On Friday, April 28th, 2023 at 06:05, Dmitry Gutov <dmitry@gutov.dev> wrote:
> On 28/04/2023 04:35, Randy Taylor wrote:
> 
> > Maybe I'm missing something, but doesn't treesit-query-string with an empty string do nothing and always error, since it queries on whatever string you pass in?
> 
> 
> It would return nil, not error (querying an empty buffer). And we should
> be checking for error.
> 
> This is the part that java-ts-mode--string-highlight-helper got right, BTW.

Sorry, I was operating under false assumptions and somehow convinced myself that treesit-query-string with an empty string returning nil meant the query wasn't supported, hence me thinking it achieved nothing. I am no longer under such delusions :).

> 
> go-ts-mode--iota-query-supported-p in your patch relies on the condition
> that the file, at the time the mode is enabled, already contains at
> least one "iota" node.

Good point. I've attached a patch using treesit-query-string with an empty string. It now works properly when iota is missing but inserted later.

> 
> > In that case, wouldn't using treesit-query-capture directly be the best? I don't understand why it's not from reading that thread. If it isn't, then treesit-query-validate is the only option, right?
> 
> 
> treesit-query-validate pops a new buffer interactively in the case of
> failure, so it's not something we should use here.