How do I extract a table from a pdf file using pymupdf

Question

My work demands I use pymupdf to extract tables from pdf files and export to csv format.

Answer 1

You are out of luck, I'm afraid. The PDF format has no internal representation of a table structure, which makes it difficult to extract tables for analysis. You have to infer the existence of a table by seeing where the columns of data have been lined up.

There are modules that will do this for you: one is Excalibur . But pymupdf is about extracting text as text and that will leave you having to do the parsing and inferencing yourself. That is a fairly ambitious project.

Answer 2

Excalibur is the GUI version of camelot

Installation https://camelot-py.readthedocs.io/en/master/user/install.html

Tutorial https://camelot-py.readthedocs.io/en/master/

This should do the job for you. There is direct export to csv. The default ouput is a dataframe which can be exported to excel or csv

Answer 3

I needed to work with pymupdf so I made a custom solution.

If it works for at least one person, I would be happy.

Please remember I did it for a specific problematic and the code may break for you.

I did it in an open way so you will be able to upgrade this code to your specific purpose.

If you make a better generalize solution, I would gladly use it, don't hesitate to post it. (this one took me 3 hours...)

import fitz  # this is pymupd, pip3 install PyMuPDF

# WARNING, this is a bad code, please use it knowing it may break easely
# Author: nah, I'm joking, nobody wants to own this shit XD


def get_page_bloc_tuple_2list(pdf_path):
    doc = fitz.open(pdf_path)
    page_bloc_tuple_2list = [page.get_text_words() for page in doc]  # for local (up to date)
    # page_bloc_tuple_2list = [page.getTextWords() for page in doc]  # for lambda (outdated)
    return page_bloc_tuple_2list


def get_line_dict_list(pdf_path):
    line_dict_list = []
    for page_block_list in get_page_bloc_tuple_2list(pdf_path):
        if len(page_block_list) == 0: continue
        word_dict_list = []
        y0_temp = page_block_list[0][1]
        y1_temp = page_block_list[0][3]
        for bloc in page_block_list:
            x0, y0, x1, y1, word, _, _, _ = bloc
            if y0 != y0_temp:
                line_dict = {
                    "y": (y0_temp, y1_temp),
                    "word_dict_list": word_dict_list
                }
                line_dict_list.append(line_dict)
                word_dict_list = []
                y0_temp = y0
                y1_temp = y1
            word_dict = {
                "word": word,
                "x": (x0, x1)
            }
            word_dict_list.append(word_dict)
    return line_dict_list


def get_word_list(line_dict):
    return [word_dict['word'] for word_dict in line_dict['word_dict_list']]


def is_title_line(line_dict, title_word_list):
    line_word_list = get_word_list(line_dict)
    for w in title_word_list:
        for sub_word in w.split(): # we need to comaprate substring
            if sub_word not in line_word_list:
                return False
    return True

def get_title_line(line_dict_list, title_word_list):
    for line_dict in line_dict_list:
        if is_title_line(line_dict, title_word_list):
            return line_dict
        

def get_word(title_x, line_dict):
    title_x0, title_x1 = title_x
    for word_dict in line_dict['word_dict_list']:
        word = word_dict['word']
        x0 = word_dict['x'][0]
        if title_x0 <= x0 and x0 <= title_x1:
           return word


def get_row_list(pdf_path, title_word_list):
    line_dict_list = get_line_dict_list(pdf_path)
    title_line = get_title_line(line_dict_list, title_word_list)
    row_list = [get_word_list(title_line)]
    title_index = line_dict_list.index(title_line)
    for idx, line_dict in enumerate(line_dict_list):
        if idx <= title_index: continue
        row = []
        for title_word_dict in title_line['word_dict_list']:
            row.append(get_word(title_word_dict['x'], line_dict))
        row_list.append(row)
    return row_list


if __name__ == "__main__":
    pdf_path = "my_filename.pdf"
    title_word_list = ["name", "surname", "whatever"]
    get_row_list(pdf_path, title_word_list)