ZeroDivisionError: float division by zero

[pratheesh-prakash]

ZeroDivisionError encountered while extracting tables using camelot.

tables = camelot.read_pdf('page_39.pdf', pages='1', flavour='stream')

---------------------------------------------------------------------------
ZeroDivisionError                         Traceback (most recent call last)
Input In [244], in <cell line: 1>()
----> 1 tables = camelot.read_pdf(arabic_english, pages='1', flavour='stream')

File ~/.venv/lib/python3.8/site-packages/camelot/io.py:113, in read_pdf(filepath, pages, password, flavor, suppress_stdout, layout_kwargs, **kwargs)
    111 p = PDFHandler(filepath, pages=pages, password=password)
    112 kwargs = remove_extra(kwargs, flavor=flavor)
--> 113 tables = p.parse(
    114     flavor=flavor,
    115     suppress_stdout=suppress_stdout,
    116     layout_kwargs=layout_kwargs,
    117     **kwargs
    118 )
    119 return tables

File ~/.venv/lib/python3.8/site-packages/camelot/handlers.py:176, in PDFHandler.parse(self, flavor, suppress_stdout, layout_kwargs, **kwargs)
    174     parser = Lattice(**kwargs) if flavor == "lattice" else Stream(**kwargs)
    175     for p in pages:
--> 176         t = parser.extract_tables(
    177             p, suppress_stdout=suppress_stdout, layout_kwargs=layout_kwargs
    178         )
    179         tables.extend(t)
    180 return TableList(sorted(tables))

File ~/.venv/lib/python3.8/site-packages/camelot/parsers/lattice.py:430, in Lattice.extract_tables(self, filename, suppress_stdout, layout_kwargs)
    426 # sort tables based on y-coord
    427 for table_idx, tk in enumerate(
    428     sorted(self.table_bbox.keys(), key=lambda x: x[1], reverse=True)
    429 ):
--> 430     cols, rows, v_s, h_s = self._generate_columns_and_rows(table_idx, tk)
    431     table = self._generate_table(table_idx, cols, rows, v_s=v_s, h_s=h_s)
    432     table._bbox = tk

File ~/.venv/lib/python3.8/site-packages/camelot/parsers/lattice.py:322, in Lattice._generate_columns_and_rows(self, table_idx, tk)
    318 t_bbox = {}
    319 v_s, h_s = segments_in_bbox(
    320     tk, self.vertical_segments, self.horizontal_segments
    321 )
--> 322 t_bbox["horizontal"] = text_in_bbox(tk, self.horizontal_text)
    323 t_bbox["vertical"] = text_in_bbox(tk, self.vertical_text)
    325 t_bbox["horizontal"].sort(key=lambda x: (-x.y0, x.x0))

File ~/.venv/lib/python3.8/site-packages/camelot/utils.py:376, in text_in_bbox(bbox, text)
    373     continue
    374 if bbox_intersect(ba, bb):
    375     # if the intersection is larger than 80% of ba's size, we keep the longest
--> 376     if (bbox_intersection_area(ba, bb) / bbox_area(ba)) > 0.8:
    377         if bbox_longer(bb, ba):
    378             rest.discard(ba)

ZeroDivisionError: float division by zero

Steps to reproduce the bug

1. Download the PDF file. https://file.io/lVEXzype0AGL
2. Import camelot and issue the following expression: tables = camelot.read_pdf(file_path, pages='1', flavour='stream')

Expected behavior

Expected to parse the page and extract tables without any error.

Code

import camelot

tables = camelot.read_pdf('page_39.pdf', pages='1', flavour='stream')

PDF

PDF

Environment

• OS: GNU/Linux
• Python version: 3.8.10
• Numpy version: 1.22.3
• OpenCV version: 4.5.5.64
• Ghostscript version: 0.7
• Camelot version: 0.10.1

[kenho211]

When trying to avoid duplicating text by discarding overlapping bounding boxes, a threshold of 80% is used (intersection of bounding boxes over the first bounding box in the pair). In your sample pdf, there is a bounding box with area = 0. Adding a line to filter out textbox with 0 area in utils.text_in_bbox will do.

`def text_in_bbox(bbox, text): """Returns all text objects present inside a bounding box.

Parameters
----------
bbox : tuple
    Tuple (x1, y1, x2, y2) representing a bounding box where
    (x1, y1) -> lb and (x2, y2) -> rt in the PDF coordinate
    space.
text : List of PDFMiner text objects.

Returns
-------
t_bbox : list
    List of PDFMiner text objects that lie inside table, discarding the overlapping ones

"""
lb = (bbox[0], bbox[1])
rt = (bbox[2], bbox[3])
t_bbox = [
    t
    for t in text
    if lb[0] - 2 <= (t.x0 + t.x1) / 2.0 <= rt[0] + 2
    and lb[1] - 2 <= (t.y0 + t.y1) / 2.0 <= rt[1] + 2
    and bbox_area(t) > 0 # Filtering out bbox with area = 0
]


# Avoid duplicate text by discarding overlapping boxes
rest = {t for t in t_bbox}
for ba in t_bbox:
    for bb in rest.copy():
        if ba == bb:
            continue
        if bbox_intersect(ba, bb):
            # if the intersection is larger than 80% of ba's size, we keep the longest
            if (bbox_intersection_area(ba, bb) / bbox_area(ba)) > 0.8:
                if bbox_longer(bb, ba):
                    rest.discard(ba)
unique_boxes = list(rest)

return unique_boxes`