[ty] IDE: add support for `object.<CURSOR>` completions #18468

BurntSushi · 2025-06-04T18:57:57Z

This PR adds logic for detecting Name Dot [Name] token patterns,
finding the corresponding ExprAttribute, getting the type of the
object and returning the members available on that object.

Here's a video demonstrating this working:

class-attr-completions-work.mp4

Ref astral-sh/ty#86

github-actions · 2025-06-04T19:00:52Z

`mypy_primer` results

No ecosystem changes detected ✅

AlexWaygood · 2025-06-04T19:03:30Z

Ref #86

you mean astral-sh/ty#86 ;)

BurntSushi · 2025-06-04T19:07:44Z

Yup. That's a hard one to get right for me haha.

AlexWaygood

Nice! I haven't reviewed the AST/tokenizing stuff in depth, but this looks reasonable to me!

crates/ty_ide/src/completion.rs

sharkdp · 2025-06-04T19:25:11Z

I am planning to review this tomorrow. One thing that surprised me when testing this locally in the playground: in a scenario like the following, I get no completions at all. I need to type one additional letter before I start to see the completions on C. The snapshot tests seem to indicate that this should work?

class C:
    foo: int
    bar: str

C.<CURSOR>

dhruvmanila · 2025-06-05T05:32:37Z

One thing that surprised me when testing this locally in the playground: in a scenario like the following, I get no completions at all. I need to type one additional letter before I start to see the completions on C. The snapshot tests seem to indicate that this should work?

This is because the server capabilities needs to be updated to include the . (dot) trigger character:

diff --git a/crates/ty_server/src/server.rs b/crates/ty_server/src/server.rs
index 84093d6ffa..4846e0219c 100644
--- a/crates/ty_server/src/server.rs
+++ b/crates/ty_server/src/server.rs
@@ -180,6 +180,7 @@ impl Server {
             completion_provider: experimental
                 .is_some_and(Experimental::is_completions_enabled)
                 .then_some(lsp_types::CompletionOptions {
+                    trigger_characters: Some(vec!['.'.to_string()]),
                     ..Default::default()
                 }),
             ..Default::default()
diff --git a/playground/ty/src/Editor/Editor.tsx b/playground/ty/src/Editor/Editor.tsx
index fca03cd08b..a25e3b4ddb 100644
--- a/playground/ty/src/Editor/Editor.tsx
+++ b/playground/ty/src/Editor/Editor.tsx
@@ -179,7 +179,7 @@ class PlaygroundServer
       monaco.languages.registerDocumentFormattingEditProvider("python", this);
   }
 
-  triggerCharacters: undefined;
+  triggerCharacters: string[] = ["."];
 
   provideCompletionItems(
     model: editor.ITextModel,

This will then inform the client / playground to request completions on . (dot).

For reference, this is Pyright's trigger characters:

triggerCharacters = { ".", "[", '"', "'" }

For context, the client needs to know when it should request completions which by default is usually [a-zA-Z] but the server can expand that set to include additional characters relevant for the server.

dhruvmanila

Thank you! I think this is a good start, I've a few suggestions on what could be done next.

It's a bit unfortunate that we don't yet have E2E test infrastructure for the server, so when testing any server related features that require editor interaction, we usually add a short video demonstrating the functionality in the PR description. I'd recommend to include that whenever you think it'd be useful which I think would be useful in this PR.

crates/ty_ide/src/completion.rs

dhruvmanila · 2025-06-05T05:55:46Z

crates/ty_ide/src/completion.rs

+        static OBJECT_DOT_EMPTY: [TokenKind; 2] = [TokenKind::Name, TokenKind::Dot];
+        static OBJECT_DOT_NON_EMPTY: [TokenKind; 3] =
+            [TokenKind::Name, TokenKind::Dot, TokenKind::Name];


I think this is a great start but it seems to have a few limitations. I don't think these needs to be addressed here but I thought to write it down as a note. For example, doing [].<CURSOR> falls back to providing completions from the scope. Maybe we shouldn't use the fallback behavior when we know that the completion was triggered from a . character? I'm not sure, you might want to play around with this in an editor context.

