Performance issue when looking up a cosmetic filter's origin in the logger

[gorhill]

Issue first raised in #2664 (comment):

2c9c98e introduced regression in logger.
On my system (Firefox ~85k cosmetic filters) up to 7 seconds is needed for cosmetic filter details layer/dialog to appear after I click on it.

In Chromium, this is about 2s.

[gorhill]

@gwarser,

I transcribed what I see in your screenshot at #2664 (comment):

Regex of filter to lookup: /[^\n]*\\*"\.googad\\*"[^\n]*/.

Content of compiled EasyList which is scanned with above regex:

compiled-easylist-cosmetic-filters.txt

When I paste/try the regex and content at https://regex101.com/, after selecting "javascript" in the left hand sidebar, I get a timing of 30-40 ms without the g flag, and 250-300 ms with the g flag (your screenshot shows over 500 ms -- edit: ok, it's 500+ ms for all filter lists, here I am testing only with EasyList). The (only) matching line is about 1/3 down from the beginning of the scanned content.

The result with the g flag is curious, given that there is only one match for that regex. Can you tell me what kind of timing you get on your side with this test?

[gorhill]

Interesting... If I add ^ and $ at the start and beginning of the regex and use the m flag, the timing decreases dramatically when using the g flag, 15-30 ms:

/^[^\n]*\\*"\.googad\\*"[^\n]*$/gm

[gwarser]

64ms without global /[^\n]*\\*"\.googad\\*"[^\n]*/
517ms with global /[^\n]*\\*"\.googad\\*"[^\n]*/g

36ms with ^...$ and mg /^[^\n]*\\*"\.googad\\*"[^\n]*$/gm