Fix: Exclude property getters/setters from function discovery

[mohammedahmed18]

Problem

When functions are defined as property getters/setters (e.g., inside Object.defineProperty with get: function getrouter() {...}), they were being discovered as optimizable functions. This caused test generation to fail with:

TypeError: Cannot read properties of undefined (reading 'bind')

Reproduction

In Express.js application.js:

Object.defineProperty(app, 'router', {
  get: function getrouter() {
    return router;
  }
});

Codeflash discovered getrouter as an optimizable function and generated tests like:

const r1 = init.getrouter();  // ❌ undefined - getrouter is not exported

But it should be:

const r1 = init.router;  // ✅ accesses property, triggers getter

Solution

Added check in _walk_tree_for_functions() to exclude function_expression nodes that are property getters/setters:

• Inside a pair node (key-value in object literal)
• Where the key is "get" or "set"

Testing

• ✅ New test suite: test_property_getter_exclusion.py
• ✅ Verified on actual Express.js code (getrouter no longer discovered)
• ✅ All existing JavaScript discovery tests pass
• ✅ Linting (prek) passes

Affected Trace IDs

From /workspace/logs analysis:

• 11b5475c-91d9-43b1-a317-33b7f4811c52 (init.getrouter)
• 78d8f1de-541a-4208-85c7-77194c70ec72 (createETagGenerator.generateETag)
• And others with similar test failures

Files Changed

• codeflash/languages/javascript/treesitter.py: Added getter/setter exclusion logic
• tests/test_property_getter_exclusion.py: Comprehensive test coverage

---

🤖 Generated with Claude Code

[claude]

Claude finished @mohammedahmed18's task in 1m 30s —— View job

---

PR Review Summary

• Triage: SMALL (12 lines of production code)
• Lint and type check (prek: ✅ passed)
• Resolve stale review threads (none found)
• Code review
• Duplicate detection
• Summary comment

---

Prek Checks

✅ ruff check and ruff format passed. No issues.

---

Code Review

Fix is correct for the stated use case, but has a potential false-positive edge case worth acknowledging.

The logic (treesitter.py:297-303):

if node.type == "function_expression" and node.parent and node.parent.type == "pair":
    property_name_node = node.parent.child_by_field_name("key")
    if property_name_node:
        property_name = self.get_node_text(property_name_node, source_bytes)
        if property_name in ("get", "set"):
            should_include = False

False positive concern: This check fires on any function_expression stored under a property key named "get" or "set", not only inside Object.defineProperty. For example:

// HTTP route handlers - "get" is just a property name, not a getter
const routes = {
  get: function handleGet(req, res) { return res.send("hello"); },
  set: function handleSet(req, res) { return res.send("ok"); }
};

Here handleGet is a directly callable, exportable function, but the fix would silently skip it from discovery. In HTTP frameworks (Express, Fastify, etc.), get/set/post/delete as property names for route handlers is a very common pattern.

More precise fix would be to walk up the parent chain and verify the pair is inside Object.defineProperty(...), or to check for the actual tree-sitter getter_definition node type (used for ES6 get foo() {} syntax). However, the Object.defineProperty case specifically uses function_expression in a pair — so the false positive only affects codebases that coincidentally use property keys literally named "get" or "set" to store named function expressions.

Test coverage issue: test_object_literal_getter_excluded (test_property_getter_exclusion.py:100) tests get router() { ... } syntax (ES6 getter shorthand), which tree-sitter parses as a getter_definition node — not function_expression. This test does not exercise the new code path. It would pass even without the new 12 lines. Consider replacing it with a test that confirms { get: function namedFn() {} } is excluded in a non-defineProperty context to at least document the known false-positive behavior (or add a test for the Object.defineProperty pattern specifically).

No bugs that crash at runtime. No security issues. No breaking API changes. No duplicate logic found in other language modules.

---

Duplicate Detection

No duplicates detected. The check is specific to languages/javascript/treesitter.py and has no parallel in the Python discovery path (Python uses AST decorators for properties, handled differently).

---

Test Coverage

New test file covers the primary use case (Express.js Object.defineProperty getter pattern). One test case (test_object_literal_getter_excluded) does not actually exercise the new code. Two test cases do exercise the new code path. Coverage is adequate for the bug being fixed.

---

Last updated: 2026-04-01T02:00:00Z

[Saga4]

We could create a list of such function types to avoid.