Re: html, css, and js modes working together

[Dmitry Gutov]

On 06.02.2017 05:26, Tom Tromey wrote:

> I'm not sure what you're referring to here.  Do some of the existing
> mixed mode things piggyback on the existing html mode?

Both https://github.com/purcell/mmm-mode/ and 
https://github.com/vspinu/polymode/ do.

> I guess my view is that it is always ok to make things in-tree work
> better with each other, even at the expense of some code that is
> out-of-tree and presumably relying on implementation details to do its
> work.

Still, defining a new major mode instead of directly reusing an existing 
one shouldn't take a lot of effort. Especially while the result is 
functional but not ideal.

And breaking packages that worked fine for many years is not nice, 
especially when it can be avoided.

> Dmitry> So far it's only used in python-mode, I think.
>
> grep shows no users in-tree, so maybe this never went in?  Or is this
> one of the various out-of-tree python modes?

lisp/progmodes/python.el uses prog-first-column and prog-widen, both of 
which refer to prog-indentation-context.

> Yeah, what I mean is that font-locking does not occur, but it would be
> good to have.  I couldn't find anything saying how this might be solved.
> I haven't come up with any really good ideas myself.  Maybe font-lock
> could also look at text properties to decide what keywords to use?

Here's one way to do it: 
https://github.com/purcell/mmm-mode/blob/master/mmm-region.el#L768-L787

> Dmitry> - Both css-mode and js-mode call syntax-ppss in their indentation
> Dmitry> code. Luckily, their syntax tables are fairly compatible. But
> Dmitry> sgml-syntax-propertize-rules calls syntax-ppss as well.
>
> I thought that a syntax-table property on the characters would make
> syntax-ppss do the "right" thing; namely, notice where the syntax table
> changes and change its parsing method accordingly.

That seems orthogonal to the possibility of breaking the cache after 
calling syntax-ppss while a different syntax table is in effect.

The characters with the syntax-table text property set should be the 
same in both "world views", but they are usually a minority.

> I didn't try your example yet.  Thanks for providing that, that's very
> helpful.

No problem. BTW, I've worked around it with a hack like this: 
https://github.com/purcell/mmm-mode/blob/master/mmm-erb.el#L93-L98