Splitting a sequence into pieces with identical elements

[candide]

Suppose you have a sequence s , a string for say, for instance this one :

spppammmmegggssss

We want to split s into the following parts :

['s', 'ppp', 'a', 'mmmm', 'e', 'ggg', 'ssss']

ie each part is a single repeated character word.

What is the pythonic way to answer this question?

A naive solution would be the following :


# -------------------------------
z='spppammmmegggssss'

zz=[]
while z:
k=1
while z[:k]==k*z[0]:
k+=1
zz+=[z[:k-1]]
z=z[k-1:]

print zz
# -------------------------------


but I guess this code is not very idiomatic

 

[Chris Rebert]

Suppose you have a sequence s , a string  for say, for instance this one :

spppammmmegggssss

We want to split s into the following parts :

['s', 'ppp', 'a', 'mmmm', 'e', 'ggg', 'ssss']

ie each part is a single repeated character word.

What is the pythonic way to answer this question?

If you're doing an operation on an iterable, always leaf thru itertools first:
http://docs.python.org/library/itertools.html

from itertools import groupby
def split_into_runs(seq):
return ["".join(run) for letter, run in groupby(seq)]


If itertools didn't exist:

def split_into_runs(seq):
if not seq: return []

iterator = iter(seq)
letter = next(iterator)
count = 1
words = []
for c in iterator:
if c == letter:
count += 1
else:
word = letter * count
words.append(word)
letter = c
count = 1
words.append(letter*count)
return words

Cheers,
Chris

 

[Tim Chase]

Suppose you have a sequence s , a string for say, for instance this one :

spppammmmegggssss

We want to split s into the following parts :

['s', 'ppp', 'a', 'mmmm', 'e', 'ggg', 'ssss']

ie each part is a single repeated character word.

While I'm not sure it's idiomatic, the overabuse of regexps in
Python certainly seems prevalent enough to be idiomatic ;-)

As such, you can use:

import re
r = re.compile(r'((.)\1*)')
#r = re.compile(r'((\w)\1*)')
s = 'spppammmmegggssss'
results = [m.group(0) for m in r.finditer(s)]

Additionally, you have all the properties of the match-object
(which includes the start/end) available too if you need).

You don't specify what you want to have happen with non-letters
(whitespace, punctuation, etc). The above just treats them like
any other character, finding repeats. If you just want "word"
characters, you can use the 2nd ("\w") version, or adjust
accordingly.

-tkc

 

[MRAB]

Suppose you have a sequence s , a string for say, for instance this
one :

spppammmmegggssss

We want to split s into the following parts :

['s', 'ppp', 'a', 'mmmm', 'e', 'ggg', 'ssss']

ie each part is a single repeated character word.

While I'm not sure it's idiomatic, the overabuse of regexps in Python
certainly seems prevalent enough to be idiomatic ;-)

As such, you can use:

import re
r = re.compile(r'((.)\1*)')
#r = re.compile(r'((\w)\1*)')

That should be \2, not \1.

Alternatively:

r = re.compile(r'(.)\1*')
#r = re.compile(r'(\w)\1*')

s = 'spppammmmegggssss'
results = [m.group(0) for m in r.finditer(s)]

Additionally, you have all the properties of the match-object (which
includes the start/end) available too if you need).

You don't specify what you want to have happen with non-letters
(whitespace, punctuation, etc). The above just treats them like any
other character, finding repeats. If you just want "word" characters,
you can use the 2nd ("\w") version, or adjust accordingly.

 

[Tim Chase]

That should be \2, not \1.

Alternatively:

r = re.compile(r'(.)\1*')

Doh, I had played with both and mis-transcribed the combination
of them into one malfunctioning regexp. My original trouble with
the 2nd one was that r.findall() (not .finditer) was only
returning the first letter of each because that's what was
matched. Wrapping it in the extra set of parens and using "\2"
returned the actual data in sub-tuples:

>>> s = 'spppammmmegggssss'
>>> import re
>>> r = re.compile(r'(.)\1*')
>>> r.findall(s) # no repeated text, just the initial letter ['s', 'p', 'a', 'm', 'e', 'g', 's']
>>> [m.group(0) for m in r.finditer(s)] ['s', 'ppp', 'a', 'mmmm', 'e', 'ggg', 'ssss']
>>> r = re.compile(r'((.)\2*)')
>>> r.findall(s)

[('s', 's'), ('ppp', 'p'), ('a', 'a'), ('mmmm', 'm'), ('e', 'e'),
('ggg', 'g'), ('ssss', 's')]

>>> [m.group(0) for m in r.finditer(s)]

['s', 'ppp', 'a', 'mmmm', 'e', 'ggg', 'ssss']

By then changing to .finditer() it made them both work the way I
wanted.

Thanks for catching my mistranscription.

-tkc