From 5e44495676c8d3c3ed1b83d4fa92ca07c0b3efd7 Mon Sep 17 00:00:00 2001 From: Jean-Luc Makiola Date: Mon, 4 May 2026 22:13:34 +0200 Subject: [PATCH] test: regression tests fuer overlap>=size und boundary ausserhalb lookback --- tests/test_chunker.py | 16 ++++++++++++++++ 1 file changed, 16 insertions(+) diff --git a/tests/test_chunker.py b/tests/test_chunker.py index 0a3b248..534beb9 100644 --- a/tests/test_chunker.py +++ b/tests/test_chunker.py @@ -52,3 +52,19 @@ def test_chunk_empty_text_returns_empty_list(): def test_chunk_carries_page_number(): chunks = chunk_text("hallo welt", size_words=500, overlap_words=50, page=7) assert chunks[0].page == 7 + + +def test_chunk_overlap_geq_size_does_not_loop_forever(): + words = [f"w{i}" for i in range(60)] + text = " ".join(words) + chunks = chunk_text(text, size_words=10, overlap_words=10, page=1) + assert len(chunks) >= 5 + assert all(len(c.text.split()) <= 10 for c in chunks) + + +def test_chunk_ignores_sentence_boundary_outside_lookback_window(): + words = [f"w{i}" for i in range(600)] + words[399] = "ende." + text = " ".join(words) + chunks = chunk_text(text, size_words=500, overlap_words=50, page=1) + assert len(chunks[0].text.split()) == 500