How do I split a list into equally-sized chunks?

Asked 2023-09-20 20:21:08 View 267,139

How do I split a list of arbitrary length into equal sized chunks?


See also: How to iterate over a list in chunks.
To chunk strings, see Split string every nth character?.

  • Before you post a new answer, consider there are already 60+ answers for this question. Please, make sure that your answer contributes information that is not among existing answers. - anyone
  • This question has a pretty official answer from Python core developer Raymond Hettinger, which refers to the official docs: stackoverflow.com/a/74120449 - anyone

Answers

Here's a generator that yields evenly-sized chunks:

def chunks(lst, n):
    """Yield successive n-sized chunks from lst."""
    for i in range(0, len(lst), n):
        yield lst[i:i + n]
import pprint
pprint.pprint(list(chunks(range(10, 75), 10)))
[[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]]

For Python 2, using xrange instead of range:

def chunks(lst, n):
    """Yield successive n-sized chunks from lst."""
    for i in xrange(0, len(lst), n):
        yield lst[i:i + n]

Below is a list comprehension one-liner. The method above is preferable, though, since using named functions makes code easier to understand. For Python 3:

[lst[i:i + n] for i in range(0, len(lst), n)]

For Python 2:

[lst[i:i + n] for i in xrange(0, len(lst), n)]

Answered   2023-09-20 20:21:08

  • Your chunks method should be added to stdlib imho - anyone
  • @selle It's already in stdlib, it's called itertools.islice(iterator, chunk_size). - anyone
  • @ankostis islice does something different: it produces one slice of the iterator. - anyone
  • @NedBatchelder islice() needs a bit of boilerplate to setup a generator out of an iterator, but look how simple this solution is. - anyone

Something super simple:

def chunks(xs, n):
    n = max(1, n)
    return (xs[i:i+n] for i in range(0, len(xs), n))

For Python 2, use xrange() instead of range().

Answered   2023-09-20 20:21:08

I know this is kind of old but nobody yet mentioned numpy.array_split:

import numpy as np

lst = range(50)
np.array_split(lst, 5)

Result:

[array([0, 1, 2, 3, 4, 5, 6, 7, 8, 9]),
 array([10, 11, 12, 13, 14, 15, 16, 17, 18, 19]),
 array([20, 21, 22, 23, 24, 25, 26, 27, 28, 29]),
 array([30, 31, 32, 33, 34, 35, 36, 37, 38, 39]),
 array([40, 41, 42, 43, 44, 45, 46, 47, 48, 49])]

Answered   2023-09-20 20:21:08

  • This allows you to set the total number of chunks, not the number of elements per chunk. - anyone
  • This method change the type of the elements [ ['a', 1] , ['b', 2] ] with chunk one may become [ ['a', '1'] , ['b', '2'] ]. If type of first element is str then all element become numpy.str_ ... - anyone
  • It also breaks the lazyness of the iterable and needs O[2x] memory. - anyone
  • @FizxMike That problem can be solved using np.split(lst, np.arange(0, len(l), chunk_size)), althoug that requires even more memory and time. - anyone

Directly from the (old) Python documentation (recipes for itertools):

from itertools import izip, chain, repeat

def grouper(n, iterable, padvalue=None):
    "grouper(3, 'abcdefg', 'x') --> ('a','b','c'), ('d','e','f'), ('g','x','x')"
    return izip(*[chain(iterable, repeat(padvalue, n-1))]*n)

The current version, as suggested by J.F.Sebastian:

#from itertools import izip_longest as zip_longest # for Python 2.x
from itertools import zip_longest # for Python 3.x
#from six.moves import zip_longest # for both (uses the six compat library)

def grouper(n, iterable, padvalue=None):
    "grouper(3, 'abcdefg', 'x') --> ('a','b','c'), ('d','e','f'), ('g','x','x')"
    return zip_longest(*[iter(iterable)]*n, fillvalue=padvalue)

I guess Guido's time machine works—worked—will work—will have worked—was working again.

