number
of backspashes.

You can find this regexp used by various major modes for comment-start-skip.


        Stefan

Re: info faces for strings and quotations

[David Kastrup]

Stefan Monnier <monnier@iro.umontreal.ca> writes:

>> 2. The pb of not interpreting " to begin a string if it is escaped (\") is
>> no doubt fixable, at the risk of complicating the regexp. Suggested fixes
>> welcome.
>
> Prefix the regexp with "\\(^\\|[^\\]\\)\\(\\\\\\\\\\)*".  The (match-end) of
> this regexp is guaranteed to not be preceded by an odd number
> of backspashes.
>
> You can find this regexp used by various major modes for comment-start-skip.

If you don't need regexps, it is usually quite faster just to check
with 
(zerop (mod (save-excursion (skip-chars-backward "\\\\")) 2))
whether you have an even number of preceding backslashes.

-- 
David Kastrup, Kriemhildstr. 15, 44793 Bochum

Re: info faces for strings and quotations

[Kim F. Storm]

Stefan Monnier <monnier@iro.umontreal.ca> writes:

>> 2. The pb of not interpreting " to begin a string if it is escaped (\") is
>> no doubt fixable, at the risk of complicating the regexp. Suggested fixes
>> welcome.
>
> Prefix the regexp with "\\(^\\|[^\\]\\)\\(\\\\\\\\\\)*".  The (match-end) of
> this regexp is guaranteed to not be preceded by an odd number
> of backspashes.

Wouldn't it be fairly trivial to enhance regex to recognize something like

         \\\{0,e}

meaning an even (possibly zero) number of \'es.

(likewise for \{0,o}, meaning an odd number of (or no) \'es )

-- 
Kim F. Storm <storm@cua.dk> http://www.cua.dk

w3 mode

[Camm Maguire]

Greetings!  Before getting into the details, I notice in general that
w3 mode works much better in xemacs.  Is this generally known, or do I
have a problem in my local setup?

Take care,
-- 
Camm Maguire			     			camm@enhanced.com
==========================================================================
"The earth is but one country, and mankind its citizens."  --  Baha'u'llah

unexec development

[Camm Maguire]

Greetings!  Currently, GCL and programs built on top of it
(maxima,acl2,axiom), use emacs' unexec by default to save intermediary
and final system images.  Compiled code is loaded and relocated
dynamically at runtime, stored in an expanding .data section via sbrk,
and finally dumped with unexec.  

I would like to extend this functionality if possible to include
compiled code which references symbols in external shared libs opened
again dynamically at runtime via dlopen.  Loading and relocating is
straightforward, but somehow I'd need to have unexec write a
new relocation record into the final image for each symbol for use by
ld.so.  

Does this sound of interest to emacs too?  Does this trigger any ideas
as to a possibly better approach?

Take care,
-- 
Camm Maguire			     			camm@enhanced.com
==========================================================================
"The earth is but one country, and mankind its citizens."  --  Baha'u'llah

Re: info faces for strings and quotations

[David Kastrup]

storm@cua.dk (Kim F. Storm) writes:

> Stefan Monnier <monnier@iro.umontreal.ca> writes:
>
>>> 2. The pb of not interpreting " to begin a string if it is escaped (\") is
>>> no doubt fixable, at the risk of complicating the regexp. Suggested fixes
>>> welcome.
>>
>> Prefix the regexp with "\\(^\\|[^\\]\\)\\(\\\\\\\\\\)*".  The (match-end) of
>> this regexp is guaranteed to not be preceded by an odd number
>> of backspashes.
>
> Wouldn't it be fairly trivial to enhance regex to recognize something like
>
>          \\\{0,e}
>
> meaning an even (possibly zero) number of \'es.
>
> (likewise for \{0,o}, meaning an odd number of (or no) \'es )

That's too ad-hoc.  Rather use \{0,,2} and \{1,,2}.

Anyway, it does not buy much that I can see here: you still need to
make sure that there is no leading \ before the expression in
question.

-- 
David Kastrup, Kriemhildstr. 15, 44793 Bochum

Re: info faces for strings and quotations

[Stefan Monnier]

> If you don't need regexps, it is usually quite faster just to check with
> (zerop (mod (save-excursion (skip-chars-backward "\\\\")) 2)) whether you
> have an even number of preceding backslashes.

