🐛 fix: report error for undefined field access on closed records by harehare · Pull Request #1365 · harehare/mq

harehare · 2026-03-01T01:15:28Z

Add UndefinedField error type and diagnostics
Report error when accessing non-existent field on closed record via selector or bracket notation
Add integration tests for undefined field access

…ecords - Add UndefinedField error type and diagnostics - Report error when accessing non-existent field on closed record via selector or bracket notation - Add integration tests for undefined field access

Copilot

Pull request overview

Adds a dedicated typechecker diagnostic for accessing missing fields on closed record types, and wires the check into both selector and bracket-access resolution paths.

Changes:

Introduces TypeError::UndefinedField with a miette diagnostic code and source labeling.
Reports UndefinedField when accessing a non-existent field on a closed record (including deferred-resolution cases).
Adds integration tests around undefined field access on closed records (currently bracket-only).

Reviewed changes

Copilot reviewed 4 out of 4 changed files in this pull request and generated 2 comments.

File	Description
crates/mq-typechecker/tests/integration_test.rs	Adds integration tests for undefined field access on closed records (bracket access; selector coverage still missing).
crates/mq-typechecker/src/lib.rs	Adds `UndefinedField` error variant and post-unification resolution for deferred selector accesses.
crates/mq-typechecker/src/infer.rs	Adds deferred-selector access tracking in the inference context to support post-unification checking.
crates/mq-typechecker/src/constraint.rs	Emits `UndefinedField` for selector access on closed records when record type is known; defers selector checks when the piped type isn’t yet resolved.

Comments suppressed due to low confidence (1)

crates/mq-typechecker/tests/integration_test.rs:1152

The two tests named test_selector_* are exercising bracket access (v[:c] / v[:a]), not selector access (e.g. | .c). This makes the test names and failure messages misleading (e.g. message mentions .c). Rename them or change the queries to actually use selector syntax so the intent matches the exercised behavior.

fn test_selector_undefined_field_on_closed_record() {
    // Accessing a non-existent field via bracket notation on a closed record should produce an error
    let result = check_types(r#"var v = {"a": 1, "b": 2} | v[:c]"#);
    assert!(
        !result.is_empty(),
        "Accessing undefined field .c on closed record should fail: {:?}",
        result
    );

Copilot · 2026-03-01T01:19:29Z

crates/mq-typechecker/tests/integration_test.rs

+fn test_selector_undefined_field_on_closed_record() {
+    // Accessing a non-existent field via bracket notation on a closed record should produce an error
+    let result = check_types(r#"var v = {"a": 1, "b": 2} | v[:c]"#);
+    assert!(
+        !result.is_empty(),
+        "Accessing undefined field .c on closed record should fail: {:?}",
+        result
+    );
+}
+
+#[test]
+fn test_selector_defined_field_on_closed_record() {
+    // Accessing an existing field via bracket notation on a closed record should succeed
+    let result = check_types(r#"var v = {"a": 1, "b": 2} | v[:a]"#);
+    assert!(
+        result.is_empty(),
+        "Bracket access to defined field :a on closed record should succeed: {:?}",
+        result
+    );
+}
+
+#[test]
+fn test_bracket_access_undefined_field_on_closed_record() {
+    // Accessing a non-existent field via bracket notation should produce an error
+    let result = check_types(r#"var v = {"key": 1, "value": "hello"} | v[:missing]"#);
+    assert!(
+        !result.is_empty(),
+        "Bracket access to undefined field :missing on closed record should fail: {:?}",
+        result
+    );


These new "undefined field access" tests only assert that some error is produced (!result.is_empty()), which could pass for an unrelated type error. Since this PR introduces TypeError::UndefinedField, the tests should assert that the returned errors contain (and ideally are dominated by) UndefinedField for the relevant field name.

@copilot
Can you fix it?

