Jason Rumney wrote:
> Lennart Borgman (gmail) wrote:
>> I have attached the relevant pieces from rng-valid.el. I think this is 
>> much easier to understand as one piece than as a diff.
> 
> It may be so for you, who wrote the changes, but for anyone else who 
> does not know what you've changed, a diff is much easier.

Ok, here comes the diff.

It is noteworthy that I actually use nxml-mode just to parse the buffer 
when it tie it together with mumamo.

You can not see that in this diff, since turning of the nxml-mode 
fontification is done in mumamo.

After "finishing" the change I sent here to rng-valid.el I realize I can 
now use the fontification from nxml-mode together with mumamo. That was 
not possible before because the nxml-mode fontification would then 
override fontification in chunks of the buffer with other major modes.

Daniel, I am starting to wonder if that is possible with your changes. I 
actually wonder if I can use the parser from nxml-mode at all. I have 
not thought much about it but it is not clear to me how it can be done yet.

My thoughts on this matter before has been that perhaps a two step 
approch to fontification would be good:

- In the first pass a font-lock keyword fontification would be done.
- In the second step a parser like nxml-mode or js2 mode could make it a 
little bit better.

This would mean that the parser have to cooperate with font-lock in some 
way. I do not know how, but the font-lock/jit lock frame work could 
possible be extended to do handle the cooperation.

I think however that there perhaps is no need for the parser to know 
about fontify-region. It would rather need to know things like how to 
tell font-lock what it has done and where major mode chunks are. (Mumamo 
does the major mode chunk dividing in font-lock-fontify-region-function.)