Find all Chinese text in a string using Python and Regex

Question

I needed to strip the Chinese out of a bunch of strings today and was looking for a simple Python regex. Any suggestions?

Are you sure you want to remove Chinese, or do you really want to remove everything that is not Latin? — SingleNegationElimination, Apr 27 '10 at 02:04
Why would it be necessary (or useful) to remove Chinese characters from a string instead of translating them? — Anderson Green, Oct 25 '13 at 18:40

score 51 · Answer 1 · edited Jul 29 '19 at 16:47

51

Python 2:

#!/usr/bin/env python
# -*- encoding: utf8 -*-


import re

sample = u'I am from 美国。We should be friends. 朋友。'
for n in re.findall(ur'[\u4e00-\u9fff]+',sample):
    print n

Python 3:

sample = 'I am from 美国。We should be friends. 朋友。'
for n in re.findall(r'[\u4e00-\u9fff]+', sample):
    print(n)

Output:

美国
朋友

About Unicode code blocks:

The 4E00—9FFF range covers CJK Unified Ideographs (CJK=Chinese, Japanese and Korean). There are a number of lower ranges that relate, to some degree, to CJK:

31C0—31EF CJK Strokes
31F0—31FF Katakana Phonetic Extensions
3200—32FF Enclosed CJK Letters and Months
3300—33FF CJK Compatibility
3400—4DBF CJK Unified Ideographs Extension A
4DC0—4DFF Yijing Hexagram Symbols
4E00—9FFF CJK Unified Ideographs

edited Jul 29 '19 at 16:47

Brad Solomon

38,521
31
149
235

answered Apr 27 '10 at 01:36

prairiedogg

6,323
8
44
52

This will not work for all Chinese characters as some are surrogate pairs when UTF-16 encoded. (Since you are using \u4e00 and \u9fff it looks like you are UTF-16) – Stephen Nutt Apr 27 '10 at 01:39
@Stephen: this is true, but the Chinese characters outside the BMP are largely variant/historical forms that are not used in modern Chinese writing, so it's unlikely to matter. Other potential issues that Prairiedogg probably doesn't care about: as you can see in the above example, the code is extracting Han characters but is ignoring Chinese punctuation; it will also ignore various other Chinese symbols (circled characters, etc); and it will do strange and terrible things to Japanese text. – Porculus Apr 27 '10 at 02:01
Actually as I'm working through my data set, I'm thinking that TokenMacGuy is correct - I really want to strip everything that's non-Latin. – prairiedogg Apr 27 '10 at 03:05

cryo · Accepted Answer · 2010-04-27T02:22:12.540

The short, but relatively comprehensive answer for narrow Unicode builds of python (excluding ordinals > 65535 which can only be represented in narrow Unicode builds via surrogate pairs):

RE = re.compile(u'[⺀-⺙⺛-⻳⼀-⿕々〇〡-〩〸-〺〻㐀-䶵一-鿃豈-鶴侮-頻並-龎]', re.UNICODE)
nochinese = RE.sub('', mystring)

The code for building the RE, and if you need to detect Chinese characters in the supplementary plane for wide builds:

# -*- coding: utf-8 -*-
import re

LHan = [[0x2E80, 0x2E99],    # Han # So  [26] CJK RADICAL REPEAT, CJK RADICAL RAP
        [0x2E9B, 0x2EF3],    # Han # So  [89] CJK RADICAL CHOKE, CJK RADICAL C-SIMPLIFIED TURTLE
        [0x2F00, 0x2FD5],    # Han # So [214] KANGXI RADICAL ONE, KANGXI RADICAL FLUTE
        0x3005,              # Han # Lm       IDEOGRAPHIC ITERATION MARK
        0x3007,              # Han # Nl       IDEOGRAPHIC NUMBER ZERO
        [0x3021, 0x3029],    # Han # Nl   [9] HANGZHOU NUMERAL ONE, HANGZHOU NUMERAL NINE
        [0x3038, 0x303A],    # Han # Nl   [3] HANGZHOU NUMERAL TEN, HANGZHOU NUMERAL THIRTY
        0x303B,              # Han # Lm       VERTICAL IDEOGRAPHIC ITERATION MARK
        [0x3400, 0x4DB5],    # Han # Lo [6582] CJK UNIFIED IDEOGRAPH-3400, CJK UNIFIED IDEOGRAPH-4DB5
        [0x4E00, 0x9FC3],    # Han # Lo [20932] CJK UNIFIED IDEOGRAPH-4E00, CJK UNIFIED IDEOGRAPH-9FC3
        [0xF900, 0xFA2D],    # Han # Lo [302] CJK COMPATIBILITY IDEOGRAPH-F900, CJK COMPATIBILITY IDEOGRAPH-FA2D
        [0xFA30, 0xFA6A],    # Han # Lo  [59] CJK COMPATIBILITY IDEOGRAPH-FA30, CJK COMPATIBILITY IDEOGRAPH-FA6A
        [0xFA70, 0xFAD9],    # Han # Lo [106] CJK COMPATIBILITY IDEOGRAPH-FA70, CJK COMPATIBILITY IDEOGRAPH-FAD9
        [0x20000, 0x2A6D6],  # Han # Lo [42711] CJK UNIFIED IDEOGRAPH-20000, CJK UNIFIED IDEOGRAPH-2A6D6
        [0x2F800, 0x2FA1D]]  # Han # Lo [542] CJK COMPATIBILITY IDEOGRAPH-2F800, CJK COMPATIBILITY IDEOGRAPH-2FA1D

def build_re():
    L = []
    for i in LHan:
        if isinstance(i, list):
            f, t = i
            try: 
                f = unichr(f)
                t = unichr(t)
                L.append('%s-%s' % (f, t))
            except: 
                pass # A narrow python build, so can't use chars > 65535 without surrogate pairs!

        else:
            try:
                L.append(unichr(i))
            except:
                pass

    RE = '[%s]' % ''.join(L)
    print 'RE:', RE.encode('utf-8')
    return re.compile(RE, re.UNICODE)

RE = build_re()
print RE.sub('', u'美国').encode('utf-8')
print RE.sub('', u'blah').encode('utf-8')

Find all Chinese text in a string using Python and Regex

2 Answers2

Linked