Schema based completions for nodes in paths #595

anderson4j · 2025-10-28T13:43:36Z

Implements similar logic that we used for paths like
MATCH (n:A)-[: to complete rels, to also work for node completions in paths like:
MATCH (n:A)-[r:R]->(:

Also uses the new lastRule property on the parsing result to get the latest node/rel pattern before the caret in both rel/node completions to avoid taking just "last pattern in parent ctx" failing on queries like MATCH (n:A)-[r:R]->(:^)-[r2:R2]->(:

Closes DEV-197

changeset-bot · 2025-10-28T13:43:40Z

⚠️ No Changeset found

Latest commit: e18ee73

Merging this PR will not cause a version bump for any packages. If these changes should not result in a new version, you're good to go. If these changes should result in a version bump, you need to add a changeset.

This PR includes no changesets

When changesets are added to this PR, you'll see the packages that this PR includes changesets for and the associated semver types

Click here to learn what changesets are, and how to add one.

Click here if you're a maintainer who wants to add a changeset to this PR

anderson4j · 2025-10-28T13:52:40Z

packages/language-support/src/autocompletion/schemaBasedCompletions.ts

+//  and returns the "labels" of rels/nodes going out/in of the node/rel in the graph schema
 function walkLabelTree(
-  relsFromLabels: Map<string, Set<string>>,
+  labelToConnectedLabelsMap: Map<string, Set<string>>,


We can use this identically if it is
Map<label, Set> (rel completion)
Map<rel types, Set>, (node completion)

I struggle to find a nice name that covers both. Maybe I should add the above comment into the code to help clarify things that way?

In the grammar/database, they're just called "labels" right? Even if they can also be relationship types. It's not perfect, but at least somewhat consistent.

An idea for a better name for the labelToConnectedLabelsMap be to call it labelSegments ? I feel it makes it clear it's part of a path/patterns but not the full thing?

Just a thought: is there a risk anywhere in our logic around people using the same names for a Reltype or label?

I have some ideas on how we maybe could make this data structure better or more efficient, but I think we can wait until the feature matures to consider that refactoring

About interference between rels/nodes with the same labels. Short answer - I see no risk

Longer:
We check for variable name/reference offset (really offset should be enough..) when fetching labels for a variable, so other variables will not interfere here -> the labels object on each symbol should be correct regardless of labels of other variables, even if the labels are the same.

Regarding the naming. Im not sure of labelSegments... First thought is that seeing a collection called labelSegments I would think "list of segments", not map to list... but then if we use it like labelSegments.get(myLabel) I suppose that makes it clear that it's a map... I checked with ChatGPT and it proposed connectedLabelsMap, which I quite like. Or maybe even just connectedLabels

I'm happy to go with connectedLabels 👍

… into nodeSACompletions

anderson4j · 2025-10-29T08:21:15Z

packages/language-support/src/tests/autocompletion/schemaBasedPatternCompletion.test.ts

        { label: 'CATCHES', kind: CompletionItemKind.TypeParameter },
        { label: 'TRAINS', kind: CompletionItemKind.TypeParameter },
+        { label: 'IS_IN', kind: CompletionItemKind.TypeParameter },
+        //Limitation: Only take preceding node into account


If we keep slicing the query at the cursor we won't even have access to the fact that a "Pokemon" node follows the rel we are completing

OskarDamkjaer · 2025-11-04T13:16:57Z

packages/language-support/src/tests/autocompletion/schemaBasedPatternCompletion.test.ts

+  test('Simple node completion pattern', () => {
+    const query = 'MATCH (t:Trainer)-[r:CATCHES]->(:';


It'd be nice to add a few more test cases (longer pattern, different relationships). Kind of like we have for the rel types?

And for example showing tests of the limitations with regards to directionality

I think directionality is covered by existing tests (like this one, you see the "limitation" comment inside), so im not sure if we need more tests that explicitly state the limitation?

OskarDamkjaer · 2025-11-04T13:17:30Z

packages/language-support/src/tests/autocompletion/completionAssertionHelpers.ts

+  if (query.length > offset) {
+    query = query.slice(0, offset);
+  }


is this not done in the auto completion function now?

Oh right, I fixed it here first when I thought id merge it in quickly, but then things came up and I did a separate PR with it.

Forgot to merge main to this, but you're right I should now remove this, good spot!

OskarDamkjaer · 2025-11-04T13:23:37Z

packages/language-support/src/autocompletion/schemaBasedCompletions.ts

+//  and returns the "labels" of rels/nodes going out/in of the node/rel in the graph schema
 function walkLabelTree(
-  relsFromLabels: Map<string, Set<string>>,
+  labelToConnectedLabelsMap: Map<string, Set<string>>,


In the grammar/database, they're just called "labels" right? Even if they can also be relationship types. It's not perfect, but at least somewhat consistent.