These solutions work because [iter(iterable)]*n (or the equivalent in the earlier version) creates one iterator, repeated n times in the list. izip_longest then effectively performs a round-robin of "each" iterator; because this is the same iterator, it is advanced by each such call, resulting in each such zip-roundrobin generating one tuple of n items.

Answered   2023-09-20 20:21:08

I'm surprised nobody has thought of using iter's two-argument form:

from itertools import islice

def chunk(it, size):
    it = iter(it)
    return iter(lambda: tuple(islice(it, size)), ())

Demo:

>>> list(chunk(range(14), 3))
[(0, 1, 2), (3, 4, 5), (6, 7, 8), (9, 10, 11), (12, 13)]

This works with any iterable and produces output lazily. It returns tuples rather than iterators, but I think it has a certain elegance nonetheless. It also doesn't pad; if you want padding, a simple variation on the above will suffice:

from itertools import islice, chain, repeat

def chunk_pad(it, size, padval=None):
    it = chain(iter(it), repeat(padval))
    return iter(lambda: tuple(islice(it, size)), (padval,) * size)

Demo:

>>> list(chunk_pad(range(14), 3))
[(0, 1, 2), (3, 4, 5), (6, 7, 8), (9, 10, 11), (12, 13, None)]
>>> list(chunk_pad(range(14), 3, 'a'))
[(0, 1, 2), (3, 4, 5), (6, 7, 8), (9, 10, 11), (12, 13, 'a')]

Like the izip_longest-based solutions, the above always pads. As far as I know, there's no one- or two-line itertools recipe for a function that optionally pads. By combining the above two approaches, this one comes pretty close:

_no_padding = object()

def chunk(it, size, padval=_no_padding):
    if padval == _no_padding:
        it = iter(it)
        sentinel = ()
    else:
        it = chain(iter(it), repeat(padval))
        sentinel = (padval,) * size
    return iter(lambda: tuple(islice(it, size)), sentinel)

Demo:

>>> list(chunk(range(14), 3))
[(0, 1, 2), (3, 4, 5), (6, 7, 8), (9, 10, 11), (12, 13)]
>>> list(chunk(range(14), 3, None))
[(0, 1, 2), (3, 4, 5), (6, 7, 8), (9, 10, 11), (12, 13, None)]
>>> list(chunk(range(14), 3, 'a'))
[(0, 1, 2), (3, 4, 5), (6, 7, 8), (9, 10, 11), (12, 13, 'a')]

I believe this is the shortest chunker proposed that offers optional padding.

As Tomasz Gandor observed, the two padding chunkers will stop unexpectedly if they encounter a long sequence of pad values. Here's a final variation that works around that problem in a reasonable way:

_no_padding = object()
def chunk(it, size, padval=_no_padding):
    it = iter(it)
    chunker = iter(lambda: tuple(islice(it, size)), ())
    if padval == _no_padding:
        yield from chunker
    else:
        for ch in chunker:
            yield ch if len(ch) == size else ch + (padval,) * (size - len(ch))

Demo:

>>> list(chunk([1, 2, (), (), 5], 2))
[(1, 2), ((), ()), (5,)]
>>> list(chunk([1, 2, None, None, 5], 2, None))
[(1, 2), (None, None), (5, None)]

Answered   2023-09-20 20:21:08

  • One-liner version: ``` from itertools import islice from functools import partial seq = [1,2,3,4,5,6,7] size = 3 result = list(iter(partial(lambda it: tuple(islice(it, size)), iter(seq)), ())) assert result == [(1, 2, 3), (4, 5, 6), (7,)] ``` - anyone

Here is a generator that work on arbitrary iterables:

def split_seq(iterable, size):
    it = iter(iterable)
    item = list(itertools.islice(it, size))
    while item:
        yield item
        item = list(itertools.islice(it, size))

Example:

>>> import pprint
>>> pprint.pprint(list(split_seq(xrange(75), 10)))
[[0, 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]]

Answered   2023-09-20 20:21:08

Don't reinvent the wheel.

