Support for snippet based completion #50

amol-mandhane · 2018-07-26T19:50:21Z

LSP servers support snippet based completion which can insert template arguments of the functions while completion.

Example of snippet based completion: https://i.stack.imgur.com/4gKHP.gif

The capability of snippet completion can be registered by setting TextDocumentClientCapabilities.completion.completionItem.snippetSupport to true, which is false by default.

Whether the completion item is plaintext or snippet is given by CompletionItem.kind field.

Yasnippet can be used to manage completion snippets. The support can be conditionally enabled if user has yasnippet installed so that package doesn't have hard dependency on yasnippet.

joaotavora · 2018-07-27T23:28:09Z

Here's a screenshot of the diff.

Wow, such technology!

Now, seriously thanks. I'll try to implement this. Ideally, It should use snippet.el, not yasnippet.el, but I never finished that refactoring...

rgrinberg · 2018-07-30T10:00:48Z

Ideally, It should use snippet.el, not yasnippet.el, but I never finished that refactoring...

Curious, why is that? yasnippet is your package so you obviously know, but I've never even heard of snippet.el

joaotavora · 2018-08-02T10:22:59Z

Curious, why is that?

Nothing special. It's just that yasnippet is way too bloated with stuff that eglot users might not care about, such as menus and snippet-organization logic.

So a few years ago I started writing snippet.el from scratch in that repository. It's the snippet expansion and navigation core of yasnippet rewritten in good elisp. It's workable already but is missing the final stretch that would make it a library that yasnippet could use.

And if it existed, eglot should use that.

yasnippet is your package

Not quite. Though technically I rewrote every line, it's initially @pluskid's package. @npostavs is the new maintainer now.

* eglot.el (eglot-client-capabilities): Declare support for snippet-based completions. (eglot-completion-at-point): Expand snippet completions with YASnippet if that is found.

joaotavora · 2018-08-17T23:31:23Z

@amol-mandhane, can you test the newest version?

garyo · 2019-10-29T22:08:18Z

I'm using vue mode with vue-language-server. I can't seem to trigger completionItem/resolve with eglot (I'm using the latest github head from yesterday, Oct 28, 2019). This issue says it should work, but I don't know how to make it happen. When I try the same server with lsp-mode, the server returns snippets in completionItem/resolve when you complete a completion, like this:

	[Trace - 05:33:09 PM] Sending request 'completionItem/resolve - (47)'.
	Params: {
	  "data": {
	    "entryName": "entries",
	    "offset": 56,
	    "uri": "file:///c:/tmp/vls-lsp-problem/test.ts"
	  },
	  "sortText": "0",
	  "kind": 3,
	  "label": "entries"
	}
	
	
	[Trace - 05:33:09 PM] Received response 'completionItem/resolve - (47)' in 20ms.
	Result: {
	  "insertTextFormat": 2,
	  "insertText": "entries(${1:obj}, ${2:key})",
	  "detail": "function entries<T>(obj: {\n    [key: string]: T;\n} | ArrayLike<T>): {\n    key: string;\n    value: T;\n}[] (+1 overload)",
	  "documentation": "Returns an array containing the property keys and values of the specified object (an associative array).\nEach entry is an object with a key and value attribute.The order of the returned array is undefined.\n\nThe generic refers to the data type of the values.",
	  "sortText": "0",
	  "kind": 3,
	  "label": "entries"
	}

It has an insertText snippet with format 2, which is right, and then yasnippet lets me fill in the args. Does eglot send completionItem/resolve? How do I trigger it to do that?

joaotavora · 2019-10-29T22:51:12Z

GaryO <[email protected]> writes:

It has an insertText snippet with format 2, which is right, and then yasnippet lets me fill in the args. Does eglot send completionItem/resolve?

It does, but not in this particular context. It is something that needs to be fixed in Eglot. I'll have a look, shouldn't be terribly hard. Do you have the rest of the log? Actually only the immediately preceding textDocument/completion request and response, if you can. Thanks, João

It was already used to resolve documentation bits of completions, but it can also be useful to resolve snippet templates and such. To resolve a completion, you need some part of a completion to start with. If it has a :data field exists and the server supports :resolveProvider, fetch the new object, otherwise use whatever we had already. * eglot.el (eglot-completion-at-point): Add another local function for resolving completions.

joaotavora · 2019-10-30T01:27:11Z

@garyo , this is a totally "blind" since I very much want to avoid installing that Vue thing as much as I can :-).

