python regular expression to split paragraphs

Question

How would one write a regular expression to use in python to split paragraphs?

A paragraph is defined by 2 linebreaks (\n). But one can have any amount of spaces/tabs together with the line breaks, and it still should be considered as a paragraph.

I am using python so the solution can use python's regular expression syntax which is extended. (can make use of (?P...) stuff)

Examples:

the_str = 'paragraph1\n\nparagraph2'
# splitting should yield ['paragraph1', 'paragraph2']

the_str = 'p1\n\t\np2\t\n\tstill p2\t   \n     \n\tp3'
# should yield ['p1', 'p2\t\n\tstill p2', 'p3']

the_str = 'p1\n\n\n\tp2'
# should yield ['p1', '\n\tp2']

The best I could come with is: r'[ \t\r\f\v]*\n[ \t\r\f\v]*\n[ \t\r\f\v]*', i.e.

import re
paragraphs = re.split(r'[ \t\r\f\v]*\n[ \t\r\f\v]*\n[ \t\r\f\v]*', the_str)

but that is ugly. Anything better?

EDIT:

Suggestions rejected:

r'\s*?\n\s*?\n\s*?' -> That would make example 2 and 3 fail, since \s includes \n, so it would allow paragraph breaks with more than 2 \ns.

Answer 1

Unfortunately there's no nice way to write "space but not a newline".

I think the best you can do is add some space with the x modifier and try to factor out the ugliness a bit, but that's questionable: (?x) (?: [ \t\r\f\v]*? \n ){2} [ \t\r\f\v]*?

You could also try creating a subrule just for the character class and interpolating it three times.

Answer 2

Are you trying to deduce the structure of a document in plain test? Are you doing what docutils does?

You might be able to simply use the Docutils parser rather than roll your own.

Answer 3

Not a regexp but really elegant:

from itertools import groupby

def paragraph(lines) :
    for group_separator, line_iteration in groupby(lines.splitlines(True), key = str.isspace) :
        if not group_separator :
            yield ''.join(line_iteration)

for p in paragraph('p1\n\t\np2\t\n\tstill p2\t   \n     \n\tp'): 
    print repr(p)

'p1\n'
'p2\t\n\tstill p2\t   \n'
'\tp3'

It's up to you to strip the output as you need it of course.

Inspired from the famous "Python Cookbook" ;-)

Answer 4

Almost the same, but using non-greedy quantifiers and taking advantage of the whitespace sequence.

\s*?\n\s*?\n\s*?