关于正则表达式：在Python中按空格分隔字符串-保留带引号的子字符串

Split a string by spaces — preserving quoted substrings — in Python

我有一个像这样的字符串：

1	this is"a test"

我正在尝试在Python中编写一些内容，以按空格将其拆分，同时忽略引号内的空格。我正在寻找的结果是：

1	['this','is','a test']

PS。我知道您会问："如果引号内有引号，那么在我的应用程序中，那将永远不会发生。

您要从shlex模块拆分。

1
2
3

>>> import shlex
>>> shlex.split('this is"a test"')
['this', 'is', 'a test']

这应该正是您想要的。

看一下shlex模块，尤其是shlex.split。

1
2
3

>>> import shlex
>>> shlex.split('this is"a test"')
['this', 'is', 'a test']

我在这里看到正则表达式方法看起来很复杂和/或错误。这让我感到惊讶，因为正则表达式语法可以轻松地描述"空格或引号引起的东西"，并且大多数正则表达式引擎(包括Python的)都可以在正则表达式上拆分。因此，如果您要使用正则表达式，为什么不直接说出您的意思呢？：

1
2
3
4

test = 'this is"a test"' # or"this is 'a test'"
# pieces = [p for p in re.split("( |[\\"'].*[\\"'])", test) if p.strip()]
# From comments, use this:
pieces = [p for p in re.split("( |\\".*?\\"|'.*?')", test) if p.strip()]

说明：

1
2
3
4

[\\"'] = double-quote or single-quote
.* = anything
( |X) = space or X
.strip() = remove space and empty-string separators

shlex可能提供更多功能。

根据您的用例，您可能还需要检出csv模块：

1
2
3
4

import csv
lines = ['this is"a string"', 'and more"stuff"']
for row in csv.reader(lines, delimiter=""):
print row

输出：

1 2	['this', 'is', 'a string'] ['and', 'more', 'stuff']

我使用shlex.split来处理70,000,000行的鱿鱼日志，它是如此之慢。所以我改了。

如果shlex有性能问题，请尝试此操作。

1
2
3
4

import re

def line_split(line):
return re.findall(r'[^"\s]\S*|".+?"', line)

由于此问题是用正则表达式标记的，因此我决定尝试使用正则表达式方法。我首先用 x00替换引号部分中的所有空格，然后按空格分割，然后将 x00替换回每个部分中的空格。

两种版本都做同样的事情，但是splitter2比splitter2更具可读性。

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15

import re

s = 'this is"a test" some text"another test"'

def splitter(s):
def replacer(m):
return m.group(0).replace("","\x00")
parts = re.sub('".+?"', replacer, s).split()
parts = [p.replace("\x00","") for p in parts]
return parts

def splitter2(s):
return [p.replace("\x00","") for p in re.sub('".+?"', lambda m: m.group(0).replace("","\x00"), s).split()]

print splitter2(s)

出于性能原因，似乎re更快。这是我使用最小贪心运算符保留外引号的解决方案：

1	re.findall("(?:".*?"\|\S)+", s)

结果：

1	['this', 'is', '"a test"']

它将aaa"bla blub"bbb之类的构造保留在一起，因为这些标记之间没有空格。如果字符串包含转义字符，则可以这样进行匹配：

1
2
3
4
5
6
7
8

>>> a ="She said "He said, \\"My name is Mark.\\"""
>>> a
'She said"He said, \"My name is Mark.\""'
>>> for i in re.findall("(?:".*?[^\\\\]"|\S)+", a): print(i)
...
She
said
"He said, "My name is Mark.""

请注意，这也通过模式的\S部分匹配空字符串""。

接受的shlex方法的主要问题是它不会忽略带引号的子字符串之外的转义字符，并且在某些特殊情况下会产生一些意外的结果。

我有以下用例，在这里我需要一个拆分函数，该函数拆分输入字符串，以便保留单引号或双引号的子字符串，并能够在这样的子字符串中转义引号。无引号的字符串中的引号不应与其他任何字符区别对待。带有预期输出的一些示例测试用例：

