Fix: Skip nested functions in JavaScript/TypeScript discovery

[mohammedahmed18]

Bug Description

Nested functions (functions defined inside other functions) were being discovered and attempted to be optimized, causing 140+ optimization runs to fail with errors like:

Invalid source code: Undefined variable(s): base, streamFn, record, writer

Root Cause

The discover_functions method in codeflash/languages/javascript/support.py was allowing nested functions to be marked for optimization. When the extraction logic ran, it only captured the nested function body without the parent scope variables, causing semantic validation failures.

Example:

export function wrapStreamFn(streamFn) {
    const base = { id: 1 };
    
    const wrapped = (model, context, options) => {  // ← This was being optimized
        return streamFn(model, context, options);   // ← But streamFn is undefined!
    };
    
    return wrapped;
}

The wrapped function depends on streamFn and base from the parent scope, but the extractor only captured the nested function itself.

Fix

Added explicit check to skip functions with parent_function set during discovery:

# Skip nested functions (functions defined inside other functions)
if func.parent_function:
    logger.debug(f"Skipping nested function: {func.name} (parent: {func.parent_function})")
    continue

Nested functions are now filtered out at discovery time before extraction is attempted.

Testing

Added tests/languages/javascript/test_discover_functions.py with 4 test cases:

1. ✅ test_discovers_top_level_function - Ensures normal functions still work
2. ✅ test_skips_nested_functions_in_closures - Main bug fix verification
3. ✅ test_discovers_class_methods - Ensures class methods still work (different from nested)
4. ✅ test_skips_nested_functions_with_multiple_levels - Handles deeply nested functions

All tests pass. Linter (prek) passes.

Impact

• Resolves 140+ trace failures with undefined variable errors
• Functions like wrapStreamFn.wrapped, createAnthropicPayloadLogger.wrapStreamFn.wrapped will no longer be attempted
• Only top-level functions and class methods will be optimized

Affected Traces

This fix resolves errors in 140+ trace IDs including:

• 02a59310-bb18-47e4-87cb-1e5144ce2d8c (wrapStreamFn.wrapped)
• And many others with nested function patterns

Note

This is paired with PR #2519 (aiservice import validation fix). Together they resolve the most common optimization failures in the openclaw run.

[claude]

Claude finished @mohammedahmed18's task in 1m 57s —— View job

---

PR Review Summary

Prek Checks

All ruff checks pass. No auto-fixable issues. Mypy has 31 pre-existing errors in support.py — none introduced by this PR.

Code Review

Dead code at support.py:179–180

# Line 166–168: all functions with parent_function are already skipped
if func.parent_function:
    logger.debug(...)
    continue

# ...lines 175–178...

# This block can never execute — func.parent_function is always falsy here
if func.parent_function:                                           # line 179
    parents.append(FunctionParent(name=func.parent_function, type="FunctionDef"))  # line 180

The continue on line 168 means any function with parent_function set never reaches line 179. This dead code was load-bearing before — it used to add nested function context to the parents list for functions that passed the old export filter. Now it's unreachable and should be removed. Fix this →

The fix logic itself (adding if func.parent_function: continue and removing the not func.parent_function exception in the export filter) is correct.

Duplicate Detection

No duplicates detected. Python's discover_functions doesn't track parent_function at all, so there's no parallel logic to align.

Test Coverage

All 4 new tests pass. Coverage is appropriate for a SMALL PR — the key happy path, the bug scenario, class methods, and multi-level nesting are all covered.

---

Last updated: 2026-04-02

[CLAassistant]

Thank you for your submission! We really appreciate it. Like many open source projects, we ask that you sign our Contributor License Agreement before we can accept your contribution.

---

Codeflash Bot seems not to be a GitHub user. You need a GitHub account to be able to sign the CLA. If you have already a GitHub account, please add the email address used for this commit to your account.
_{You have signed the CLA already but the status is still pending? Let us recheck it.}