That's true.  The "\\(^\\|[^\\]\\)\\(\\\\\\\\\\)*" prefix can slow down
regexp searching significantly.


        Stefan

Re: w3 mode

[Stefan Monnier]

> Greetings!  Before getting into the details, I notice in general that
> w3 mode works much better in xemacs.  Is this generally known, or do I
> have a problem in my local setup?

IIRC W3 was developed on XEmacs (Emacs didn't have image support at that
time), so it's no surprise it works better there, especially since there
hasn't been much work put into W3 development recently (e.g. since Emacs
provides support for images).


        Stefan

Re: w3 mode

[Camm Maguire]

Greetings, and thanks for the reply!

So I take it that no one is actively working on this, or some other
web mode, for GNU emacs at present?  Just wondering.

Take care,

Stefan Monnier <monnier@iro.umontreal.ca> writes:

> > Greetings!  Before getting into the details, I notice in general that
> > w3 mode works much better in xemacs.  Is this generally known, or do I
> > have a problem in my local setup?
> 
> IIRC W3 was developed on XEmacs (Emacs didn't have image support at that
> time), so it's no surprise it works better there, especially since there
> hasn't been much work put into W3 development recently (e.g. since Emacs
> provides support for images).
> 
> 
>         Stefan
> 
> 
> 

-- 
Camm Maguire			     			camm@enhanced.com
==========================================================================
"The earth is but one country, and mankind its citizens."  --  Baha'u'llah

Re: w3 mode

[Mark Plaksin]

Camm Maguire <camm@enhanced.com> writes:

> Greetings, and thanks for the reply!
>
> So I take it that no one is actively working on this, or some other
> web mode, for GNU emacs at present?  Just wondering.

Emacs-w3m is a nice web browser for Emacs.  It uses w3m.
http://emacs-w3m.namazu.org/

Re: w3 mode

[Kevin Rodgers]

Camm Maguire wrote:
 > So I take it that no one is actively working on this, or some other
 > web mode, for GNU emacs at present?  Just wondering.

Definitely not;

http://emacs-w3m.namazu.org/

-- 
Kevin Rodgers

Re: info faces for strings and quotations

[Luc Teirlinck]

Juri Linkov wrote:

   But OTOH, there are some places where it fails on legitimate constructs.

That is unavoidable.  Info files do not contain enough information to
do the job reliably.  The .texi source files do but using the .texi
files is taboo, because it would not allow the Texinfo language to
change independently of Emacs.  The XML output format does.  Oliver
Scholz is working on that.

   While I think this feature is useful enough to be ON by default,
   expect a lot of bug reports unless these problems are fixed.

That is exactly why the feature should not only be off by default, but
that in addition the user should be told that the feature does not aim
for perfection and occasionally can result in anomalous fontification.
The user has to know that these problems are a known and unavoidable
consequence of the heuristic, as opposed to fixable bugs, and that
filing a bug report would be futile.  The user should know that if he
is not willing to live with these problems, he should simply not turn
the feature on.

Sincerely,

Luc.

RE: info faces for strings and quotations

[Drew Adams]

[-- Attachment #1: Type: text/plain, Size: 3022 bytes --]

This version works well. The fontify function is in line below; the full
patch is attached. This seems to do everything the previous version did,
plus treat initial \" correctly. Things like this from (ccmode)Indentation
Functions are not highlighted now, which is the correct behavior: (\"Hello
world!\");

Thanks to everyone who contributed suggestions. I went with David's
suggestion for treating initial \".

Per Luc's suggestion, I added a note to the doc string of
`Info-fontify-quotations-flag' explaining that fontification can never be
100% reliable, that occasional inappropriate highlighting is not a bug, and
that highlighting can be turned off. It is now off by default.

I won't be working any more on this. I think it is ready for prime time, and
I hope the decision will be to add it to Emacs.

BTW, with this highlighting, I noticed these additional Info text bugs (the
first one I had noticed earlier, but forgot to point out):

 - (eintr)the-the -- "the' should be "the"
 - (gnus)Fancy Mail Splitting: (any "mypackage@somewhere\" -
"bugs-mypackage" "mypkg.list") -- I'm guessing that the backslash doesn't
belong here.
 - (elisp)Specification List -- `(spec .  [(more specs...)])') -- ' is in
wrong place.