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21

我最终得到了以下函数来拆分字符串，以便所有输入字符串的预期输出结果：

1
2
3
4
5
6
7
8
9

import re

def quoted_split(s):
def strip_quotes(s):
if s and (s[0] == '"' or s[0] =="'") and s[0] == s[-1]:
return s[1:-1]
return s
return [strip_quotes(p).replace('\"', '"').replace("\'","'") \
for p in re.findall(r'"(?:\\.|[^"])*"|\'(?:\\.|[^\'])*\'|[^\s]+', s)]

以下测试应用程序检查其他方法(现在为shlex和csv)和自定义拆分实现的结果：

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78

#!/bin/python2.7

import csv
import re
import shlex

from timeit import timeit

def test_case(fn, s, expected):
try:
if fn(s) == expected:
print '[ OK ] %s -> %s' % (s, fn(s))
else:
print '[FAIL] %s -> %s' % (s, fn(s))
except Exception as e:
print '[FAIL] %s -> exception: %s' % (s, e)

def test_case_no_output(fn, s, expected):
try:
fn(s)
except:
pass

def test_split(fn, test_case_fn=test_case):
test_case_fn(fn, 'abc def', ['abc', 'def'])
test_case_fn(fn,"abc \\s def", ['abc', '\\s', 'def'])
test_case_fn(fn, '"abc def" ghi', ['abc def', 'ghi'])
test_case_fn(fn,"'abc def' ghi", ['abc def', 'ghi'])
test_case_fn(fn, '"abc \" def" ghi', ['abc" def', 'ghi'])
test_case_fn(fn,"'abc \' def' ghi", ["abc ' def", 'ghi'])
test_case_fn(fn,"'abc \\s def' ghi", ['abc \\s def', 'ghi'])
test_case_fn(fn, '"abc \\s def" ghi', ['abc \\s def', 'ghi'])
test_case_fn(fn, '"" test', ['', 'test'])
test_case_fn(fn,"'' test", ['', 'test'])
test_case_fn(fn,"abc'def", ["abc'def"])
test_case_fn(fn,"abc'def'", ["abc'def'"])
test_case_fn(fn,"abc'def' ghi", ["abc'def'", 'ghi'])
test_case_fn(fn,"abc'def'ghi", ["abc'def'ghi"])
test_case_fn(fn, 'abc"def', ['abc"def'])
test_case_fn(fn, 'abc"def"', ['abc"def"'])
test_case_fn(fn, 'abc"def" ghi', ['abc"def"', 'ghi'])
test_case_fn(fn, 'abc"def"ghi', ['abc"def"ghi'])
test_case_fn(fn,"r'AA' r'.*_xyz$'", ["r'AA'","r'.*_xyz$'"])

def csv_split(s):
return list(csv.reader([s], delimiter=' '))[0]

def re_split(s):
def strip_quotes(s):
if s and (s[0] == '"' or s[0] =="'") and s[0] == s[-1]:
return s[1:-1]
return s
return [strip_quotes(p).replace('\"', '"').replace("\'","'") for p in re.findall(r'"(?:\\.|[^"])*"|\'(?:\\.|[^\'])*\'|[^\s]+', s)]

if __name__ == '__main__':
print 'shlex
'
test_split(shlex.split)
print

print 'csv
'
test_split(csv_split)
print

print 're
'
test_split(re_split)
print

iterations = 100
setup = 'from __main__ import test_split, test_case_no_output, csv_split, re_split
import shlex, re'
def benchmark(method, code):
print '%s: %.3fms per iteration' % (method, (1000 * timeit(code, setup=setup, number=iterations) / iterations))
benchmark('shlex', 'test_split(shlex.split, test_case_no_output)')
benchmark('csv', 'test_split(csv_split, test_case_no_output)')
benchmark('re', 'test_split(re_split, test_case_no_output)')

