fix: preserve zero-width characters and ligatures in table cell extraction

src/table.py

@@ -348,6 +348,7 @@ def extract_cells(textpage, cell, markdown=False):
         A string with the text extracted from the cell.
     """
     text = ""
+    cell_rect = pymupdf.Rect(cell)
     for block in textpage.extractRAWDICT()["blocks"]:
         if block["type"] != 0:
             continue
@@ -388,12 +389,21 @@ def extract_cells(textpage, cell, markdown=False):
                 ):
                     continue  # skip spans outside cell
 
-                # only include chars with more than 50% bbox overlap
+                # only include chars with more than 50% bbox overlap.
+                # A char with an empty bbox (e.g. second character of an expanded
+                # ligature with quad corrections unset, or a zero-width combining mark)
+                # has no area; fall back to testing if its origin is inside the cell:
+                # issue #5185.
                 span_text = ""
                 for char in span["chars"]:
                     this_char = char["c"]
                     bbox = pymupdf.Rect(char["bbox"])
-                    if abs(bbox & cell) > 0.5 * abs(bbox):
+                    if bbox.is_empty:
+                        in_cell = pymupdf.Point(char["origin"]) in cell_rect
+                    else:
+                        in_cell = abs(bbox & cell_rect) > 0.5 * abs(bbox)
+
+                    if in_cell:
                         span_text += this_char
                     elif this_char in white_spaces:
                         span_text += " "

tests/test_5185.py

@@ -0,0 +1,34 @@
+"""
+Regression test for issue #5185: table cell extraction dropped characters with
+an empty bbox, e.g. the second character of an expanded ligature when quad
+corrections are unset ("fine" became "fne").
+"""
+
+from pymupdf.table import extract_cells
+
+
+class _TextPage:
+    """Minimal stand-in for a TextPage, providing extractRAWDICT() only."""
+
+    def __init__(self, chars):
+        bbox = (0, 0, 500, 30)
+        span = {"bbox": bbox, "flags": 0, "char_flags": 0, "chars": chars}
+        line = {"bbox": bbox, "dir": (1, 0), "spans": [span]}
+        self.rawdict = {"blocks": [{"type": 0, "bbox": bbox, "lines": [line]}]}
+
+    def extractRAWDICT(self):
+        return self.rawdict
+
+
+def test_5185():
+    cell = (0, 0, 100, 30)
+    chars = [
+        {"c": "f", "bbox": (10, 10, 15, 20), "origin": (10, 18)},
+        # zero-width bbox of the expanded "fi" ligature, inside the cell
+        {"c": "i", "bbox": (15, 10, 15, 20), "origin": (15, 18)},
+        {"c": "n", "bbox": (15, 10, 20, 20), "origin": (15, 18)},
+        {"c": "e", "bbox": (20, 10, 25, 20), "origin": (20, 18)},
+        # zero-width char outside the cell must still be excluded
+        {"c": "x", "bbox": (150, 10, 150, 20), "origin": (150, 18)},
+    ]
+    assert extract_cells(_TextPage(chars), cell) == "fine"