tokenize a string keeping delimiters in Python

Question

Is there any equivalent to str.split in Python that also returns the delimiters?

I need to preserve the whitespace layout for my output after processing some of the tokens.

Example:

>>> s="\tthis is an  example"
>>> print s.split()
['this', 'is', 'an', 'example']

>>> print what_I_want(s)
['\t', 'this', ' ', 'is', ' ', 'an', '  ', 'example']

Thanks!

Solution 1

How about

import re
splitter = re.compile(r'(\s+|\S+)')
splitter.findall(s)

Solution 2

>>> re.compile(r'(\s+)').split("\tthis is an  example")
['', '\t', 'this', ' ', 'is', ' ', 'an', '  ', 'example']

Solution 3

the re module provides this functionality:

>>> import re
>>> re.split('(\W+)', 'Words, words, words.')
['Words', ', ', 'words', ', ', 'words', '.', '']

(quoted from the Python documentation).

For your example (split on whitespace), use re.split('(\s+)', '\tThis is an example').

The key is to enclose the regex on which to split in capturing parentheses. That way, the delimiters are added to the list of results.

Edit: As pointed out, any preceding/trailing delimiters will of course also be added to the list. To avoid that you can use the .strip() method on your input string first.

Solution 4

Have you looked at pyparsing? Example borrowed from the pyparsing wiki:

>>> from pyparsing import Word, alphas
>>> greet = Word(alphas) + "," + Word(alphas) + "!"
>>> hello1 = 'Hello, World!'
>>> hello2 = 'Greetings, Earthlings!'
>>> for hello in hello1, hello2:
...     print (u'%s \u2192 %r' % (hello, greet.parseString(hello))).encode('utf-8')
... 
Hello, World! → (['Hello', ',', 'World', '!'], {})
Greetings, Earthlings! → (['Greetings', ',', 'Earthlings', '!'], {})

View more solutions

Comments

[Dominic Rodger]

+1 - Interesting question, splitlines seems to have a keepends parameter, but no such thing for split. Seems odd (docs.python.org/library/stdtypes.html#str.splitlines).

[Admin]

elegant and easily expandable (think (\s+|\w+|\S+)).

[Admin]

not using the OP's string masks the fact that the empty string is included as the first element of the returned list.

[Tim Pietzcker]

Thanks. I edited my post accordingly (although in this case, the OP's spec ("want to preserve whitespace") and his example were contradictory).

[ghostdog74]

no need regex or creating your own wheels if you have python 2.5 onwards.. see my answer.

[fortran]

No, it wasn't... there was one example of the current behaviour, and another of the desired one.