Trigger character logic

[erooke]

Is the logic for deciding what the triggering character was document somewhere? For a plugin I wrote it would be nice if the string \^2 considered \ as its triggering character however it seems when ^ is entered it takes over as the new triggering character.

I was able to bodge around this manually by registering both \ and ^ as triggers and manually changing offsets when the trigger was ^ (erooke/blink-cmp-latex@61927e9)

I poked around trying to find the logic for keyword/trigger detection but nothing immediately popped up.

[Saghen]

Yeah whenever we see a trigger character, we drop the existing context and create a new one, so there's no knowledge of the previous trigger character. This is by design to follow the LSP spec. I think for your case, having language specific keyword regexes (such as one that includes \ and ^) could work around this: #1599 (comment)

[erooke]

Language specific keyword regexes would be a great step. Honestly the completions I'm doing may not be the best fit for blink.

The completion is completing unicode characters by their latex macro counter part. This results in some very symbol heavy completions (\1/2 and \mathbb{A}`) for example, I would probably end up needing source specific keyword regexes as the goal is to have this source work regardless of the language.