Copilot · 2026-03-01T01:19:29Z

crates/mq-typechecker/tests/integration_test.rs

+// --- Undefined field access tests ---
+
+#[test]
+fn test_selector_undefined_field_on_closed_record() {
+    // Accessing a non-existent field via bracket notation on a closed record should produce an error
+    let result = check_types(r#"var v = {"a": 1, "b": 2} | v[:c]"#);
+    assert!(
+        !result.is_empty(),
+        "Accessing undefined field .c on closed record should fail: {:?}",
+        result
+    );
+}
+
+#[test]
+fn test_selector_defined_field_on_closed_record() {
+    // Accessing an existing field via bracket notation on a closed record should succeed
+    let result = check_types(r#"var v = {"a": 1, "b": 2} | v[:a]"#);
+    assert!(
+        result.is_empty(),
+        "Bracket access to defined field :a on closed record should succeed: {:?}",
+        result
+    );
+}
+
+#[test]
+fn test_bracket_access_undefined_field_on_closed_record() {
+    // Accessing a non-existent field via bracket notation should produce an error
+    let result = check_types(r#"var v = {"key": 1, "value": "hello"} | v[:missing]"#);
+    assert!(
+        !result.is_empty(),
+        "Bracket access to undefined field :missing on closed record should fail: {:?}",
+        result
+    );
+}
+
+#[test]
+fn test_bracket_access_defined_field_on_closed_record() {
+    // Accessing an existing field via bracket notation should succeed
+    let result = check_types(r#"var v = {"key": 1, "value": "hello"} | v[:key]"#);
+    assert!(
+        result.is_empty(),
+        "Bracket access to defined field :key on closed record should succeed: {:?}",
+        result
+    );
+}


PR description mentions reporting errors for undefined field access via selector or bracket notation, but the added integration coverage here only exercises bracket access (v[:missing], etc.). Add at least one failing and one passing test for selector access on a closed record (e.g. {"a": 1} | .missing and {"a": 1} | .a) to ensure the deferred selector-resolution path is covered.

Copilot · 2026-03-01T02:20:19Z

@harehare I've opened a new pull request, #1366, to work on those changes. Once the pull request is ready, I'll request review from you.

… access tests Co-authored-by: harehare <[email protected]>

Fix undefined field access tests to assert TypeError::UndefinedField specifically

Copilot

Pull request overview

Copilot reviewed 4 out of 4 changed files in this pull request and generated 3 comments.

Copilot · 2026-03-01T04:21:16Z

crates/mq-typechecker/src/constraint.rs

+                    // Piped type not yet resolved — defer check to post-unification
+                    if let Some(name) = field_name {
+                        ctx.add_deferred_selector_access(crate::infer::DeferredSelectorAccess {
+                            symbol_id,
+                            piped_ty: piped_ty.clone(),
+                            field_name: name.to_string(),
+                            range: get_symbol_range(hir, symbol_id),
+                        });
+                    }


The PR description mentions reporting UndefinedField via both selector (.field) and bracket ([:field]) notation, but the added integration tests only exercise bracket access. Add at least one test that uses selector syntax (both defined and undefined fields) to cover the new deferred selector resolution path.

Copilot · 2026-03-01T04:21:16Z

crates/mq-typechecker/src/lib.rs

        }

+        // Resolve deferred selector field accesses (.field on records)
+        Self::resolve_selector_field_accesses(&mut ctx);


resolve_selector_field_accesses adds new Constraint::Equal constraints (to bind selector type to the resolved record field type), but TypeChecker::check doesn’t run unification afterwards. If there are no later phases that call unify::solve_constraints (e.g. no deferred overloads/user calls), selector result types can remain as unbound type variables and downstream type inference can be incorrect. Consider making resolve_selector_field_accesses return a bool like resolve_record_field_accesses and re-running unify::solve_constraints when any selector was resolved (or bind the symbol type directly instead of adding a constraint).

Suggested change

Self::resolve_selector_field_accesses(&mut ctx);

Self::resolve_selector_field_accesses(&mut ctx);

// Re-run unification to propagate newly added selector field constraints.

// This ensures that selector result types are fully resolved before

// subsequent phases rely on them.

unify::solve_constraints(&mut ctx);

Copilot · 2026-03-01T04:21:17Z

crates/mq-typechecker/tests/integration_test.rs

+fn test_selector_undefined_field_on_closed_record() {
+    // Accessing a non-existent field via bracket notation on a closed record should produce an error
+    let result = check_types(r#"var v = {"a": 1, "b": 2} | v[:c]"#);
+    assert!(
+        result
+            .iter()
+            .any(|e| matches!(e, TypeError::UndefinedField { field, .. } if field == "c")),
+        "Accessing undefined field :c on closed record should produce UndefinedField error: {:?}",
+        result
+    );
+}
+
+#[test]
+fn test_selector_defined_field_on_closed_record() {
+    // Accessing an existing field via bracket notation on a closed record should succeed
+    let result = check_types(r#"var v = {"a": 1, "b": 2} | v[:a]"#);
+    assert!(