A follow-up to this would be to expand the token kind that can occur before the . character so that completions on objects other than a variable work like list, dictionary, strings, etc. For example, in [].<CURSOR> we would provide completions for list attributes. We might have to be careful here because we can't just look at the previous token as it's only [].<CURSOR> (a syntactically valid list) should provide completion and not ].<CURSOR>. Also, the list could be arbitrary long which means that there would could be n number of previous tokens to look before we see the [ corresponding to the ] in which case maybe we should rely on the AST instead.

Another example is when an attribute access happens directly on an instantiated class like A().<CURSOR>.

Another one: Anything parenthesized: (a). or { a: True }....

I had a similar thought while playing with this and reviewing the code. There are not that many places in Python's syntax where a . can appear(?). The ellipsis ..., float literals 3.14, and relative imports such as from .mod import x come to mind. This is very hand-wavy, but maybe this suggests that an "opt out" approach could work? That is, instead of explicitly matching on a set of patterns, maybe we should always try to find an enclosing attribute-expression if we see DOT or DOT NAME? This would mean we would also start the enclosing-node search for things like a float literal (3.<CURSOR>), but maybe that's not a problem?

Yeah I like that idea @sharkdp!

(I'll work on this in another PR.)

crates/ty_ide/src/completion.rs

MichaReiser

Nice!

I think we need to set the completion options in the playground and the LSP setup because . isn't a trigger character by default. From the LSP specification

	/**
	 * The additional characters, beyond the defaults provided by the client (typically
	 * [a-zA-Z]), that should automatically trigger a completion request. For example
	 * `.` in JavaScript represents the beginning of an object property or method and is
	 * thus a good candidate for triggering a completion request.
	 *
	 * Most tools trigger a completion request automatically without explicitly
	 * requesting it using a keyboard shortcut (e.g. Ctrl+Space). Typically they
	 * do so when the user starts to type an identifier. For example if the user
	 * types `c` in a JavaScript file code complete will automatically pop up
	 * present `console` besides others as a completion item. Characters that
	 * make up identifiers don't need to be listed here.
	 */
	triggerCharacters?: string[];

MichaReiser · 2025-06-05T07:10:08Z

crates/ty_ide/src/completion.rs

+use crate::find_node::covering_node;

 #[derive(Debug, Clone)]
 pub struct Completion {


Maybe something for another PR but It would be great if we could extend Completion here with a kind that tells VS code what the completion is (is it a variable, a field, etc...). This is currently hard coded in the playground to always be Variable but that will be incorrect with this PR.

Yeah @dhruvmanila has mentioned this a few times, but I've been focusing on breadth (making completions work in more places) over depth (improving completion quality). Maybe I should pause on breadth for the moment and look at depth.

MichaReiser · 2025-06-05T07:13:42Z

crates/ty_ide/src/completion.rs

+        static OBJECT_DOT_EMPTY: [TokenKind; 2] = [TokenKind::Name, TokenKind::Dot];
+        static OBJECT_DOT_NON_EMPTY: [TokenKind; 3] =
+            [TokenKind::Name, TokenKind::Dot, TokenKind::Name];


Another one: Anything parenthesized: (a). or { a: True }....

BurntSushi · 2025-06-05T13:24:37Z

For the trigger characters, nice catch @sharkdp and @dhruvmanila. I didn't notice that because I've been testing this by specifically opting into auto-completions. My editor doesn't show them automatically. I'll look into changing that.

BurntSushi · 2025-06-05T14:57:14Z

Thanks all for the feedback! I've addressed what I think is easy to address in this PR. Otherwise, in follow-up work, I would like to focus on improving the somewhat naive token-based detection here with something more flexible. i.e., To make things like [].<CURSOR> and (a).<CURSOR> work.

crates/ty_ide/src/completion.rs

This PR adds logic for detecting `Name Dot [Name]` token patterns, finding the corresponding `ExprAttribute`, getting the type of the object and returning the members available on that object. Ref #86

BurntSushi · 2025-06-05T15:09:43Z

we usually add a short video demonstrating the functionality in the PR description. I'd recommend to include that whenever you think it'd be useful which I think would be useful in this PR.

Yes, thank you! That's a good idea. I've added a video to the top-level comment here.

In speaking with @MichaReiser, it seems like it would be a good idea for me to be testing this in VS Code as well. So I'll look into setting that up.

dhruvmanila · 2025-06-06T05:52:18Z

Yes, thank you! That's a good idea. I've added a video to the top-level comment here.

