v0.6.0 BigUpdate

tests/test_word_fix.py

@@ -14,8 +14,11 @@ from md2gost.word_fix import (
     find_page_break_row,
     fix_continuations,
     is_caption_orphaned,
+    continuation_pass_limit,
     is_orphan_header_break,
+    locate_break_row,
     parse_caption_text,
+    resume_page_before_cut,
     split_vmerge_spans_at_break,
 )
 
@@ -61,6 +64,68 @@ def test_find_page_break_row():
     assert find_page_break_row([14, 14, 14, 14, 14, 14, 15, 15]) == 7
 
 
+def test_resume_page_before_cut():
+    assert resume_page_before_cut(1) == 1
+    assert resume_page_before_cut(2) == 1
+    assert resume_page_before_cut(15) == 14
+
+
+def test_locate_break_row_matches_full_scan_from_page_1():
+    samples = [
+        [1, 1, 2, 2],
+        [2, 2, 3],
+        [1, 2],
+        [14, 14, 14, 14, 14, 14, 15, 15],
+        [1, 1, 1],
+        [8] * 30 + [9] * 10 + [12] * 5,
+    ]
+    for pages in samples:
+        def page_at(row, pages=pages):
+            return pages[row - 1]
+
+        found, cut = locate_break_row(len(pages), page_at, 1)
+        assert found == find_page_break_row(pages)
+        if found:
+            assert cut == pages[found - 1]
+
+
+def test_locate_break_row_starts_at_page_before_cut():
+    """A long table is not read from row 1 when the window starts near the cut."""
+    pages = [10] * 80 + [14] * 8 + [15] * 120
+    probes: list[int] = []
+
+    def page_at(row):
+        probes.append(row)
+        return pages[row - 1]
+
+    found, cut = locate_break_row(len(pages), page_at, min_page=14)
+    assert found == 81
+    assert cut == 14
+    assert len(probes) < 40
+
+    continuation = [15] * 12 + [16] * 200
+    cont_probes: list[int] = []
+
+    def continuation_at(row):
+        cont_probes.append(row)
+        return continuation[row - 1]
+
+    found, cut = locate_break_row(
+        len(continuation),
+        continuation_at,
+        min_page=resume_page_before_cut(15),
+    )
+    assert (found, cut) == (13, 16)
+    assert len(cont_probes) < 30
+
+
+def test_continuation_pass_limit():
+    assert continuation_pass_limit(0, 0) == 10
+    assert continuation_pass_limit(3, 0) == 13
+    assert continuation_pass_limit(3, 1) == 14
+    assert continuation_pass_limit(3, 5) == 18
+
+
 def test_orphan_header_and_caption_guards():
     """Method guide: no title-only / title+column-headers alone at page end."""
     # Split would leave only heading row on previous page
@@ -104,6 +169,68 @@ def test_fix_continuations_missing_file():
     assert "не найден" in r.message.lower() or "Файл" in r.message
 
 
+def test_word_com_apartment_pairs_init(monkeypatch):
+    import types
+
+    calls = []
+    fake = types.ModuleType("pythoncom")
+    fake.CoInitialize = lambda: calls.append("init")
+    fake.CoUninitialize = lambda: calls.append("uninit")
+    monkeypatch.setitem(sys.modules, "pythoncom", fake)
+    from md2gost.word_fix import word_com_apartment
+
+    with word_com_apartment():
+        calls.append("body")
+    assert calls == ["init", "body", "uninit"]
+
+
+def test_word_com_apartment_skips_uninit_when_init_fails(monkeypatch):
+    import types
+
+    calls = []
+    fake = types.ModuleType("pythoncom")
+
+    def boom():
+        raise RuntimeError("coinit")
+
+    fake.CoInitialize = boom
+    fake.CoUninitialize = lambda: calls.append("uninit")
+    monkeypatch.setitem(sys.modules, "pythoncom", fake)
+    from md2gost.word_fix import word_com_apartment
+
+    with pytest.raises(RuntimeError, match="coinit"):
+        with word_com_apartment():
+            calls.append("body")
+    assert calls == []
+
+
+@pytest.mark.skipif(sys.platform != "win32", reason="Windows only")
+def test_second_thread_can_dispatch_com():
+    """A fresh thread must call CoInitialize itself; module import does it once."""
+    try:
+        import win32com.client  # noqa: F401
+    except ImportError:
+        pytest.skip("pywin32 not installed")
+    import threading
+
+    from md2gost.word_fix import word_com_apartment
+
+    errors: list[BaseException] = []
+
+    def work():
+        try:
+            with word_com_apartment():
+                win32com.client.Dispatch("Scripting.Dictionary")
+        except Exception as exc:
+            errors.append(exc)
+
+    for _ in range(2):
+        thread = threading.Thread(target=work)
+        thread.start()
+        thread.join()
+    assert errors == []
+
+
 @pytest.mark.skipif(sys.platform != "win32", reason="Windows only")
 @pytest.mark.skipif(os.environ.get("MD2GOST_TEST_WORD") != "1", reason="set MD2GOST_TEST_WORD=1 to run Word COM smoke")
 def test_fix_continuations_com_smoke(tmp_path):