[READY] Add support for didChangeWorkspaceFolders

[bstaletic]

Perhaps it's finally time to add this capability...

There are a couple missing pieces:

1. So far, this does not take into account the extra confs.
2. Clangd completer does not have GetProjectRootFiles() implemented, but also clangd does not want to listen to this notification. I guess it knows better than us?
3. Java completer also does not have GetProjectRootFiles() implemented, but actually wants this notification.
4. Do we need _project_directory and _server_workspace_dirs?
5. Tests... after we figure out the answers to the above!

---

This change is 

[codecov]

Codecov Report

Merging #1731 (3de31d7) into master (f11e293) will increase coverage by 0.09%.
The diff coverage is 100.00%.

Additional details and impacted files

@@            Coverage Diff             @@
##           master    #1731      +/-   ##
==========================================
+ Coverage   95.43%   95.52%   +0.09%     
==========================================
  Files          83       83              
  Lines        8168     8185      +17     
  Branches      163      163              
==========================================
+ Hits         7795     7819      +24     
+ Misses        323      316       -7     
  Partials       50       50              

[bstaletic]

This pull request actually allows rust-analyzer to report diagnostics for two projects opened in the same vim instance. For example:

1. ycmd/tests/rust/testdata/common
2. ycmd/tests/rust/testdata/macro

[bstaletic]

Rebased.

This pull request is behaving correctly for rust, go and java. C++ server does not support this notification and we respect that.

Other than that, I have tried it with haskell and it did what I expected, but haskell server does not really do anything with the notification.

[puremourning]

Looks good!

Answers

1. extra cones... I think you added this?
2. clangd does its own copile_commands.json scanning and does a good job of associating files with dbs, let's leave them to it!
3. java completer is special, and I can see that you've handled it like the special sausage that it is
4. I think we keep both. The project root is still special in that it's the rootUri in the init and it's the CWD of the server
5. I would like to see some tests for the other engines if we can. The java one was interesting that it shows the server is bonkers and seems to duplicate the messages :(

Reviewable status: 0 of 2 LGTMs obtained (waiting on @bstaletic)

---

ycmd/completers/java/java_completer.py line 421 at r1 (raw file):

  def _OpenProject( self, request_data, args ):

I wonder if we can just ditch this now? I doubt anyone uses it. I added it for a specific purpose to work around this limitation, but TBH YcmCompleter RestartServer works just as well, and this is just a bit of a hack on top of that.

I vote remove. WDYT?

[bstaletic]

Right, I should have updated the opening comment.

extra cones... I think you added this?

Extra confs can report statically known workspaces. I am unsure if we will need something like extra_conf.GetWorkspaceForFile(file). I am happy to defer that to a point after we get a lot more field experience and thus know whether we need that or not.

clangd

Agreed.

java [snip] special sausage

Loved that!

I would like to see some tests for the other engines if we can.

Originally, I tried to write a test for rust. But it tends to publish one set of diagnostics, change its mind, then publish another set of diagnostics. But not always. Our tests for the diagnostic poll don't deal well with that, so I gave up on testing with rust.

I can try with go...

Reviewable status: 0 of 2 LGTMs obtained (waiting on @bstaletic and @puremourning)

---

ycmd/completers/java/java_completer.py line 421 at r1 (raw file):

Previously, puremourning (Ben Jackson) wrote…

I wonder if we can just ditch this now? I doubt anyone uses it. I added it for a specific purpose to work around this limitation, but TBH YcmCompleter RestartServer works just as well, and this is just a bit of a hack on top of that.

I vote remove. WDYT?

I initially stayed away from that because it is technically a breaking change. I can remove it, sure.

[bstaletic]

Ha! Gopls works great!

Reviewable status: 0 of 2 LGTMs obtained (waiting on @bstaletic and @puremourning)

---

ycmd/completers/java/java_completer.py line 421 at r1 (raw file):

Previously, bstaletic (Boris Staletic) wrote…

I initially stayed away from that because it is technically a breaking change. I can remove it, sure.

Removed.

[bstaletic]

Okay, rust test is there too, although it needed a bit of secret jank sauce.

Reviewable status: 0 of 2 LGTMs obtained (waiting on @puremourning)

[bstaletic]

Wiped off the jank sauce. 100% diff coverage hit.

Reviewable status: 0 of 2 LGTMs obtained (waiting on @puremourning)

[puremourning]

Reviewed 9 of 12 files at r1, 10 of 10 files at r2, all commit messages.
Reviewable status: 1 of 2 LGTMs obtained (waiting on @bstaletic)

[mergify]

Thanks for sending a PR!