 - Drew

;; The regexp has these parts: double-quoted string or single-quoted stuff.
;;
;; String has, inside "...", zero or more of these characters:
;;   - any character except \ and "
;;   - \ followed by any character
;;
;; Single-quoted stuff has, inside `...': any character except '
;;
;; The regexp matches `...' or "...". The latter case also includes
;; pseudo-strings that start with \". The second branch of the
;; conditional tests this: if that is the match, then it moves past
;; the \.
(defun info-fontify-quotations ()
  "Fontify text between double-quotes (\"...\") and single-quotes (`...')
For single-quotes, use face `info-quoted-name'.
For double-quotes, use face `info-string'."
  (while (re-search-forward             ; Match `...' or "..."
          "\"\\(?:[^\\\"]\\|\\\\\\(?:.\\|[\n]\\)\\)*\"\\|`[^']+'"
          nil t)
    (cond ((eq ?` (aref (match-string 0) 0))
           (put-text-property (1+ (match-beginning 0)) (1- (match-end 0))
                              'font-lock-face 'info-quoted-name))
          ((and (goto-char (match-beginning 0)) ; "...": If preceded by \,
skip it
                (= 1 (mod (save-excursion (skip-chars-backward "\\\\")) 2)))
           (forward-char 1))
          ((goto-char (match-end 0))    ; "..." not preceded by \
           (put-text-property (match-beginning 0) (match-end 0)
                              'font-lock-face 'info-string)))))


-----Original Message-----From: Stefan, quoting David, replying to Stefan
> If you don't need regexps, it is usually quite faster just to check with
> (zerop (mod (save-excursion (skip-chars-backward "\\\\")) 2)) whether you
> have an even number of preceding backslashes.

That's true.  The "\\(^\\|[^\\]\\)\\(\\\\\\\\\\)*" prefix can slow down
regexp searching significantly.

[-- Attachment #2: diff-info-highlight-quotations-fourth.txt --]
[-- Type: text/plain, Size: 4215 bytes --]

diff -c "c:/emacs-21.3.50/lisp/info.el" "c:/drews-lisp-20/info-w-quotes-fourth.el"
*** c:/emacs-21.3.50/lisp/info.el	Mon Jul 26 09:42:06 2004
--- c:/drews-lisp-20/info-w-quotes-fourth.el	Thu Oct  7 21:30:04 2004
***************
*** 65,70 ****
--- 65,94 ----
  The Lisp code is executed when the node is selected.")
  (put 'Info-enable-active-nodes 'risky-local-variable t)
  
+ (defcustom Info-fontify-quotations-flag nil
+   "*Non-nil means `info' fontifies text between quotes.
+ This applies to double-quote strings (\"...\") and text between
+ single-quotes (`...').
+ 
+ Note: This fontification can never be 100% reliable. It aims to
+ be useful in most Info texts, but it can occasionally result in
+ fontification that you might not expect. This is not a bug; it is
+ part of the design to be able to appropriately fontify a great
+ variety of texts. Set this flag to nil if you do not find this
+ fontification useful."
+   :type 'boolean
+   :group 'info)
+ 
+ (defface info-quoted-name               ; For `...'
+   '((t (:inherit font-lock-constant-face)))
+   "Face used for quoted names (`...') in `info'."
+   :group 'info)
+ 
+ (defface info-string                    ; For "..."
+   '((t (:inherit font-lock-string-face)))
+   "Face used for strings (\"...\") in `info'."
+   :group 'info)
+ 
  (defface info-node
    '((((class color) (background light)) :foreground "brown" :weight bold :slant italic)
      (((class color) (background dark)) :foreground "white" :weight bold :slant italic)
***************
*** 3416,3421 ****
--- 3440,3449 ----
                (skip-chars-backward " \t,")
                (put-text-property (point) header-end 'invisible t)))))
  
+       ;; Fontify `...' and "..."
+       (goto-char (point-min))
+       (when Info-fontify-quotations-flag (info-fontify-quotations))
+ 
        ;; Fontify titles
        (goto-char (point-min))
        (when not-fontified-p
***************
*** 3669,3676 ****
                                 '(font-lock-face info-xref
                                                  mouse-face highlight
                                                  help-echo "mouse-2: go to this URL"))))
- 
        (set-buffer-modified-p nil))))
  \f
  