输出：

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69

shlex

[ OK ] abc def -> ['abc', 'def']
[FAIL] abc \s def -> ['abc', 's', 'def']
[ OK ]"abc def" ghi -> ['abc def', 'ghi']
[ OK ] 'abc def' ghi -> ['abc def', 'ghi']
[ OK ]"abc " def" ghi -> ['abc" def', 'ghi']
[FAIL] 'abc \' def' ghi -> exception: No closing quotation
[ OK ] 'abc \s def' ghi -> ['abc \\s def', 'ghi']
[ OK ]"abc \s def" ghi -> ['abc \\s def', 'ghi']
[ OK ]"" test -> ['', 'test']
[ OK ] '' test -> ['', 'test']
[FAIL] abc'def -> exception: No closing quotation
[FAIL] abc'def' -> ['abcdef']
[FAIL] abc'def' ghi -> ['abcdef', 'ghi']
[FAIL] abc'def'ghi -> ['abcdefghi']
[FAIL] abc"def -> exception: No closing quotation
[FAIL] abc"def" -> ['abcdef']
[FAIL] abc"def" ghi -> ['abcdef', 'ghi']
[FAIL] abc"def"ghi -> ['abcdefghi']
[FAIL] r'AA' r'.*_xyz$' -> ['rAA', 'r.*_xyz$']

csv

[ OK ] abc def -> ['abc', 'def']
[ OK ] abc \s def -> ['abc', '\\s', 'def']
[ OK ]"abc def" ghi -> ['abc def', 'ghi']
[FAIL] 'abc def' ghi -> ["'abc","def'", 'ghi']
[FAIL]"abc " def" ghi -> ['abc \', 'def"', 'ghi']
[FAIL] 'abc \' def' ghi -> ["'abc","\'","def'", 'ghi']
[FAIL] 'abc \s def' ghi -> ["'abc", '\\s',"def'", 'ghi']
[ OK ]"abc \s def" ghi -> ['abc \\s def', 'ghi']
[ OK ]"" test -> ['', 'test']
[FAIL] '' test -> ["''", 'test']
[ OK ] abc'def -> ["abc'def"]
[ OK ] abc'def' -> ["abc'def'"]
[ OK ] abc'def' ghi -> ["abc'def'", 'ghi']
[ OK ] abc'def'ghi -> ["abc'def'ghi"]
[ OK ] abc"def -> ['abc"def']
[ OK ] abc"def" -> ['abc"def"']
[ OK ] abc"def" ghi -> ['abc"def"', 'ghi']
[ OK ] abc"def"ghi -> ['abc"def"ghi']
[ OK ] r'AA' r'.*_xyz$' -> ["r'AA'","r'.*_xyz$'"]

re

[ OK ] abc def -> ['abc', 'def']
[ OK ] abc \s def -> ['abc', '\\s', 'def']
[ OK ]"abc def" ghi -> ['abc def', 'ghi']
[ OK ] 'abc def' ghi -> ['abc def', 'ghi']
[ OK ]"abc " def" ghi -> ['abc" def', 'ghi']
[ OK ] 'abc \' def' ghi -> ["abc ' def", 'ghi']
[ OK ] 'abc \s def' ghi -> ['abc \\s def', 'ghi']
[ OK ]"abc \s def" ghi -> ['abc \\s def', 'ghi']
[ OK ]"" test -> ['', 'test']
[ OK ] '' test -> ['', 'test']
[ OK ] abc'def -> ["abc'def"]
[ OK ] abc'def' -> ["abc'def'"]
[ OK ] abc'def' ghi -> ["abc'def'", 'ghi']
[ OK ] abc'def'ghi -> ["abc'def'ghi"]
[ OK ] abc"def -> ['abc"def']
[ OK ] abc"def" -> ['abc"def"']
[ OK ] abc"def" ghi -> ['abc"def"', 'ghi']
[ OK ] abc"def"ghi -> ['abc"def"ghi']
[ OK ] r'AA' r'.*_xyz$' -> ["r'AA'","r'.*_xyz$'"]

shlex: 0.281ms per iteration
csv: 0.030ms per iteration
re: 0.049ms per iteration

因此，性能要比shlex好得多，并且可以通过预编译正则表达式来进一步提高性能，在这种情况下，它的性能将优于csv方法。

要保留引号，请使用以下功能：

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18

def getArgs(s):
args = []
cur = ''
inQuotes = 0
for char in s.strip():
if char == ' ' and not inQuotes:
args.append(cur)
cur = ''
elif char == '"' and not inQuotes:
inQuotes = 1
cur += char
elif char == '"' and inQuotes:
inQuotes = 0
cur += char
else:
cur += char
args.append(cur)
return args

为了解决某些Python 2版本中的unicode问题，我建议：

1 2	from shlex import split as _split split = lambda a: [b.decode('utf-8') for b in _split(a.encode('utf-8'))]

