Re: Issues with japanese support

* Masayuki Ataka (2005-02-16) writes:

> From: Ralf Angeli <[email protected]>
> Subject: Re: Issues with japanese support
> Date: Wed, 16 Feb 2005 10:28:48 +0100
>
>> Although I have a `re-search-forward' phobia since we had these
>> performance problems related to filling, I think this would be a
>> pragmatic solution.  At least there is not much difference in
>> execution time between the following two implementations (work only
>> with Emacs):
>> 
> Let me name the code for convenience.
>
> fill-cvs-emacs:
[...]
> fill-replace:
[...]
> And new code.
>
> fill-skip-char:
[...]
> In my computer and Emacs-21.4, `fill-re-search' is faster
> than `fill-cvs-emacs':
>   (fill-cvs-emacs) 2.065401077270508
>   (fill-replace)   1.1991240978240967
>   (fill-skip-char) 0.9685571193695068
> But, after byte-compile, `fill-cvs-emacs' becomes much
> faster:
>   (fill-cvs-emacs) 0.5713150501251221
>   (fill-replace)   1.1394240856170654
>   (fill-skip-char) 0.9528110027313232

Argh, byte compilation ... (c:

> Yes.  `re-search-forward' has a performance problem.
> `fill-cvs-emacs' is the fastest.  `fill-skip-char' is better
> than `re-search-forward'.  So, our new function will be...
>
> (defun LaTeX-fill-delete-newlines (from to justify nosqueeze squeeze-after)
>   ;; COMPATIBILITY for Emacs <= 22.1

This should be "... for Emacs < 22.1 and XEmacs".

>   (if (fboundp 'fill-delete-newlines)
>       (fill-delete-newlines from to justify nosqueeze squeeze-after)
>     (if (featurep 'xemacs)
>         (when (featurep 'mule)
>           (goto-char from)
>           (while (re-search-forward "\\(\\cj\\)\n\\(\\cj\\)" to t)

It would be good to use something like the value of
`word-across-newline' in the regular expression,
i.e. "\\(\\cj\\|\\cc\\|\\ct\\)\n\\(\\cj\\|\\cc\\|\\ct\\)".  If the
variable `word-across-newline' should be used, one somehow will have
to silence the byte compiler.

>               (skip-chars-backward "^\n")
>               (delete-char -1)))
>       ;; This else-sentence was copied from the function `fill-delete-newlines'
>       ;; in `fill.el' (CVS Emacs, 2005-02-17).
>       (while (search-forward "\n" nil t)
>         (let ((prev (char-before (match-beginning 0)))
>               (next (following-char)))
>           (when (or (aref (char-category-set next) ?|)
>                     (aref (char-category-set prev) ?|))
>             (delete-char -1)))))
>
> (snip) (snip) (snip)
> ;; Is comment appropriate?

Yep.  You could add something like "... and adapted accordingly"
because the checks for the fill-space text property were removed.
(AFAICS this property is not used in Emacs 21.3.)

>> What I don't like is the name `LaTeX-word-across-newline'.  What about
>> something matching better its value/purpose, like
>> `LaTeX-nospace-between-words-char-regexp'.  Argh, I created a
>> monster.  A little bit shorter: `LaTeX-nospace-between-char-regexp'.
>> Any better suggestions?
>
> I like it.

With the above function we don't even need it.

-- 
Ralf