Salvage text from a broken pdf file

salvage-pdf.py

#!/usr/bin/env python3
import re, zlib, sys

def main(argv):
    print("Salvage text from a broken pdf file")
    print("Usage: salvage-pdf.py [<input_file.pdf> [<output_file.txt>]]")
    pdf = argv[1] if len(argv) > 1 else sys.stdin.fileno()
    input = open(pdf, "rb").read()
    output = salvage_pdf(input)
    txt = argv[2] if len(argv) > 2 else sys.stdout.fileno()
    with open(txt, "wb") as file: file.write(output)

def salvage_pdf(input):
    stream = re.compile(rb'/FlateDecode.*?stream(.*?)endstream', re.S)
    streams = []
    for s in stream.findall(input):
        s = s.strip(b'\r\n')
        try:
            streams.append(zlib.decompress(s))
        except:
            pass # ignore decompression failures

    charmap = extract_charmap(streams)
    # print(charmap)

    text = [extract_text(data, charmap) for data in streams]
    output = [line for line in text if line] # remove empty lines
    return b"\n\n".join(output)

# https://stackoverflow.com/questions/40036588/in-pdf-if-encoding-and-tounicode-are-both-present-in-pdf-how-to-map-the-text-e
def extract_charmap(streams):
    charmap = {}
    # TODO: add support for beginbfrange\n(.+)\nendbfrange
    for data in streams:
        m = re.search(b"beginbfchar\n(.+?)\nendbfchar", data, re.DOTALL)
        if m:
            for l in m.group(1).split(b"\n"):
                pair = re.search(b'<(.+?)> *<(.+?)>',l)
                charmap[pair.group(1)] = pair.group(2)
    return charmap

# https://stackoverflow.com/questions/29467539/encoding-of-pdf-text-string
def extract_text(data, charmap):
    if not re.search(rb'BT[ \n]', data): return b''
    # extract lines ending with Tj or TJ
    lines = [line for line in data.split(b"\n") if (line[-2:]==b'Tj' or line[-2:]==b'TJ')]
    text = [extract_line(line, charmap) for line in lines]
    return b"\n".join(text)

def extract_line(line, charmap):
    l = decode_line(line, charmap)
    # l = re.sub(rb'\\222', '’'.encode('utf-8'), l)
    # l = re.sub(rb'\\225', '•'.encode('utf-8'), l)
    return re.sub(rb'\\([0-9]{3})', lambda m: num2char(m.group(1),8,'latin-1').decode('cp1252').encode('utf-8'), l)

def decode_line(line, charmap):
    if re.search(b'<.+>Tj|\[<.+>\]TJ', line):
        # remove everything except plain text inside brackets: <xxx>Tj or [<xxx>...<yyy>]TJ
        l = re.sub(b'^.*?<|>Tj|>\]TJ|>.*?<',b'',line)
        mapped = re.sub(b'.{4}', lambda m: (charmap.get(m.group()) or ''), l)
        string = re.sub(b'.{4}', lambda m: num2char(m.group()), mapped)
        return string
    # remove everything except plain text inside brackets: (xxx)Tj or [(xxx)...(yyy)]TJ
    l= re.sub(b'^.*?\(|\)Tj|\)\]TJ|\)[^)]*?\(', b'', line)
    # remove escape for brackets
    l= re.sub(rb'\\([\(\)])', rb'\1', l)
    return l

def num2char(hex, base=16, encoding='utf-8'):
    try:
        return chr(int(hex, base)).encode(encoding)
    except Exception as e:
        print(hex)
        print(e)
        return b'?'

if __name__ == "__main__": main(sys.argv)