速度测试的不同答案：

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18

import re
import shlex
import csv

line = 'this is"a test"'

%timeit [p for p in re.split("( |\\".*?\\"|'.*?')", line) if p.strip()]
100000 loops, best of 3: 5.17 μs per loop

%timeit re.findall(r'[^"\s]\S*|".+?"', line)
100000 loops, best of 3: 2.88 μs per loop

%timeit list(csv.reader([line], delimiter=""))
The slowest run took 9.62 times longer than the fastest. This could mean that an intermediate result is being cached.
100000 loops, best of 3: 2.4 μs per loop

%timeit shlex.split(line)
10000 loops, best of 3: 50.2 μs per loop

上面讨论的shlex的unicode问题(最佳答案)似乎在2.7.2+中已(间接)解决了
http://bugs.python.org/issue6988#msg146200

(因为我无法发表评论，所以单独回答)

嗯，似乎无法找到" Reply"按钮……无论如何，此答案基于Kate的方法，但是正确地将字符串与包含转义引号的子字符串分开，并且还删除了子字符串的开始和结束引号：

1	[i.strip('"').strip("'") for i in re.split(r'(\s+\|(?<!\\)".?(?<!\\)"\|(?<!\\)\'.?(?<!\\)\')', string) if i.strip()]

这适用于'This is" a \\\"test\\\"\\\'s substring"'之类的字符串(不幸的是，必须使用疯狂的标记来防止Python删除转义符)。

如果不需要返回列表中的字符串中的结果转义符，则可以使用此函数的稍有改动的版本：

1	[i.strip('"').strip("'").decode('string_escape') for i in re.split(r'(\s+\|(?<!\\)".?(?<!\\)"\|(?<!\\)\'.?(?<!\\)\')', string) if i.strip()]

我建议：

测试字符串：

1	s = 'abc"ad" \'fg\'"kk\'rdt\'" zzz"34"zzz"" \'\''

还要捕获"和"：

1 2	import re re.findall(r'"[^"]"\|\'[^\']\'\|[^"\'\s]+',s)

结果：

1	['abc', '"ad"',"'fg'", '"kk\'rdt\'"', 'zzz', '"34"', 'zzz', '""',"''"]

忽略空的"和"：

1 2	import re re.findall(r'"[^"]+"\|\'[^\']+\'\|[^"\'\s]+',s)

结果：

1	['abc', '"ad"',"'fg'", '"kk\'rdt\'"', 'zzz', '"34"', 'zzz']

如果您不关心子字符串，除了简单

1	>>> 'a short sized string with spaces '.split()

性能：

1
2
3
4

>>> s =" ('a short sized string with spaces '*100).split()"
>>> t = timeit.Timer(stmt=s)
>>> print"%.2f usec/pass" % (1000000 * t.timeit(number=100000)/100000)
171.39 usec/pass

或字符串模块

1 2	>>> from string import split as stringsplit; >>> stringsplit('a short sized string with spaces '*100)

性能：字符串模块似乎比字符串方法的性能更好

1
2
3
4

>>> s ="stringsplit('a short sized string with spaces '*100)"
>>> t = timeit.Timer(s,"from string import split as stringsplit")
>>> print"%.2f usec/pass" % (1000000 * t.timeit(number=100000)/100000)
154.88 usec/pass

或者您可以使用RE引擎

1
2
3
4

>>> from re import split as resplit
>>> regex = '\s+'
>>> medstring = 'a short sized string with spaces '*100
>>> resplit(regex, medstring)

性能

1
2
3
4

>>> s ="resplit(regex, medstring)"
>>> t = timeit.Timer(s,"from re import split as resplit; regex='\s+'; medstring='a short sized string with spaces '*100")
>>> print"%.2f usec/pass" % (1000000 * t.timeit(number=100000)/100000)
540.21 usec/pass

对于非常长的字符串，您不应将整个字符串加载到内存中，而应拆分行或使用迭代循环

尝试这个：

1
2
3
4
5
6
7
8
9
10

def adamsplit(s):
result = []
inquotes = False
for substring in s.split('"'):
if not inquotes:
result.extend(substring.split())
else:
result.append(substring)
inquotes = not inquotes
return result

一些测试字符串：