Is there any equivalent to split for arrays?
a = [1, 3, 4, 6, 8, 5, 3, 4, 5, 8, 4, 3]
separator = [3, 4] (len(separator) can be any)
b = a.split(separator)
b = [[1], [6, 8, 5], [5, 8, 4, 3]]
4 Answers
No, but we could write a function to do such a thing, and then if you need it to be an instance method you could subclass or encapsulate list.
def separate(array,separator):
results = []
a = array[:]
i = 0
while i<=len(a)-len(separator):
if a[i:i+len(separator)]==separator:
results.append(a[:i])
a = a[i+len(separator):]
i = 0
else: i+=1
results.append(a)
return results
If you wanted this to work as an instance method, we could do the following to encapsulate the list:
class SplitableList:
def __init__(self,ar): self.ary = ar
def split(self,sep): return separate(self.ary,sep)
# delegate other method calls to self.ary here, for example
def __len__(self): return len(self.ary)
a = SplitableList([1,3,4,6,8,5,3,4,5,8,4,3])
b = a.split([3,4]) # returns desired result
or we could subclass list like so:
class SplitableList(list):
def split(self,sep): return separate(self,sep)
a = SplitableList([1,3,4,6,8,5,3,4,5,8,4,3])
b = a.split([3,4]) # returns desired result
2 Comments
Yoav Glazner
I think it doesn't work for [1, 3,4], it will return [[1]] instead of [[1], []]
Matthew
You are right, and I wasn't sure if that empty list should occur or not (it should if the desire is to behave like the string split operator). I have edited the code to address that.
No there is not. You will have to write your own
or take this one:
def split(a, sep):
pos = i = 0
while i < len(a):
if a[i:i+len(sep)] == sep:
yield a[pos:i]
pos = i = i+len(sep)
else:
i += 1
yield a[pos:i]
print list(split(a, sep=[3, 4]))
Comments
You could join the list to a string using a non-numeric separator, then do the split:
>>> s = " {} ".format(" ".join(map(str, a)))
>>> s
' 1 3 4 6 8 5 3 4 5 8 4 3 '
>>> [[int(y) for y in x.split()] for x in s.split(" 3 4 ")]
[[1], [6, 8, 5], [5, 8, 4, 3]]
The 2 extra spaces in the string take account of edge cases (e.g. a = [1, 3, 4]).
Comments
For efficiency (50 x) on big arrays, there is a np.split method which can be used. the difficulty is to delete the separator :
from pylab import *
a=randint(0,3,10)
separator=arange(2)
ind=arange(len(a)-len(separator)+1) # splitting indexes
for i in range(len(separator)): ind=ind[a[ind]==separator[i]]+1 #select good candidates
cut=dstack((ind-len(separator),ind)).flatten() # begin and end
res=np.split(a,cut)[::2] # delete separators
print(a,cut,res)
gives:
[0 1 2 0 1 1 2 0 1 1] [0 2 3 5 7 9] [[],[2],[1, 2],[1]]
separatorin your invented call.itertools.groupby?