So I created a new branch and did a commit there which you should test. Maybe it fixes your issue. If it doesn't we'll iterate until it does.

garyo · 2019-10-30T16:14:25Z

Progress -- I now get snippets in a pure typescript buffer. Per your other message, I already had yasnippet and yas-global-mode set up, and I can confirm eglot is telling the server it supports snippets. Also, the log above is actually from a typescript (jsts-ls) server (my mistake, sorry), and with your patch eglot is working fine with that server now.

I guess vetur/vls doesn't support function completion snippets though. I don't have them working with lsp-mode either. I'm not getting any insertText in the completionItem/resolve reply for a function completion. Hmm, I thought it did -- it has its own global snippets, which are working, but who knows.

client-request (id:75) Wed Oct 30 11:54:38 2019:
(:jsonrpc "2.0" :id 75 :method "textDocument/completion" :params
          (:textDocument
           (:uri "file:///c:/tmp/vls-lsp-problem/test.vue")
           :position
           (:line 12 :character 11)
           :context
           (:triggerKind 1)))

server-reply (id:75) Wed Oct 30 11:54:39 2019:
(:jsonrpc "2.0" :id 75 :result
          (:isIncomplete :json-false :items
                         [(:uri "file:///c:/tmp/vls-lsp-problem/test.vue" :position
                                (:line 12 :character 11)
                                :label "arguments" :sortText "00" :kind 6 :data
                                (:languageId "typescript" :uri "file:///c:/tmp/vls-lsp-problem/test.vue" :offset 194))
                          (:uri "file:///c:/tmp/vls-lsp-problem/test.vue" :position
                                (:line 12 :character 11)
                                :label "fooFunction" :sortText "01" :kind 3 :data
                                (:languageId "typescript" :uri "file:///c:/tmp/vls-lsp-problem/test.vue" :offset 194))
              ...(lots more here)...

client-request (id:82) Wed Oct 30 11:54:39 2019:
(:jsonrpc "2.0" :id 82 :method "completionItem/resolve" :params
          (:uri "file:///c:/tmp/vls-lsp-problem/test.vue" :position
                (:line 12 :character 11)
                :label
                #("fooFunction" 0 1
                  (eglot--lsp-item #1))
                :sortText "01" :kind 3 :data
                (:languageId "typescript" :uri "file:///c:/tmp/vls-lsp-problem/test.vue" :offset 194)))

server-reply (id:82) Wed Oct 30 11:54:39 2019:
(:jsonrpc "2.0" :id 82 :result
          (:uri "file:///c:/tmp/vls-lsp-problem/test.vue" :position
                (:line 12 :character 11)
                :label "fooFunction" :sortText "01" :kind 3 :detail "function fooFunction(alpha: string, beta: number): string" :documentation
                (:kind "markdown" :value "")))

And, I just tested vls in VS Code: same thing. The "special" vue snippets work there (as they do now with your patch in eglot), but no snippets on function completion. So I think your patch in your scratch branch is fine!

joaotavora · 2019-10-30T16:47:56Z

Also, the log above is actually from a typescript (jsts-ls) server (my mistake, sorry), and with your patch eglot is working fine with that server now.

So if I understand correctly the semi-blind untested completionItem/resolve thing is working.

And, I just tested vls in VS Code: same thing. The "special" vue snippets work there (as they do now with your patch in eglot), but no snippets on function completion. So I think your patch in your scratch branch is fine!

Cool!

garyo · 2019-10-30T17:07:36Z

So if I understand correctly the semi-blind untested completionItem/resolve thing is working.

Yes, it's working for me!

It was already used to resolve documentation bits of completions, but it can also be useful to resolve snippet templates and such. To resolve a completion, you need some part of a completion to start with. If it has a :data field exists and the server supports :resolveProvider, fetch the new object, otherwise use whatever we had already. * eglot.el (eglot-completion-at-point): Add another local function for resolving completions.

* eglot.el (eglot-client-capabilities): Declare support for snippet-based completions. (eglot-completion-at-point): Expand snippet completions with YASnippet if that is found. (eglot-note, eglot-warning, eglot-error): Diagnostic overlay priorities have to be slightly lower than yasnippet's, which must be reasonably high.

