Python 中 UTF-8 编码的坑：字符串多种表示方法

UTF-8 中定义了一些组合字符，这些字符会与它前面的非组合字符组合显示成一个字符，一般用它来添加加重或者变音标记。

同时呢，某些常用的加重字符也会有自己的单一编码值，这些字符叫做预组合字符（precomposed characters）。

这就带来一个很恐怖的后果，某些 UTF-8 的字符可能有两种表示方法！例如单词 naïve 可以写作这6个字符 nai\u0308ve ，也可能写作5个字符 na\u00EFve。这样一来，在程序中处理这类字符时就会出现一些很诡异的结果：

例如下面这段 python 代码

import re
s1 = "nai\u0308ve"
s2 = "na\u00EFve"

if s1 == s2:
    print(s1,"is equal to",s2)
else:
    print(s1,"is not equal to",s2)

regexp = '^.....$'
if re.match(regexp,s1):
    print(regexp,"is matching",s1)
else:
    print(regexp,"is not matching",s1)

if re.match(regexp,s2):
    print(regexp,"is matching",s2)
else:
    print(regexp,"is not matching",s2)

print("length of",s1,"is",len(s1))
print("length of",s2,"is",len(s2))

结果为：

naïve is not equal to naïve
^.....$ is not matching naïve
^.....$ is matching naïve
length of naïve is 6
length of naïve is 5

解决方法是用 unicodedata 库中的 normalize 函数来对字符串进行归一化（normalization）

import re
from unicodedata import normalize
s1 = normalize('NFC',"nai\u0308ve")
s2 = normalize('NFC',"na\u00EFve")

if s1 == s2:
    print(s1,"is equal to",s2)
else:
    print(s1,"is not equal to",s2)

regexp = '^.....$'
if re.match(regexp,s1):
    print(regexp,"is matching",s1)
else:
    print(regexp,"is not matching",s1)

if re.match(regexp,s2):
    print(regexp,"is matching",s2)
else:
    print(regexp,"is not matching",s2)

print("length of",s1,"is",len(s1))
print("length of",s2,"is",len(s2))

其结果为

naïve is equal to naïve
^.....$ is matching naïve
^.....$ is matching naïve
length of naïve is 5
length of naïve is 5