textmatch

#!/usr/bin/env python3
# -*- coding: utf-8 -*-

import sys, os, argparse, difflib
from extras import ColoredArgParser
from extras import FileRead
from extras import colorize
from extras import trydecodingbytes
from extras import ColoredSetting

DEFAULT_MODULE = 'difflib'

def is_valid_file(parser, f):
    if f != '-' and not os.path.isfile(f):
        parser.error("%s: No such file" % f)
    else:
        return f

def parse_args():
    parser = ColoredArgParser(description = 'Compute text similarity matching using difflib.SequenceMatcher(built-in) and python-Levenshtein / FuzzyWuzzy if installed.')
    parser.add_argument('-o', '--original',
        type = lambda f: is_valid_file(parser, f),
        default = '-',
        help = 'The original file.')
    parser.add_argument('-d', '--duplicate',
        type = lambda f: is_valid_file(parser, f),
        default = '-',
        help = 'The duplicate file.')
    parser.add_argument('--original-encoding',
        dest = 'original_encoding',
        default = None,
        help = 'Text encoding of original.')
    parser.add_argument('--duplicate-encoding',
        dest = 'duplicate_encoding',
        default = None,
        help = 'Text encoding of duplicate.')
    parser.add_argument('-m', '--module',
        action = 'append',
        choices = ['difflib', 'python-Levenshtein', 'FuzzyWuzzy'],
        help = 'Specify which modules or libraries to used. Supported difflib, python-Levenshtein, FuzzyWuzzy. the default is difflib.')
    parser.add_argument('--color',
        choices = ['auto', 'always', 'never'],
        default = 'auto',
        dest = 'colored',
        help = 'When to colored, the default is auto.')
    return parser.parse_args()

def on_levenshtein(x, y):
    import Levenshtein
    return Levenshtein.ratio(x, y)

def on_fuzzywuzzy(x, y):
    from fuzzywuzzy import fuzz
    return fuzz.ratio(x, y)

def main(args = None):
    if args is None:
        return 0
    ColoredSetting(args.colored)
    readhook = (lambda f: ''.join(f.read().splitlines())) if args.original_encoding else (lambda f: b''.join(f.read().splitlines()))
    mode = 'rt' if args.original_encoding else 'rb'
    openhook = FileRead.hook_open(encoding = args.original_encoding)
    fr = FileRead(args.original, mode = mode, prompt_when_stdin = 'Enter the original(Press Ctrl-D when finished).', openhook = openhook, readhook = readhook)
    original_text = next(iter(fr))
    original_filename = fr.filename()
    if not args.original_encoding:
        original_text, args.original_encoding = trydecodingbytes(original_text)

    readhook = (lambda f: ''.join(f.read().splitlines())) if args.duplicate_encoding else (lambda f: b''.join(f.read().splitlines()))
    mode = 'rt' if args.duplicate_encoding else 'rb'
    openhook = FileRead.hook_open(encoding = args.duplicate_encoding)
    fr = FileRead(args.duplicate, mode = mode, prompt_when_stdin = 'Enter the duplicate(Press Ctrl-D when finished).', openhook = openhook, readhook = readhook)
    duplicate_text = next(iter(fr))
    duplicate_filename = fr.filename()
    if not args.duplicate_encoding:
        duplicate_text, args.duplicate_encoding = trydecodingbytes(duplicate_text)

    isjunk = lambda x: x.isspace()
    switcher = {
        'difflib' : lambda x, y: difflib.SequenceMatcher(isjunk = isjunk, a = x, b = y).ratio(),
        'python-Levenshtein' : on_levenshtein,
        'FuzzyWuzzy' : on_fuzzywuzzy,
    }
    for m in args.module or ( DEFAULT_MODULE, ):
        ratio = switcher.get(m)(original_text.strip(), duplicate_text.strip())
        prompt = 'The ratio after compared between %s(%s) and %s(%s) using %s is' % (
            repr(original_filename),
            args.original_encoding or 'Unknown',
            repr(duplicate_filename),
            args.duplicate_encoding or 'Unknown',
            m)
        print(colorize('%120s: %f' % (prompt, ratio),
            enabling = ColoredSetting().is_colorize(sys.stdout),
            fgcolor = 'green',
            set = ( 'bold', )))

if __name__ == '__main__':
    sys.exit(main(parse_args()))