find_tables(): keep quad corrections on for the line-based finder (#5198)

src/table.py

@@ -3242,6 +3242,15 @@ def find_tables(
             tset = TableSettings.resolve(settings=settings)
             page.table_settings = tset
 
+            if not boxes:
+                # Without layout boxes this is the line-based finder, whose
+                # characters need the minimum glyph heights set above. Those
+                # only take effect with quad corrections on; without them each
+                # char keeps its exact glyph box, a "." sits lower than the
+                # digits around it and is grouped onto a line of its own
+                # ("0.74" becomes "0 74\n."). Turn corrections on, whatever
+                # get_layout() above or an import of pymupdf4llm left them at.
+                pymupdf.TOOLS.unset_quad_corrections(False)
             TEXTPAGE = make_chars(page, clip=clip)  # create character list of page
             make_edges(
                 page,

tests/test_tables.py

@@ -2017,3 +2017,44 @@ def test_5241():
     names_after = list(tab.header.names)
     assert names_before == names_after, "Column headers changed by to_pandas()"
 
+
+def test_5198():
+    """Cell text keeps decimal points when quad corrections are off.
+
+    Importing pymupdf4llm turns quad corrections off for the whole process.
+    find_tables() then built its characters from exact glyph boxes, so the
+    "." of "0.74" sat lower than the digits, was grouped onto a line of its
+    own and the cell read "0 74\\n.".
+    """
+    doc = pymupdf.open()
+    page = doc.new_page(width=595, height=842)
+    x0, y0, x1, y1, xm, ym = 72, 72, 472, 152, 272, 112
+    for p1, p2 in [
+        ((x0, y0), (x1, y0)),
+        ((x0, ym), (x1, ym)),
+        ((x0, y1), (x1, y1)),
+        ((x0, y0), (x0, y1)),
+        ((xm, y0), (xm, y1)),
+        ((x1, y0), (x1, y1)),
+    ]:
+        page.draw_line(p1, p2)
+    for text, point in [
+        ("P for interaction", (80, 100)),
+        ("Cooking fuel type", (280, 100)),
+        ("0.74", (80, 140)),
+        ("12.5 mg", (280, 140)),
+    ]:
+        page.insert_text(point, text, fontsize=11)
+
+    quad_before = bool(pymupdf.TOOLS.unset_quad_corrections())
+    pymupdf.TOOLS.unset_quad_corrections(True)  # as importing pymupdf4llm does
+    try:
+        tables = page.find_tables(use_layout=False).tables
+        assert [t.extract() for t in tables] == [
+            [["P for interaction", "Cooking fuel type"], ["0.74", "12.5 mg"]]
+        ]
+        # find_tables() restores the setting it found
+        assert pymupdf.TOOLS.unset_quad_corrections() is True
+    finally:
+        pymupdf.TOOLS.unset_quad_corrections(quad_before)
+        doc.close()