Thank you!

In speaking with @MichaReiser, it seems like it would be a good idea for me to be testing this in VS Code as well. So I'll look into setting that up.

Sounds good! The README over at https://github.com/astral-sh/ty-vscode should be good on how to do it. Feel free to provide any feedback if anything was confusing :)

BurntSushi requested review from AlexWaygood, MichaReiser, carljm, dcreager and sharkdp as code owners June 4, 2025 18:57

BurntSushi removed request for carljm and dcreager June 4, 2025 18:58

AlexWaygood added server Related to the LSP server ty Multi-file analysis & type inference labels Jun 4, 2025

AlexWaygood approved these changes Jun 4, 2025

View reviewed changes

crates/ty_ide/src/completion.rs Outdated Show resolved Hide resolved

crates/ty_ide/src/completion.rs Show resolved Hide resolved

dhruvmanila approved these changes Jun 5, 2025

View reviewed changes

MichaReiser approved these changes Jun 5, 2025

View reviewed changes

BurntSushi force-pushed the ag/completion-object branch from c5fd576 to 5addef3 Compare June 5, 2025 14:55

AlexWaygood reviewed Jun 5, 2025

View reviewed changes

crates/ty_ide/src/completion.rs Outdated Show resolved Hide resolved

BurntSushi force-pushed the ag/completion-object branch from 5addef3 to 51608ce Compare June 5, 2025 14:58

AlexWaygood reviewed Jun 5, 2025

View reviewed changes

crates/ty_ide/src/completion.rs Outdated Show resolved Hide resolved

[ty] IDE: add support for object.<CURSOR> completions

7ba780f

This PR adds logic for detecting `Name Dot [Name]` token patterns, finding the corresponding `ExprAttribute`, getting the type of the object and returning the members available on that object. Ref #86

BurntSushi force-pushed the ag/completion-object branch from 51608ce to 7ba780f Compare June 5, 2025 15:07

BurntSushi merged commit 5510020 into main Jun 5, 2025
35 checks passed

BurntSushi deleted the ag/completion-object branch June 5, 2025 15:15

[ty] IDE: add support for object.<CURSOR> completions #18468

[ty] IDE: add support for object.<CURSOR> completions #18468

Uh oh!

Conversation

BurntSushi commented Jun 4, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

github-actions bot commented Jun 4, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

mypy_primer results

Uh oh!

AlexWaygood commented Jun 4, 2025

Uh oh!

BurntSushi commented Jun 4, 2025

Uh oh!

AlexWaygood left a comment

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Uh oh!

sharkdp commented Jun 4, 2025

Uh oh!

dhruvmanila commented Jun 5, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

dhruvmanila left a comment

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Uh oh!

dhruvmanila Jun 5, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

MichaReiser Jun 5, 2025

Choose a reason for hiding this comment

Uh oh!

sharkdp Jun 5, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

BurntSushi Jun 5, 2025

Choose a reason for hiding this comment

Uh oh!

BurntSushi Jun 5, 2025

Choose a reason for hiding this comment

Uh oh!

Uh oh!

MichaReiser left a comment

Choose a reason for hiding this comment

Uh oh!

MichaReiser Jun 5, 2025

Choose a reason for hiding this comment

Uh oh!

BurntSushi Jun 5, 2025

Choose a reason for hiding this comment

Uh oh!

MichaReiser Jun 5, 2025

Choose a reason for hiding this comment

Uh oh!

BurntSushi commented Jun 5, 2025

Uh oh!

BurntSushi commented Jun 5, 2025

Uh oh!

Uh oh!

Uh oh!

BurntSushi commented Jun 5, 2025

Uh oh!

Uh oh!

dhruvmanila commented Jun 6, 2025

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

6 participants

[ty] IDE: add support for `object.<CURSOR>` completions #18468

[ty] IDE: add support for `object.<CURSOR>` completions #18468

BurntSushi commented Jun 4, 2025 •

edited

Loading

github-actions bot commented Jun 4, 2025 •

edited

Loading

`mypy_primer` results

dhruvmanila commented Jun 5, 2025 •

edited

Loading

dhruvmanila Jun 5, 2025 •

edited

Loading

sharkdp Jun 5, 2025 •

edited

Loading