bug#62116: RFE: eglot: support window.showDocument LSP RPC

[Felician Nemeth]

> Felicián, can you show your full patch again, perhaps with a unit test
> using the ada language server?

I've attached the patch.

However, I failed to write a unit test.  At the end of
eglot-handle-request buffer-name shows the expected main.adb, but at the
end of the my ert-deftest it is the old, incorrect value.  This puzzles
my, because if I interactively repeat the test steps, everything is OK.
I copy the test code at the end of the email.

> Also, can you comment on why you think Alan Donovan's patch has that
> non-synchronous find-file mechanism?  Since Alan, doesn't reply, what do
> you conjecture is the reason his patch goes through this trouble?
>
> I'd also like someone to address my comments of that patch, which is
> similar to yours:
> https://debbugs.gnu.org/cgi/bugreport.cgi?bug=62116#10.  Among others
>
> 1. if you think the Emacs frame should be raised and/or focused when a
>    window.showDocument comes in.

I always use Emacs in single frame, so I have no experience in this
area, but I don't understand why x-focus-frame is necessary.  The
showDocument request should not come out of the blue, and if the user is
actively working on a source file, the frame is already in focus.

> 2. if browse-url should be used for non-file: urls (it might open a
>    browser window outside of Emacs, which IMO is fine.)

My original patch used browse-url as well.  browse-url is highly
customizable, and I think it is right thing to use here.  Users can
customize it to relay the URL to eww, firefox, etc even for specific
URLs.

If I understand it correctly, you argued that if takeFocus is false,
Eglot should still show the requested file, but the point should remain
were it was.  I think you're right, but I haven't updated my patch yet.

Alan's patch also contains a recenter call, which I think is unnecessary
and would probably annoy me, when the target point was already visible
before the showDocument request.

On the other hand, if the showDocument request contains a selection,
both patches could be enhanced to better visualize this selection.

Because of the timers, Alan's eglot-handle-request returns success even
if the inner lambda fails.  It should probably check if filename exists
before calling run-with-idle-timer.

However, I don't like approach of using timers.  I think a better
alterative is to put a modified lambda function in an event queue, and
Eglot should process the queue after it sent its reply to the client
request.

Finally, I think Alan's copyright status is still unclear.

And this is the my failed attempt at writing a test:

(ert-deftest eglot-test-window/showDocument ()
  "Test handling a window/showDocument server request."
  (skip-unless (executable-find "ada_language_server"))
  (let (server)
    (eglot--with-fixture
     '(("project" .
        (("project.gpr" .
          "Project P is\n   for Main use (\"main.adb\");\nend P;")
         ("main.ads" . "procedure Main;")
         ("main.adb" . "procedure Main is null;"))))
     (with-current-buffer
         (eglot--find-file-noselect "project/main.ads")
       (eglot--sniffing (:server-requests s-requests
                         :server-replies s-replies
                         :client-requests c-requests
                         :client-replies c-replies)
         (should (setq server (eglot--tests-connect)))
         (eglot-execute-command server "als-other-file"
                                (vector (eglot--TextDocumentIdentifier)))
         (let (cmd-id show-doc-id)
           (eglot--wait-for (c-requests 2)
               (&key id method &allow-other-keys)
               (setq cmd-id id)
               (string= method "workspace/executeCommand"))
           (eglot--wait-for (s-requests 2)
               (&key id method &allow-other-keys)
             (setq show-doc-id id)
             (string= method "window/showDocument"))
           (eglot--wait-for (c-replies 1)
               (&key id error &allow-other-keys)
               (and (eq id show-doc-id) (null error)))
           (eglot--wait-for (s-replies 1)
               (&key id &allow-other-keys)
               (eq id cmd-id))))
         (should (string-equal (buffer-name) "main.adb"))))))

0001-Eglot-support-window-showDocument-request.patch
Description: Text Data