Re: eglot-cthier.el --- show call hierarchy with Eglot (formally called eglot-hierarchy)

[Harald Kirsch <pifpafpuf@gmx.de>]

Hi Filippo.

On 11.12.24 07:05, Filippo Argiolas wrote:
> Harald Kirsch <pifpafpuf@gmx.de> writes:
>
>> This shows a call hierarchy as a foldable tree, unfolded on demand.
>>
>> Through nice feedback here, on codeberg and on @HaraldKi@nrw.social, it
>> got a significant update with bug fixes, enhancements and code improvements.
>>
>> I test this with eglipse-jdts (Java) and tsc (TypeScript).
>>
>> https://codeberg.org/harald/eglot-supplements#call-hierarchy
>>
>> Feedback welcome.
>
> Thanks Harald, love it!
>
> Seems to work with clangd too if you want to add it to the list of
> supported servers. Let me know if you want me to test something
> specific.

A major pain point was buffer/window/frame handling. If you want to get
creative, messing with the window/frames of the hierarchy and the source
code to see what fails would be nice.

> How about an option to unfold the tree by default? I was testing on a
> small project, probably it gets quite bigger on large projects and
> it's better to leave it folded?

I have no recursion-test yet. In any case I would probably add a custom
variable for the max depth to unfold initially. (Classic shift of
responsibility to the user :-) But this should not be difficult.

https://codeberg.org/harald/eglot-supplements/issues/11

> Not sure if it's a clangd thing, I noticed that when I request the call
> hierarchy on anything other than a function call or declaration I get a
> "Error running timer: (args-out-of-range [] 0)". Maybe it could be
> handled more nicely? Not sure it's something to notify the user about,
> guess silently doing nothing would be fine.

Hmmpf, I thought I had this fixed. But server's seem to be creative of
how to provide a null result. I'll check and may contact you again for
details or a test.

https://codeberg.org/harald/eglot-supplements/issues/17
https://codeberg.org/harald/eglot-supplements/issues/15


> Also, when I get to a leaf in the tree it shows with the '+' prefix,
> when expanded it becomes '-' but there is nothing to expand so it just
> changes the prefix. Maybe the leaf nodes should be made non-expandable?

Indeed. But: nodes are only expanded on demand, so on first try, there
is no easy way around. But clearly, once known, the leaf node should be
distinguished.

https://codeberg.org/harald/eglot-supplements/issues/16

Thanks for the feedback
Harald