tests/test_document_loader.py

"""

makes sure that large text units are split into overlapping chunks without
losing source metadata, and that FAR section references can be identified from text.

Source metadata:
    Metadata such as the source file name and section identifies where each
    chunk originated. Preserving it allows retrieved chunks to remain linked
    to their original document and section

Why:
    Reliable chunk metadata improves traceability in retrieval workflows, while
    FAR section detection helps organize and cite regulatory content accurately
"""

'''
params:
    text: document content being processed

    source_name: name of the source document, e.g. ``policy.txt``

    source_type: file type of the source document, such as ``.TXT``

    section: section metadata associated with the source text

    chunk_size: max target size of each generated chunk, measured in characters

    overlap: no. of characters repeated between adjacent chunks to preserve context
'''

from document_loader import TextUnit, infer_far_section, units_to_chunks


def test_chunking_preserves_source_metadata():
    units = [TextUnit(text="A" * 2400, source_name="policy.txt", source_type="TXT", section="Section 1")]
    chunks = units_to_chunks(units, chunk_size=1000, overlap=100)
    assert len(chunks) >= 3
    assert all(c.source_name == "policy.txt" for c in chunks)
    assert all(c.section == "Section 1" for c in chunks)


def test_far_section_detection():
    assert infer_far_section("FAR 10.001 requires market research") == "FAR 10.001"