  ;; When an Info buffer is killed, make sure the associated tags buffer
--- 3697,3733 ----
                                 '(font-lock-face info-xref
                                                  mouse-face highlight
                                                  help-echo "mouse-2: go to this URL"))))
        (set-buffer-modified-p nil))))
+ 
+ 
+ ;; The regexp has these parts: double-quoted string or single-quoted stuff.
+ ;;
+ ;; String has, inside "...", zero or more of these characters:
+ ;;   - any character except \ and "
+ ;;   - \ followed by any character
+ ;;
+ ;; Single-quoted stuff has, inside `...': any character except '
+ ;;
+ ;; The regexp matches `...' or "...". The latter case also includes
+ ;; pseudo-strings that start with \". The second branch of the
+ ;; conditional tests this: if that is the match, then it moves past
+ ;; the \.
+ (defun info-fontify-quotations ()
+   "Fontify text between double-quotes (\"...\") and single-quotes (`...')
+ For single-quotes, use face `info-quoted-name'.
+ For double-quotes, use face `info-string'."
+   (while (re-search-forward             ; Match `...' or "..."
+           "\"\\(?:[^\\\"]\\|\\\\\\(?:.\\|[\n]\\)\\)*\"\\|`[^']+'"
+           nil t)
+     (cond ((eq ?` (aref (match-string 0) 0))
+            (put-text-property (1+ (match-beginning 0)) (1- (match-end 0))
+                               'font-lock-face 'info-quoted-name))
+           ((and (goto-char (match-beginning 0)) ; "...": If preceded by \, skip it
+                 (= 1 (mod (save-excursion (skip-chars-backward "\\\\")) 2)))
+            (forward-char 1))
+           ((goto-char (match-end 0))    ; "..." not preceded by \
+            (put-text-property (match-beginning 0) (match-end 0)
+                               'font-lock-face 'info-string)))))
  \f
  
  ;; When an Info buffer is killed, make sure the associated tags buffer

[-- Attachment #3: Type: text/plain, Size: 142 bytes --]

_______________________________________________
Emacs-devel mailing list
Emacs-devel@gnu.org
http://lists.gnu.org/mailman/listinfo/emacs-devel

Re: info faces for strings and quotations

[Richard Stallman]

    Apart from being a very useful feature, this also helps to find errors
    visually in Info manuals.

Thanks for noticing--will you fix them?

    But OTOH, there are some places where it fails on legitimate constructs.
    Most serious cases I found where this feature completely fails are

    1. (info "(eintr)else") with

      `"It's a tiger!"'; but when you evaluate `(type-of-animal 'zebra)',
      you will see `"It's not fierce!"'.

What Texinfo code generates that?  It looks like someone put @samp
around a Lisp string constant.  If so, that is incorrect usage; it
should be @code, not @samp.

Re: w3 mode

[Richard Stallman]

    Greetings!  Before getting into the details, I notice in general that
    w3 mode works much better in xemacs.

That is so vague that it cannot be of any help.
If you state a specific behavior that you don't like,
maybe we can improve it.

Re: info faces for strings and quotations

[Luc Teirlinck]

Richard Stallman wrote:

       1. (info "(eintr)else") with

	 `"It's a tiger!"'; but when you evaluate `(type-of-animal 'zebra)',
	 you will see `"It's not fierce!"'.

   What Texinfo code generates that?  It looks like someone put @samp
   around a Lisp string constant.  If so, that is incorrect usage; it
   should be @code, not @samp.

That text _is_ already in @code.  Juri apparently cut and pasted that
from an *info* buffer.  It is impossible to tell from a .info file
whether `abc' was produced by @code{abc} or @samp{abc}.  You can see
the difference if you view the dvi output.

Sincerely,

Luc.

Re: w3 mode

[David Kastrup]

Richard Stallman <rms@gnu.org> writes:

>     Greetings!  Before getting into the details, I notice in general that
>     w3 mode works much better in xemacs.
>
> That is so vague that it cannot be of any help.
> If you state a specific behavior that you don't like,
> maybe we can improve it.

