from datetime import datetime
from src.chunking.header_chunker import HeaderBasedChunker
from src.config import ChunkingConfig
from src.models import Document
class TestParentChildChunking:
def test_creates_parent_and_child_chunks_when_enabled(self):
config = ChunkingConfig(
strategy="header_based",
min_chunk_chars=100,
max_chunk_chars=400,
overlap_chars=0,
parent_retrieval_enabled=True,
parent_chunk_min_chars=500,
parent_chunk_max_chars=1000,
)
chunker = HeaderBasedChunker(config)
content = """# Main Title
This is the introduction with some content that should be reasonably long.
## Section One
First section content with details about topic A. More content here to make
the chunk larger. Adding extra text to ensure we have enough characters.
## Section Two
Second section content about topic B. More details and explanations here.
Additional content to reach minimum chunk size for testing purposes.
## Section Three
Third section content covering topic C. Further elaboration and examples.
More text to ensure this section is substantial enough for chunking.
"""
doc = Document(
id="test_doc",
content=content,
metadata={},
links=[],
tags=[],
file_path="/test/doc.md",
modified_time=datetime.now(),
)
chunks = chunker.chunk_document(doc)
child_chunks = [c for c in chunks if "_parent_" not in c.chunk_id]
# Should have both parents and children when parent retrieval is enabled
# and content is long enough
assert len(chunks) > 0
# Child chunks should have parent_chunk_id set
for child in child_chunks:
if child.parent_chunk_id:
assert child.parent_chunk_id.startswith("test_doc_parent_")
def test_no_parent_chunks_when_disabled(self):
config = ChunkingConfig(
strategy="header_based",
min_chunk_chars=100,
max_chunk_chars=400,
overlap_chars=0,
parent_retrieval_enabled=False,
)
chunker = HeaderBasedChunker(config)
content = """# Title
Some content here.
## Section
More content in this section.
"""
doc = Document(
id="test_doc",
content=content,
metadata={},
links=[],
tags=[],
file_path="/test/doc.md",
modified_time=datetime.now(),
)
chunks = chunker.chunk_document(doc)
# No parent chunks when disabled
parent_chunks = [c for c in chunks if "_parent_" in c.chunk_id]
assert len(parent_chunks) == 0
# All chunks should have no parent_chunk_id
for chunk in chunks:
assert chunk.parent_chunk_id is None
def test_child_chunks_reference_correct_parent(self):
config = ChunkingConfig(
strategy="header_based",
min_chunk_chars=50,
max_chunk_chars=200,
overlap_chars=0,
parent_retrieval_enabled=True,
parent_chunk_min_chars=300,
parent_chunk_max_chars=800,
)
chunker = HeaderBasedChunker(config)
content = """# Doc Title
Introduction paragraph with enough text to form a chunk.
## First Section
Content for section one with sufficient length for a chunk.
## Second Section
Content for section two with adequate length for testing.
## Third Section
Content for section three with more text for the chunk.
"""
doc = Document(
id="test_doc",
content=content,
metadata={},
links=[],
tags=[],
file_path="/test/doc.md",
modified_time=datetime.now(),
)
chunks = chunker.chunk_document(doc)
parent_chunks = {c.chunk_id: c for c in chunks if "_parent_" in c.chunk_id}
child_chunks = [c for c in chunks if "_parent_" not in c.chunk_id]
# Each child with a parent_chunk_id should reference an existing parent
for child in child_chunks:
if child.parent_chunk_id:
assert child.parent_chunk_id in parent_chunks
def test_parent_content_contains_child_content(self):
config = ChunkingConfig(
strategy="header_based",
min_chunk_chars=50,
max_chunk_chars=200,
overlap_chars=0,
parent_retrieval_enabled=True,
parent_chunk_min_chars=300,
parent_chunk_max_chars=1000,
)
chunker = HeaderBasedChunker(config)
content = """# Document
Intro text that should be included.
## Section A
Content for section A with enough text.
## Section B
Content for section B with enough text.
"""
doc = Document(
id="test_doc",
content=content,
metadata={},
links=[],
tags=[],
file_path="/test/doc.md",
modified_time=datetime.now(),
)
chunks = chunker.chunk_document(doc)
parent_chunks = {c.chunk_id: c for c in chunks if "_parent_" in c.chunk_id}
child_chunks = [c for c in chunks if "_parent_" not in c.chunk_id]
# Child content should be part of parent content
for child in child_chunks:
if child.parent_chunk_id and child.parent_chunk_id in parent_chunks:
parent = parent_chunks[child.parent_chunk_id]
# The child content (without overlap markers) should be in parent
child_text = child.content
if child_text.startswith("[..."):
# Remove overlap prefix
child_text = child_text.split("]\n\n", 1)[-1]
# Check content is part of parent (allowing for whitespace differences)
assert any(
line.strip() in parent.content
for line in child_text.split("\n")
if line.strip()
)
