feat(test): add leaderboard

[paul-nechifor]

Problem

Closes DIM-XXX

Solution

Breaking Changes

How to Test

Contributor License Agreement

• I have read and approved the CLA.

[greptile-apps]

Greptile Summary

Adds a new bin/test-speed-leaderboard script that runs pytest, collects test durations from JUnit XML, then uses git blame on each test function's source to attribute total test time to individual committers — producing a ranked leaderboard sorted by time-per-line.

• The script follows the same patterns as the existing bin/coverage-by-author tool (same git blame porcelain parsing, same repo root detection).
• Uses AST parsing and caching for efficient function line resolution, and blame caching to avoid redundant git calls.
• Minor issue: ast.walk in find_function_lines may match class methods when looking for module-level test functions.

Confidence Score: 4/5

• This PR is safe to merge — it adds a standalone developer utility script with no production code impact.
• Single new file that is a developer-only CLI tool (bin script). Follows existing codebase conventions closely (mirrors bin/coverage-by-author patterns). No production code is modified. One minor AST traversal issue noted but unlikely to cause practical problems.
• No files require special attention.

Important Files Changed

Filename	Overview
bin/test-speed-leaderboard	New 220-line Python script that runs pytest, parses JUnit XML, resolves test functions via AST, and uses git blame to attribute test durations to committers. Well-structured with caching; minor AST walk issue when class_name is None.

Flowchart

%%{init: {'theme': 'neutral'}}%%
flowchart TD
    A[Run pytest with JUnit XML output] --> B[Parse JUnit XML for test cases]
    B --> C[For each test case]
    C --> D[Convert classname to file path]
    D --> E[AST parse source file & find function lines]
    E --> F[git blame on function line range]
    F --> G[Attribute duration proportionally by author line count]
    G --> C
    C --> H[Sort authors by time-per-line]
    H --> I[Print ranked leaderboard table]

Loading

_{Last reviewed commit: 4aa2910}

[greptile-apps]

ast.walk may match class methods for module-level tests

When class_name is None (i.e. the test is a module-level function), ast.walk(tree) traverses every node in the AST including methods inside classes. If a class method happens to share the same name as a module-level test function, this will return whichever one ast.walk yields first — which has no guaranteed order. This could attribute the wrong line range to a test.

Consider restricting the no-class branch to only search tree.body (top-level statements) instead of walking the full tree:

Suggested change

	for node in ast.walk(tree):
	if class_name and isinstance(node, ast.ClassDef) and node.name == class_name:
	for item in node.body:
	if isinstance(item, (ast.FunctionDef, ast.AsyncFunctionDef)):
	if item.name == func_name:
	return item.lineno, item.end_lineno
	elif not class_name and isinstance(node, (ast.FunctionDef, ast.AsyncFunctionDef)):
	if node.name == func_name:
	return node.lineno, node.end_lineno
	def find_function_lines(tree, class_name, func_name):
	"""Return (start_line, end_line) of a test function in an AST."""
	for node in ast.walk(tree):
	if class_name and isinstance(node, ast.ClassDef) and node.name == class_name:
	for item in node.body:
	if isinstance(item, (ast.FunctionDef, ast.AsyncFunctionDef)):
	if item.name == func_name:
	return item.lineno, item.end_lineno
	if not class_name:
	for node in tree.body:
	if isinstance(node, (ast.FunctionDef, ast.AsyncFunctionDef)):
	if node.name == func_name:
	return node.lineno, node.end_lineno
	return None, None

[leshy]

approved without a review this is a fun helper