Incremential parsing (or more generally - Optimization)

[InAnYan]

I'm not sure if we will ever run into this issue. Premature optimization is evil.

But just think: how diagnostics (and other stuff) is generated?

User makes edit -> vscode sends to lsp the source code -> jabls reparses bib -> generates diagnostics a new

How often edits are sent and file is reparsed and analyses are rerunned?

Rust LSP starts analyses after save. More complex LSPs (like for Python) able to analyze data in real time.

But we'll see in future

[koppor]

When tackling this, we change the whole BibTeX loading and saving to the idea of json-tree -- have the BibTeX as String in memory and have indexes on the String. On any update, modify the string and the indexes.