bug#57447: 28.1.90; Can font-lock stop requesting fontification of invisible text?

[Ihor Radchenko <yantar92@gmail.com>]

Eli Zaretskii <eliz@gnu.org> writes:

>> Can anything be done to make Emacs fontify less (or none) of the hidden
>> text?
>
> I explained that recently, in a discussion in which you participated:
> the display engine skips invisible text, but that cannot completely
> avoid fontifications of some of the invisible text.
>
> The technical reason for that is that the display engine considers the
> 'fontified' property before it considers the 'invisible' property (and
> any other properties).  Since fontification-functions put text
> properties on buffer text, the 'fontified' property _must_ be examined
> first, and the fontification-functions _must_ be invoked _before_ the
> properties are examined; otherwise we'd risk behaving incorrectly due
> to outdated or not-yet-existing properties.

This issue surfaced exactly because I was hoping that the overheads are
not going to be too terrible if we leave Emacs decide what to fontify.
In the past, Org explicitly suppressed fontification of all the hidden
text.

I did expect some invisible text to be fontified, but did not expect 10x
increase in the requested text size.

> One way this could be improved is by making jit-lock-chunk-size
> smaller.  This will make the probability of such "over-fontification"
> lower.

I tried to reduce the chunk size and at appears to help. However, I am
not sure if there are other side effects of reducing
jit-lock-chunk-size.

> If a mode _knows_ that no fontifications could ever produce invisible
> properties or overlays, then its fontification-functions could stop
> whenever they encounter invisible text.

This is not so simple in Org, unfortunately. Sometimes, we do want to
fontify invisible text for the purposes of imenu/agenda/etc.

The code required to handle invisibility manually gets quite complex:
https://git.savannah.gnu.org/cgit/emacs/org-mode.git/commit/?id=dd0a723603ec4c9085a10ac8fbf864fdb026e48a

>> The fontification log is:
>> 
>> org-font-lock: About to fontify 13481..15094
>> org-font-lock: Fontified drawer(13372..13487) completely.
>
> Thanks, but this log is insufficient to analyze whether the code
> functions correctly.  To do such an analysis one must know which text
> parts (in therms of buffer positions) were invisible and which
> weren't.  Also, you haven't shown the code which produces the log, so
> the exact meanings of "about to fontify", "fontfied drawer", and the
> rest of the log messages, as well as the numerical notations like
> "13487:+235", are unclear.

"About to fontify" refers to region boundaries used to call
font-lock-fontify-region. The other messages refer to the actual
fontification that was performed.

The code producing the messages is in
https://github.com/yantar92/org/blob/feature/org-font-lock-element/lisp/org-font-lock-core.el
Search for org-font-lock-verbose in the code.

If you want to test it yourself, clone https://github.com/yantar92/org
and use feature/org-font-lock-element branch with the file I provided,
unfold the "Home" headline, and then unfold the "Ideas ..." headline.

To enable messages, you need to set org-font-lock-verbose to non-nil.

-- 
Ihor Radchenko,
Org mode contributor,
Learn more about Org mode at https://orgmode.org/.
Support Org development at https://liberapay.com/org-mode,
or support my work at https://liberapay.com/yantar92