An idea for a better name for the labelToConnectedLabelsMap be to call it labelSegments ? I feel it makes it clear it's part of a path/patterns but not the full thing?

OskarDamkjaer · 2025-11-04T13:24:52Z

packages/language-support/src/autocompletion/schemaBasedCompletions.ts

+      }
+
+      // limitation: not direction-aware (ignores <- vs ->)
+      // limitation: not checking relationship variable reuse


Suggested change

// limitation: not checking relationship variable reuse

// limitation: not checking node label repetition

OskarDamkjaer · 2025-11-04T13:27:29Z

packages/language-support/src/autocompletion/schemaBasedCompletions.ts

+//  and returns the "labels" of rels/nodes going out/in of the node/rel in the graph schema
 function walkLabelTree(
-  relsFromLabels: Map<string, Set<string>>,
+  labelToConnectedLabelsMap: Map<string, Set<string>>,


Just a thought: is there a risk anywhere in our logic around people using the same names for a Reltype or label?

OskarDamkjaer · 2025-11-04T13:30:35Z

packages/language-support/src/autocompletion/schemaBasedCompletions.ts

+      let currentRelEntry = nodesFromRelsSet.get(rel.relType);
+      if (!currentRelEntry) {
+        nodesFromRelsSet.set(rel.relType, new Set());
+        currentRelEntry = nodesFromRelsSet.get(rel.relType);


I think it'd be cleaned to do a .has check first, and avoid the mutation of currentRelEntry

OskarDamkjaer · 2025-11-04T13:32:55Z

packages/language-support/src/autocompletion/schemaBasedCompletions.ts

+//  and returns the "labels" of rels/nodes going out/in of the node/rel in the graph schema
 function walkLabelTree(
-  relsFromLabels: Map<string, Set<string>>,
+  labelToConnectedLabelsMap: Map<string, Set<string>>,


I have some ideas on how we maybe could make this data structure better or more efficient, but I think we can wait until the feature matures to consider that refactoring

… into nodeSACompletions

…etion method

OskarDamkjaer

Would be nice to have a simple undirected match test

MATCH (t:Trainer)-[r:CATCHES]-(:';

and I think it could be good for a longer path as well

anderson4j added 2 commits October 28, 2025 13:31

adds schema-based completions for nodes too

78a2b43

adds cursor pos handling, and tests for node SB completions

c540c5b

anderson4j added the auto completion Issues related to the auto-completion label Oct 28, 2025

anderson4j commented Oct 28, 2025

View reviewed changes

anderson4j added 2 commits October 28, 2025 16:35

Merge branch 'main' of https://github.com/neo4j/cypher-language-support…

6b1bb61

… into nodeSACompletions

removes check not needed when slicing, slices in testing too

8021db9

anderson4j commented Oct 29, 2025

View reviewed changes

OskarDamkjaer reviewed Nov 4, 2025

View reviewed changes

anderson4j added 4 commits November 4, 2025 15:10

Merge branch 'main' of https://github.com/neo4j/cypher-language-support…

3fd79ec

… into nodeSACompletions

remove (extra) slice in testing now that we do it in shared autocompl…

d08fed1

…etion method

minor cleanup

9c0afcd

expanded testing

501e94d

OskarDamkjaer approved these changes Nov 5, 2025

View reviewed changes

adds tests using undirected rels

e18ee73

anderson4j merged commit e7d209f into main Nov 5, 2025
8 checks passed

anderson4j deleted the nodeSACompletions branch November 5, 2025 15:06

		test('Simple node completion pattern', () => {
		const query = 'MATCH (t:Trainer)-[r:CATCHES]->(:';

	// limitation: not checking relationship variable reuse
	// limitation: not checking node label repetition

Schema based completions for nodes in paths #595

Schema based completions for nodes in paths #595

Uh oh!

Conversation

anderson4j commented Oct 28, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

changeset-bot bot commented Oct 28, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

⚠️ No Changeset found

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

anderson4j Nov 4, 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!

Choose a reason for hiding this comment

Uh oh!

anderson4j Oct 29, 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!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

anderson4j Nov 4, 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!

Choose a reason for hiding this comment

Uh oh!

anderson4j Nov 4, 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!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

OskarDamkjaer left a comment

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

3 participants

anderson4j commented Oct 28, 2025 •

edited

Loading

changeset-bot bot commented Oct 28, 2025 •

edited

Loading

anderson4j Nov 4, 2025 •

edited

Loading

anderson4j Oct 29, 2025 •

edited

Loading

anderson4j Nov 4, 2025 •

edited

Loading

anderson4j Nov 4, 2025 •

edited

Loading