Re: How to edit "mixed" (e.g. HTML/JavaScript) files?

[David Hansen <david.hansen@gmx.net>]

On Thu, 08 Mar 2007 17:06:53 +0100 Lennart Borgman wrote:
> I guess you mean that it is bad to switch major mode because of that
> it takes long time, or are there other reasons?
>
> If it is the time to change that is important, what do you believe
> is that takes too much time?

I guess mainly the first font-lock run.

> When it comes to fontification the approach with indirect buffers
> can not work, unfortunately, since text properties (which are used
> for fontification) are shared between the buffers.

multi-mode works (or at least tries to works) the way that
fontification happens region by region only in the relevant buffer.
In theory this should work (long time ago that i had a look at the
sources).

> This together with that a major mode wants to fontify the whole
> buffer

There's `font-lock-dont-widen' (I think one of the glitches of
multi-mode is that it doesn't set this and syntactic fontification
in one mode may screws it in the other mode).

> makes it impossible as far as I can see. (There is nothing
> in Emacs 22 beta that can remedy this either.)

Have a look at the multi-mode sources.  It's a relatively small,
clean coded and good commented package (no not by me ;).

For my taste this is a minor problem anyway as long as the text
around the part of code I'm currently editing looks OK (and if not
there's M-o M-o).

What terrible sucks is indentation if some construct spreads over
more than one region.  But at least it works reliable within one
region.  MMM mode can't even that.

I assume NXML mode won't be that happy too if you narrow it to one
region (or does it widen for it's syntax analysis?  Would probably
be even worse).

>> I hope this topic will get some more attention after the upcoming
>> (?) release of Emacs 22.
>
> Yes, that would be good. But I think we need some experiences then
> about what can work.


> As far as I can see switching major mode might not be that bad if it
> could be applied to a part of a buffer.

This always implies unnecessary refontification which IMHO is *very*
bad.

David