20:58, August 12, 2022, Eli Zaretskii <eliz@gnu.org>:
From: Dmitry Gutov <dgutov@yandex.ru>
Cc: Eli Zaretskii <eliz@gnu.org>,
"56682@debbugs.gnu.org" <56682@debbugs.gnu.org>,
"monnier@iro.umontreal.ca" <monnier@iro.umontreal.ca>
Date: Fri, 12 Aug 2022 19:35:57 +0300
I have eliminated two performance problems frim js-json-mode that are uncharacteristic for font-lock in
general. One used a pathological matcher (simply removed), and another uses the "frameworks" facility
which simply always rescans the full buffer from the beginning. That one can/should be fixed later, but I have
disabled it for JSON (where it doesn't make sense anyway).
You should know that the rest of js-mode's font-lock keywords are still there. I experimented with leaving just
the bare minimum needed for JSON, but the effect was very minimal (like 20% improvement).
I didn't yet have time to try the branch, but what you say begs a
question: does the branch attempt to solve the problems of JSON files
only, or does it attempt to solve a more general problem?
There is nothing json-specific in the branch.
All the js/json stuff is on master already.