Hongyi Zhao <[email protected]> writes: > The most significant advantage I've found is the on-the-fly > diagnostics. The server flags issues like undefined labels, incorrect > command usage, or missing packages as you type, often before you even > think about compiling. This shortens the feedback cycle dramatically > and catches small mistakes instantly.
Thanks. In my workflow, these were non-issues: I added and referenced labels/citations exclusively with RefTeX, so there were no undefined labels, commands were provided by AUCTeX-completion, and I was happy that the editor didn't add any new packages to my file behind my back. > Modern LSPs also add seamless "go to definition" for labels/citations > and richer context-aware autocompletion (e.g., showing abstracts for > \cite candidates), which feels very much like a modern IDE. Check `reftex-view-crossref'. > Many people now seem to be adopting a hybrid approach: > > 1. Using LSP for the live error-checking and smart navigation. > 2. Still relying on AUCTeX for its superior compilation process (C-c > C-c) and the amazing preview-latex feature, which LSP can't replace. > > It's less of a replacement and more of a powerful enhancement. Just > thought I'd share my perspective on how things have evolved recently. Yes, this seems to the reality now. Don't get me wrong, I'm not arguing against a LSP-server, I think it just didn't fit the way I wrote LaTeX files. Personally, I think the killer feature of AUCTeX (or Emacs in general) is bound to M-q. Best, Arash