These test names/comments say “selector” but the queries are bracket accesses (v[:c], v[:a]). This makes it hard to tell what syntax is actually being covered. Either rename these to test_bracket_access_* or change the queries to use selector syntax (e.g. v.c) so the test intent matches the code.

fix(typechecker): report error for undefined field access on closed r…

8bbf2a5

…ecords - Add UndefinedField error type and diagnostics - Report error when accessing non-existent field on closed record via selector or bracket notation - Add integration tests for undefined field access

Copilot AI review requested due to automatic review settings March 1, 2026 01:15

harehare changed the title ~~fix(typechecker): report error for undefined field access on closed r…~~ fix(typechecker): report error for undefined field access on closed records Mar 1, 2026

Copilot started reviewing on behalf of harehare March 1, 2026 01:15 View session

Copilot AI reviewed Mar 1, 2026

View reviewed changes

Initial plan

798380f

Copilot AI mentioned this pull request Mar 1, 2026

Fix undefined field access tests to assert TypeError::UndefinedField specifically #1366

Merged

Copilot AI and others added 3 commits March 1, 2026 02:25

fix(typechecker): assert UndefinedField error type in undefined field…

6e7195d

… access tests Co-authored-by: harehare <[email protected]>

Merge pull request #1366 from harehare/copilot/sub-pr-1365

729c1e8

Fix undefined field access tests to assert TypeError::UndefinedField specifically

🎨 Format code

92f94b2

Copilot AI review requested due to automatic review settings March 1, 2026 04:17

Copilot started reviewing on behalf of harehare March 1, 2026 04:18 View session

harehare changed the title ~~fix(typechecker): report error for undefined field access on closed records~~ 🐛 fix: report error for undefined field access on closed records Mar 1, 2026

Copilot AI reviewed Mar 1, 2026

View reviewed changes

harehare merged commit 76e5ed6 into main Mar 1, 2026
10 checks passed

harehare deleted the fix/typechecker-undefined-field-error branch March 1, 2026 04:41

BrewTestBot mentioned this pull request Mar 11, 2026

mq 0.5.18 Homebrew/homebrew-core#271763

Merged

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

🐛 fix: report error for undefined field access on closed records#1365

🐛 fix: report error for undefined field access on closed records#1365
harehare merged 5 commits intomainfrom
fix/typechecker-undefined-field-error

harehare commented Mar 1, 2026 •

edited

Loading

Uh oh!

Copilot AI left a comment

Uh oh!

Copilot AI Mar 1, 2026

Uh oh!

harehare Mar 1, 2026

Uh oh!

Copilot AI Mar 1, 2026

Uh oh!

Copilot AI commented Mar 1, 2026

Uh oh!

Copilot AI left a comment

Uh oh!

Copilot AI Mar 1, 2026

Uh oh!

Copilot AI Mar 1, 2026

Uh oh!

Copilot AI Mar 1, 2026

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

3 participants

-        Self::resolve_selector_field_accesses(&mut ctx);
+        Self::resolve_selector_field_accesses(&mut ctx);
+        // Re-run unification to propagate newly added selector field constraints.
+        // This ensures that selector result types are fully resolved before
+        // subsequent phases rely on them.
+        unify::solve_constraints(&mut ctx);

Conversation

harehare commented Mar 1, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

Copilot AI left a comment

Choose a reason for hiding this comment

Pull request overview

Reviewed changes

Uh oh!

Copilot AI Mar 1, 2026

Choose a reason for hiding this comment

Uh oh!

harehare Mar 1, 2026

Choose a reason for hiding this comment

Uh oh!

Copilot AI Mar 1, 2026

Choose a reason for hiding this comment

Uh oh!

Copilot AI commented Mar 1, 2026

Uh oh!

Copilot AI left a comment

Choose a reason for hiding this comment

Pull request overview

Uh oh!

Copilot AI Mar 1, 2026

Choose a reason for hiding this comment

Uh oh!

Copilot AI Mar 1, 2026

Choose a reason for hiding this comment

Uh oh!

Copilot AI Mar 1, 2026

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

3 participants

harehare commented Mar 1, 2026 •

edited

Loading