Incorrect detection of code blocks. about gnome-shell-pano HOT 3 CLOSED

This is really hard to detect. I use highlight.js to determine the language but it is not always accurate. Also, I experimented with guesslang models that vscode uses to detect the code. That seems a little bit better but extension bundle size skyrockets(because of TensorFlow dependencies). I am open to ideas about other solutions.

from gnome-shell-pano.

Highlightjs seems to have no state for plain text input. Imo, doing some basic analysis before passing it would be a better solution for now.

Im thinking of having different conditions with weights.

Some sample conditions with weights I can think of:
Line starts with // - 1
Line starts with # - 0.75
Has words seperated by only dots - 0.5
Has // somewhere in a line - 0.2
Has more than one line with only { or } - 1
Has a ( following a word without space- 0.25
Has () following a word - 1
Contains some number of predifined keywords like if while for - 1
Has more than one PascalCase snake_case or camelCase words - 1
Has some math symbols - 1
Has multiple < and > with a word in between - 1

This is just a rough idea and having some library handle this in a better way would be the best solution.

Found this while writing: https://softwareengineering.stackexchange.com/a/87622
Those are some better conditions.

from gnome-shell-pano.

I am considering removing code blocks.

from gnome-shell-pano.