For what value of "we"?  AFAIR, w3 development is pretty stalled, and
code that deals with Emacs-21 images is more or less just present in
the CVS version (available from Savannah), maybe in the last
unofficial tarball as well.

Since w3 does not appear to be developed actively any more, it is
unclear who could do just what about the situation, and why.

-- 
David Kastrup, Kriemhildstr. 15, 44793 Bochum

Re: info faces for strings and quotations

[Robert J. Chassell]

        Apart from being a very useful feature, this also helps to
        find errors visually in Info manuals.

    Thanks for noticing--will you fix them?

I already fixed the typos in the intro.  It is a useful tool for that.

          `"It's a tiger!"'; but when you evaluate `(type-of-animal 'zebra)',
          you will see `"It's not fierce!"'.

    What Texinfo code generates that?  It looks like someone put @samp
    around a Lisp string constant.  If so, that is incorrect usage; it
    should be @code, not @samp.

It is @code.  That is what makeinfo produces for Info.  In a Web
browser such as Emacs W3, Emacs W3M, or Firefox, or when printed, you
see only the double quotes.

-- 
    Robert J. Chassell                         
    bob@rattlesnake.com                         GnuPG Key ID: 004B4AC8
    http://www.rattlesnake.com                  http://www.teak.cc

Re: info faces for strings and quotations

[Richard Stallman]

    That text _is_ already in @code.

I was misremembering the difference, sorry.

Re: unexec development

[Camm Maguire]

Greetings!  Just another ping on this issue.  Is anyone working on
unexec anymore?  Does anyone have interest in doing so?

Take care,

Camm Maguire <camm@enhanced.com> writes:

> Greetings!  Currently, GCL and programs built on top of it
> (maxima,acl2,axiom), use emacs' unexec by default to save intermediary
> and final system images.  Compiled code is loaded and relocated
> dynamically at runtime, stored in an expanding .data section via sbrk,
> and finally dumped with unexec.  
> 
> I would like to extend this functionality if possible to include
> compiled code which references symbols in external shared libs opened
> again dynamically at runtime via dlopen.  Loading and relocating is
> straightforward, but somehow I'd need to have unexec write a
> new relocation record into the final image for each symbol for use by
> ld.so.  
> 
> Does this sound of interest to emacs too?  Does this trigger any ideas
> as to a possibly better approach?
> 
> Take care,
> -- 
> Camm Maguire			     			camm@enhanced.com
> ==========================================================================
> "The earth is but one country, and mankind its citizens."  --  Baha'u'llah
> 
> 
> _______________________________________________
> Emacs-devel mailing list
> Emacs-devel@gnu.org
> http://lists.gnu.org/mailman/listinfo/emacs-devel
> 
> 
> 

-- 
Camm Maguire			     			camm@enhanced.com
==========================================================================
"The earth is but one country, and mankind its citizens."  --  Baha'u'llah

Re: unexec development

[Jan D.]

> Greetings!  Just another ping on this issue.  Is anyone working on
> unexec anymore?  Does anyone have interest in doing so?
>

Can you describe what the benefits for Emacs would be?  Or why you want  
this.

	Jan D.

> Camm Maguire <camm@enhanced.com> writes:
>
>> Greetings!  Currently, GCL and programs built on top of it
>> (maxima,acl2,axiom), use emacs' unexec by default to save intermediary
>> and final system images.  Compiled code is loaded and relocated
>> dynamically at runtime, stored in an expanding .data section via sbrk,
>> and finally dumped with unexec.
>>
>> I would like to extend this functionality if possible to include
>> compiled code which references symbols in external shared libs opened
>> again dynamically at runtime via dlopen.  Loading and relocating is
>> straightforward, but somehow I'd need to have unexec write a
>> new relocation record into the final image for each symbol for use by
>> ld.so.
>>
>> Does this sound of interest to emacs too?  Does this trigger any ideas
>> as to a possibly better approach?
>>
>> Take care,
>> --  
>> Camm Maguire			     			camm@enhanced.com
>> ====================================================================== 
>> ====
>> "The earth is but one country, and mankind its citizens."  --   
>> Baha'u'llah
>>
>>
>> _______________________________________________
>> Emacs-devel mailing list
>> Emacs-devel@gnu.org
>> http://lists.gnu.org/mailman/listinfo/emacs-devel
>>
>>
>>
>
> -- 
> Camm Maguire			     			camm@enhanced.com
> ======================================================================= 
> ===
> "The earth is but one country, and mankind its citizens."  --   
> Baha'u'llah
>
>
> _______________________________________________
> Emacs-devel mailing list
> Emacs-devel@gnu.org
> http://lists.gnu.org/mailman/listinfo/emacs-devel

Re: Re: unexec development

[Camm Maguire]

Greetings, and thanks for your reply!

