Optimize the url filter by using non-capturing regex groups

[borissmidt]

It seems especially the URL_PATTERN couldn't be optimized by the regex compiler because all groups are capturing.

I was originally looking at the golang ide console speed and there one of the heavy filter plugins is the GoConsoleFilter but i can't really update the code to try the differences.

But i see the complictated regex where not all matching groups are used. So i suspect this gives a similar issue to the URL regex change i do here.

(this is in the GoConsoleFilter)

MESSAGE_PATTERN = Pattern.compile("(?:^|\\s)(?<slog>source=)?(?<filename>\\S+\\.\\w+):(?<line>\\d+)(:(?<col>\\d+))?(?=[:\\s]|$).*");

[trespasserw]

Unfortunately, those are public fields with both internal and external (i.e., by 3rd-party plugins) usages relying on the group numbers, so we can't just change the patterns.

Re GoConsoleFilter -- could you please file an issue at https://youtrack.jetbrains.com, possibly with a CPU profile attached?

[borissmidt]

I've created a ticket:
https://youtrack.jetbrains.com/issue/GO-18089/improve-terminal-speed-for-GoConsoleFilter

Unfortunately, those are public fields with both internal and external (i.e., by 3rd-party plugins) usages relying on the group numbers, so we can't just change the patterns.

As an alternative could we deprecate the existing ones and create a '_V2' version which should be used instead?

Or to limit the scope, these new regexes could be used just in the PatternBasedFileHyperlinkRawDataFinder.