It was already used to resolve documentation bits of completions, but it can also be useful to resolve snippet templates and such. To resolve a completion, you need some part of a completion to start with. If it has a :data field exists and the server supports :resolveProvider, fetch the new object, otherwise use whatever we had already. * eglot.el (eglot-completion-at-point): Add another local function for resolving completions.

* eglot.el (eglot-client-capabilities): Declare support for snippet-based completions. (eglot-completion-at-point): Expand snippet completions with YASnippet if that is found. (eglot-note, eglot-warning, eglot-error): Diagnostic overlay priorities have to be slightly lower than yasnippet's, which must be reasonably high.

It was already used to resolve documentation bits of completions, but it can also be useful to resolve snippet templates and such. To resolve a completion, you need some part of a completion to start with. If it has a :data field exists and the server supports :resolveProvider, fetch the new object, otherwise use whatever we had already. * eglot.el (eglot-completion-at-point): Add another local function for resolving completions.

* eglot.el (eglot-client-capabilities): Declare support for snippet-based completions. (eglot-completion-at-point): Expand snippet completions with YASnippet if that is found. (eglot-note, eglot-warning, eglot-error): Diagnostic overlay priorities have to be slightly lower than yasnippet's, which must be reasonably high. #50: joaotavora/eglot#50

It was already used to resolve documentation bits of completions, but it can also be useful to resolve snippet templates and such. To resolve a completion, you need some part of a completion to start with. If it has a :data field exists and the server supports :resolveProvider, fetch the new object, otherwise use whatever we had already. * eglot.el (eglot-completion-at-point): Add another local function for resolving completions. #50: joaotavora/eglot#50

* eglot.el (eglot-client-capabilities): Declare support for snippet-based completions. (eglot-completion-at-point): Expand snippet completions with YASnippet if that is found. (eglot-note, eglot-warning, eglot-error): Diagnostic overlay priorities have to be slightly lower than yasnippet's, which must be reasonably high. GitHub-reference: close joaotavora/eglot#50

It was already used to resolve documentation bits of completions, but it can also be useful to resolve snippet templates and such. To resolve a completion, you need some part of a completion to start with. If it has a :data field exists and the server supports :resolveProvider, fetch the new object, otherwise use whatever we had already. * eglot.el (eglot-completion-at-point): Add another local function for resolving completions. GitHub-reference: per joaotavora/eglot#50

joaotavora added a commit that referenced this issue Aug 10, 2018

Close #50: Support snippet completions

2b1db17

* eglot.el (eglot-client-capabilities): Declare support for snippet-based completions. (eglot-completion-at-point): Expand snippet completions with YASnippet if that is found.

joaotavora closed this as completed in 3ffea45 Aug 17, 2018

garyo mentioned this issue Oct 30, 2019

Completion with Vue vls returns no completions #319

Closed

joaotavora mentioned this issue Feb 16, 2022

Eglot completion with snippet got extra chars #836

Closed

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Support for snippet based completion #50

Support for snippet based completion #50

amol-mandhane commented Jul 26, 2018

joaotavora commented Jul 27, 2018

rgrinberg commented Jul 30, 2018

joaotavora commented Aug 2, 2018

joaotavora commented Aug 17, 2018

garyo commented Oct 29, 2019

joaotavora commented Oct 29, 2019 via email

joaotavora commented Oct 30, 2019

garyo commented Oct 30, 2019 •

edited

Loading

joaotavora commented Oct 30, 2019

garyo commented Oct 30, 2019

Support for snippet based completion #50

Support for snippet based completion #50

Comments

amol-mandhane commented Jul 26, 2018

joaotavora commented Jul 27, 2018

rgrinberg commented Jul 30, 2018

joaotavora commented Aug 2, 2018

joaotavora commented Aug 17, 2018

garyo commented Oct 29, 2019

joaotavora commented Oct 29, 2019 via email

joaotavora commented Oct 30, 2019

garyo commented Oct 30, 2019 • edited Loading

joaotavora commented Oct 30, 2019

garyo commented Oct 30, 2019

garyo commented Oct 30, 2019 •

edited

Loading