Re: Simplify the tree-sitter parser creation API

[Yuan Fu <casouri@gmail.com>]

> On May 15, 2022, at 7:39 AM, Daniel Martín <mardani29@yahoo.es> wrote:
> 
> 
> Currently, there are three functions to create a tree-sitter parser:
> 
> - treesit-get-parser-create
> - treesit-get-parser
> - treesit-parser-create
> 
> I think this set of similarly named functions can be reduced, so that
> developers don't need to remember a lot of APIs.  How I understand the
> problem of creating a tree-sitter parser is that we need three things
> from the developer:
> 
> - A programming language (a symbol).
> - A buffer (optional, default to the current buffer).
> - Whether to create or reuse an existing parser for that language.
> 
> I think it would be simpler to have just one parser creation API like
> 
> (treesit-make-parser LANGUAGE &optional BUFFER ALWAYS-NEW)
> 
> There are similar API patterns like this one already in Emacs.  With
> this proposed API, the common case, which is to create a parser for the
> current buffer, and to try to reuse one if it already exists, is very
> simple from the POV of an ELisp developer integrating with tree-sitter:
> 
> (treesit-make-parser 'c)
> 
> Any thoughts?
> 

Thanks! I think it makes sense. Though I’d keep the treesit-parser-create function because it directly corresponds to the tree-sitter C API. I can extend the C definition of treesit-parser-create to 

(treesit-parser-create LANGUAGE &optional BUFFER NO-REUSE)

Yuan