from chunkhound.code_mapper.public_utils import (
is_empty_research_result,
merge_sources_metadata,
)
def test_merge_sources_metadata_deduplicates_files_and_chunks() -> None:
"""merge_sources_metadata should deduplicate files and chunk ranges."""
results = [
{
"metadata": {
"sources": {
"files": ["scope/a.py", "scope/b.py"],
"chunks": [
{"file_path": "scope/a.py", "start_line": 1, "end_line": 10},
# Duplicate of the same chunk; should be collapsed.
{"file_path": "scope/a.py", "start_line": 1, "end_line": 10},
],
},
"aggregation_stats": {"files_total": 10, "chunks_total": 20},
},
},
{
"metadata": {
"sources": {
"files": ["scope/b.py", "scope/c.py"],
"chunks": [
{"file_path": "scope/b.py", "start_line": 5, "end_line": 15},
{"file_path": "scope/c.py", "start_line": 7, "end_line": 9},
],
},
# Later aggregation stats should be able to override earlier zeros.
"aggregation_stats": {"files_total": 10, "chunks_total": 20},
},
},
]
unified_files, unified_chunks, total_files, total_chunks = merge_sources_metadata(
results
)
# Files are deduplicated and normalized as keys in the mapping.
assert set(unified_files.keys()) == {"scope/a.py", "scope/b.py", "scope/c.py"}
# Chunks are deduplicated by (file_path, start_line, end_line).
assert len(unified_chunks) == 3
assert {
"file_path": "scope/a.py",
"start_line": 1,
"end_line": 10,
} in unified_chunks
assert {
"file_path": "scope/b.py",
"start_line": 5,
"end_line": 15,
} in unified_chunks
assert {
"file_path": "scope/c.py",
"start_line": 7,
"end_line": 9,
} in unified_chunks
# Aggregation stats are propagated when present.
assert total_files == 10
assert total_chunks == 20
def test_is_empty_research_result_detects_skipped_synthesis() -> None:
"""is_empty_research_result should detect DeepResearch 'no context' responses."""
result = {
"answer": (
"No relevant code context found for: 'foo'.\n\n"
"Try a more code-specific question."
),
"metadata": {
"skipped_synthesis": True,
"depth_reached": 0,
"nodes_explored": 1,
"chunks_analyzed": 0,
"files_analyzed": 0,
},
}
assert is_empty_research_result(result)
non_empty = {
"answer": "# Heading\n\nSome real content.",
"metadata": {},
}
assert not is_empty_research_result(non_empty)
