More options for completion filtering

[ddovod]

Here's the first try to add more options for completion filtering

[ddovod]

For mecaseInsensitivePrefixMatcher works better than others so far. Will try another approaches

[jacobdufault]

Happy to merge when implementation is finalized

[jacobdufault]

Can you add some timing to see if the virtual function call is negatively impacting how long it takes to filter completion results?

[ddovod]

I'll check it a bit later, no problem. I'm sure that virtual function call will not spend much more time than plain function call, especially comparing to all that fuzzy-matching stuff = )

[jacobdufault]

Make sure to document all possible options here

[ddovod]

Done

[jacobdufault]

Strange, but this should have a virtual dtor for correct behavior (also, the = default should probably go in a .cc file)

[bstaletic]

The = default part can behave differently if moved to the .cc file.
https://accu.org/index.php/journals/2379

[ddovod]

No idea whats happened but now it works fine. Probably some weird stuff on my side

[jacobdufault]

There may be a function in utils.h to do this

[ddovod]

Done. And also moved StartsWithIgnoreCase to utils

[ddovod]

Hi guys. Sorry, was too busy all the week

[ddovod]

Also which version of clang-format do you use to format code? Mine is 6.0.0 and I'm getting some diff on untouched files when formatting all src directory

[ddovod]

I'm experiencing some weird stuff. Basically I've found out that cquery fuzzy matching algorithm works good enough. The thing is completion results are reordered, probably without respecting score_ parameter.
This is what I see with cquery matcher:

But if I add this piece of code

  int i = 0;
  for (auto& el : items) {
      el.label = std::to_string(++i) + " : " + el.label;
  }

right before finalize() is called in FilterAndSortCompletionResponse, I see this:

Looks like somebody sorts candidates after they was sorted once. Could you please reproduce it on your side?

[jacobdufault]

Looks like somebody sorts candidates after they was sorted once. Could you please reproduce it on your side?

If cquery is not explicitly sorting then that is almost certainly your completion client not honoring the priority.

[ddovod]

Damn, there're soo many packages between me and code I'm trying to edit) maybe you can help me a bit? I just need to understand who is sorting candidates (lsp-mode, emacs-cquery, company-lsp). Maybe you can point me to some elisp code which receives completion candidates from cquery and gives it to some lsp-mode functions?

[jacobdufault]

Sorry, I'm not very familiar with how emacs LSP works.

[ddovod]

Okay, no problem, I'll try to dig into it