feat(lint): add nursery rule noUselessReturn

[ff1451]

I used Claude Code to assist with generating test cases and documentation.

Summary

Closes #8205.

Port ESLint's no-useless-return rule to Biome as a nursery rule
noUselessReturn.

A bare return; at the end of a function body is redundant — removing it does not change the function's behavior. This rule
detects such cases and provides a safe fix to remove them.

The implementation uses an AST-based approach: for each JsReturnStatement without an argument, it walks ancestors to check whether the return is in tail position relative to its enclosing function. It bails out for returns inside loops, switch
statements, and finally blocks.

Test Plan

• Snapshot tests in crates/biome_js_analyze/tests/specs/nursery/noUselessReturn/
  • invalid.js: 12 cases including end-of-function, after statements, if/else branches, nested ifs, try/catch blocks, arrow
    functions, labeled statements, and try-finally.
  • valid.js: 14 cases including return with value, early returns, loops, switch, finally block, empty function, and nested
    function scopes.
• All tests pass with cargo test -p biome_js_analyze -- no_useless_return.

Docs

Rule documentation is provided via inline rustdoc in the rule implementation, including invalid and valid examples with
expect_diagnostic annotations.

[changeset-bot]

🦋 Changeset detected

Latest commit: 3a444e4

The changes in this PR will be included in the next version bump.

Not sure what this means? Click here to learn what changesets are.

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

[coderabbitai]

Walkthrough

Adds a new JavaScript lint rule NoUselessReturn that detects redundant return; statements when they are the last reachable statement in a non-top-level function. The rule skips returns with values, excludes top-level contexts, analyses tail-position through blocks/conditionals/try/catch (conservatively excluding finally), emits a diagnostic and offers an auto-fix that removes the useless return. Also adds NoUselessReturnOptions, exposes the options module, and includes comprehensive valid/invalid tests.

Suggested reviewers

• dyc3
• ematipico
• Netail

🚥 Pre-merge checks | ✅ 4

✅ Passed checks (4 passed)

Check name	Status	Explanation
Title check	✅ Passed	The title clearly and concisely summarises the main change: adding a new nursery lint rule called noUselessReturn.
Description check	✅ Passed	The description is well-related to the changeset, explaining the rule's purpose, implementation approach, test coverage, and documentation.
Linked Issues check	✅ Passed	The PR successfully implements the objective from #8205: porting ESLint's no-useless-return rule to Biome with the name noUselessReturn, with AST-based detection and safe auto-fix.
Out of Scope Changes check	✅ Passed	All changes are scoped to implementing the noUselessReturn rule; no extraneous modifications outside the stated objective are present.

_{✏️ Tip: You can configure your own custom pre-merge checks in the settings.}

✨ Finishing touches

🧪 Generate unit tests (beta)

• Create PR with unit tests
• Post copyable unit tests in a comment

---

No actionable comments were generated in the recent review. 🎉

---

Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.

❤️ Share

• X
• Mastodon
• Reddit
• LinkedIn

_{Comment @coderabbitai help to get the list of available commands and usage tips.}

[coderabbitai]

Actionable comments posted: 1

🤖 Fix all issues with AI agents

In `@crates/biome_js_analyze/src/lint/nursery/no_useless_return.rs`:
- Line 75: The rule registration currently marks the ESLint mapping as identical
using RuleSource::Eslint("no-useless-return").same(), but the implementation
diverges from ESLint (AST-only vs CFG behavior); change the mapping to
.inspired() so it reflects different behavior. Locate the use of
RuleSource::Eslint("no-useless-return") in the no_useless_return rule
registration and replace .same() with .inspired(), ensuring any documentation or
rule metadata that relies on sameness is updated accordingly.

🧹 Nitpick comments (1)

crates/biome_js_analyze/src/lint/nursery/no_useless_return.rs (1)

1-7: Nit: inconsistent import style for biome_js_syntax types.

Some types (JsFinallyClause, JsFunctionBody, etc.) are imported at the top, while others (JsBlockStatement, JsIfStatement, JsModule, JsScript, etc.) are used fully-qualified in the function bodies (lines 102–103, 215–221). Either approach is fine, but mixing both is a bit untidy.

Add missing imports and use short names throughout

 use biome_js_syntax::{
-    JsDoWhileStatement, JsFinallyClause, JsForInStatement, JsForOfStatement, JsForStatement,
-    JsFunctionBody, JsReturnStatement, JsStatementList, JsSwitchStatement, JsWhileStatement,
+    JsBlockStatement, JsCatchClause, JsDoWhileStatement, JsElseClause, JsFinallyClause,
+    JsForInStatement, JsForOfStatement, JsForStatement, JsFunctionBody, JsIfStatement,
+    JsLabeledStatement, JsModule, JsReturnStatement, JsScript, JsStatementList,
+    JsSwitchStatement, JsTryFinallyStatement, JsTryStatement, JsWhileStatement,
 };

[dyc3]

We have ControlFlowGraph for analyzing control flow. Would that make the implementation of this rule more simple?

[codspeed-hq]

Merging this PR will not alter performance

✅ 58 untouched benchmarks
⏩ 95 skipped benchmarks¹

---

_{Comparing ff1451:feat/no-useless-return/#8205 (3a444e4) with main (9d4cfa3)}

Footnotes

1. 95 benchmarks were skipped, so the baseline results were used instead. If they were deleted from the codebase, click here and archive them to remove them from the performance reports. ↩

[ff1451]

I initially tried a CFG-based implementation using ControlFlowGraph as the query type, but the code grew to ~350 lines with additional dependencies (RoaringBitmap, FxHashSet) and required traversing all CFG blocks with forward-flow analysis to determine if a return was useless.

Since the core problem — "is this return; the last statement before the function ends?" — is fundamentally a positional question rather than a reachability one, the AST approach answers it more naturally by walking ancestors to check tail position (~170 lines, no extra dependencies).

If you'd like to review the CFG-based approach, I can share that implementation for comparison.
@dyc3

[dyc3]

Makes sense, thanks for the explanation.