"Jan D." <jan.h.d@swipnet.se> writes:

> > Greetings!  Just another ping on this issue.  Is anyone working on
> > unexec anymore?  Does anyone have interest in doing so?
> >
> 
> Can you describe what the benefits for Emacs would be?  Or why you
> want  this.
> 

Well, the benefits for emacs might not accrue in the immediate future.
To my very limited understanding, emacs only byte-compiles code loaded
into and executed from its heap.  GCL can compile, load, and execute
native code this way.  In principle at least, emacs might want to do
likewise at some point in the future for performance reasons.  I agree
it is far from pressing.  But the GCL compiler is written in lisp and
probably would work for emacs without too much work.

In such a setup, it makes sense that a user might want to compile and
load code into their heap which references external functions in
shared libraries which could be dlopened at runtime.  All is well
until the unexec, when the mapping from the heap to the dlopened
shared library is lost, as the start address of the latter is not
reliable..  Perhaps one might say to just relink by hand when the
saved image is restarted.  This is indeed a possibility.  But at least
in the case of gcl, this cycle of compile,load,save,restart occurs
many times in the course of building a particular application.  There
should be no reason we couldn't push the relinking work onto ld.so,
with the added advantage that the binary could be prelinked and thus
accelerated at the end.  Doing this in GCL would be to essentially
duplicate the functionality of ld.so.

As an aside, I notice that even without this, gcl images cannot be
prel at present, and I assume the same is for emacs, as the images are
formed by unexec.  This must be related to the lack of initialization
of some elf section I'd think.  Something else that could be addressed
at the same time.

Take care,


> 	Jan D.
> 
> > Camm Maguire <camm@enhanced.com> writes:
> >
> >> Greetings!  Currently, GCL and programs built on top of it
> >> (maxima,acl2,axiom), use emacs' unexec by default to save intermediary
> >> and final system images.  Compiled code is loaded and relocated
> >> dynamically at runtime, stored in an expanding .data section via sbrk,
> >> and finally dumped with unexec.
> >>
> >> I would like to extend this functionality if possible to include
> >> compiled code which references symbols in external shared libs opened
> >> again dynamically at runtime via dlopen.  Loading and relocating is
> >> straightforward, but somehow I'd need to have unexec write a
> >> new relocation record into the final image for each symbol for use by
> >> ld.so.
> >>
> >> Does this sound of interest to emacs too?  Does this trigger any ideas
> >> as to a possibly better approach?
> >>
> >> Take care,
> >> --
> >> Camm Maguire			     			camm@enhanced.com
> >> ======================================================================
> >> ====
> >> "The earth is but one country, and mankind its citizens."  --
> >> Baha'u'llah
> >>
> >>
> >> _______________________________________________
> >> Emacs-devel mailing list
> >> Emacs-devel@gnu.org
> >> http://lists.gnu.org/mailman/listinfo/emacs-devel
> >>
> >>
> >>
> >
> > -- 
> > Camm Maguire			     			camm@enhanced.com
> > =======================================================================
> > ===
> > "The earth is but one country, and mankind its citizens."  --
> > Baha'u'llah
> >
> >
> > _______________________________________________
> > Emacs-devel mailing list
> > Emacs-devel@gnu.org
> > http://lists.gnu.org/mailman/listinfo/emacs-devel
> 
> 
> 
> _______________________________________________
> Gcl-devel mailing list
> Gcl-devel@gnu.org
> http://lists.gnu.org/mailman/listinfo/gcl-devel
> 
> 
> 

-- 
Camm Maguire			     			camm@enhanced.com
==========================================================================
"The earth is but one country, and mankind its citizens."  --  Baha'u'llah

Re: unexec development

[Richard Stallman]

We're not doing any work to extend unexec for Emacs's sake.
However, it would not be objectionable to extend it for GCL's sake.