Tree-sitter support for Emacs 29+ #282
Conversation
er-basic-expansions.el (er/mark-ts-node): New function for expanding the region using Tree-sitter parse tree nodes when Tree-sitter is available and can parse the buffer around point. The new function is safe to add to the global base value of `er/try-expand-list` since it leaves the region unchanged when Tree-sitter information is not available.
|
Thanks this looks pretty great! Would you say it is ready for use? |
|
That was a swift response of yours indeed! I have been using it for two weeks without any issues now. In that light, and given the low complexity of the change: yes, ready for use. |
|
Excellent, thank you! I'm sorry that I have to ask this, but what is your status regarding the Emacs copyright paperwork? |
No worries, that is a very valid question. I have signed the FSF copyright assignment for contributions to Emacs. I am thus fine with your licence, and having my code on ELPA. |
|
Perfect, thank you! 🙏 |
|
👍 Thank you for taking my proposal on board, and so fast; glad you found it helpful. |
I presume this tries to achieve very similar goals as PR #279, but uses a different approach. Instead of iterating some list of languages, it adds Tree-sitter region expansion to the global default value of
er/try-expand-list. The new expansion function checks whether Tree-sitter support is available in this instance of Emacs, and for the stuff around point, and stays out of the way when it isn't. Since Tree-sitter support in Emacs 29+ requires the user to make some extra configuration steps, such as installing external language grammars, selecting which<language>-ts-modes to use etc., it is impossible to predict for which languages Tree-sitter support is available for a particular user, or not. Hence the generic approach of adding it to the basic expansion try list, and allowing it to kick in when a Tree-sitter parse tree is available only. Whenever the user will add a new Tree-sitter grammar to his/her configuration in the future, expand-region will instantly start working in that language, too.Warning re. tests: The new expansion function is added to the end of
er/try-expand-list, but may still break existing tests. To devise new tests for the new function, the approach would need to be something along the lines of: pick a language for which there is a built-in<language>-ts-mode, but which is not supported by expand-region, and for which the person running the tests has a Tree-sitter language grammar installed and activated Tree-sitter support in Emacs. Since it seems close to impossible to reliably guess the state of Tree-sitter language support in everybody's copy of Emacs, It does not seem realistically feasible to come up with a test case which will work on everyone's Emacs 29+.