UPDATE: A complete solution is found in Python 3.12+ itertools.batched.

Given

import itertools as it
import collections as ct

import more_itertools as mit


iterable = range(11)
n = 3

Code

itertools.batched++

list(it.batched(iterable, n))
# [[0, 1, 2], [3, 4, 5], [6, 7, 8], [9, 10]]

Details

The following non-native approaches were suggested prior to Python 3.12:

more_itertools+

list(mit.chunked(iterable, n))
# [[0, 1, 2], [3, 4, 5], [6, 7, 8], [9, 10]]

list(mit.sliced(iterable, n))
# [range(0, 3), range(3, 6), range(6, 9), range(9, 11)]

list(mit.grouper(n, iterable))
# [(0, 1, 2), (3, 4, 5), (6, 7, 8), (9, 10, None)]

list(mit.windowed(iterable, len(iterable)//n, step=n))
# [(0, 1, 2), (3, 4, 5), (6, 7, 8), (9, 10, None)]

list(mit.chunked_even(iterable, n))
# [[0, 1, 2], [3, 4, 5], [6, 7, 8], [9, 10]]

(or DIY, if you want)

The Standard Library

list(it.zip_longest(*[iter(iterable)] * n))
# [(0, 1, 2), (3, 4, 5), (6, 7, 8), (9, 10, None)]
d = {}
for i, x in enumerate(iterable):
    d.setdefault(i//n, []).append(x)
    

list(d.values())
# [[0, 1, 2], [3, 4, 5], [6, 7, 8], [9, 10]]
dd = ct.defaultdict(list)
for i, x in enumerate(iterable):
    dd[i//n].append(x)
    

list(dd.values())
# [[0, 1, 2], [3, 4, 5], [6, 7, 8], [9, 10]]

References

+ A third-party library that implements itertools recipes and more. > pip install more_itertools

++Included in Python Standard Library 3.12+. batched is similar to more_itertools.chunked.

Answered   2023-09-20 20:21:08

Simple yet elegant

L = range(1, 1000)
print [L[x:x+10] for x in xrange(0, len(L), 10)]

or if you prefer:

def chunks(L, n): return [L[x: x+n] for x in xrange(0, len(L), n)]
chunks(L, 10)

Answered   2023-09-20 20:21:08

How do you split a list into evenly sized chunks?

"Evenly sized chunks", to me, implies that they are all the same length, or barring that option, at minimal variance in length. E.g. 5 baskets for 21 items could have the following results:

>>> import statistics
>>> statistics.variance([5,5,5,5,1]) 
3.2
>>> statistics.variance([5,4,4,4,4]) 
0.19999999999999998

A practical reason to prefer the latter result: if you were using these functions to distribute work, you've built-in the prospect of one likely finishing well before the others, so it would sit around doing nothing while the others continued working hard.

Critique of other answers here

When I originally wrote this answer, none of the other answers were evenly sized chunks - they all leave a runt chunk at the end, so they're not well balanced, and have a higher than necessary variance of lengths.

For example, the current top answer ends with:

[60, 61, 62, 63, 64, 65, 66, 67, 68, 69],
[70, 71, 72, 73, 74]]

Others, like list(grouper(3, range(7))), and chunk(range(7), 3) both return: [(0, 1, 2), (3, 4, 5), (6, None, None)]. The None's are just padding, and rather inelegant in my opinion. They are NOT evenly chunking the iterables.

Why can't we divide these better?

Cycle Solution

A high-level balanced solution using itertools.cycle, which is the way I might do it today. Here's the setup:

from itertools import cycle
items = range(10, 75)
number_of_baskets = 10

Now we need our lists into which to populate the elements:

baskets = [[] for _ in range(number_of_baskets)]

Finally, we zip the elements we're going to allocate together with a cycle of the baskets until we run out of elements, which, semantically, it exactly what we want:

for element, basket in zip(items, cycle(baskets)):
    basket.append(element)

Here's the result:

>>> from pprint import pprint
>>> pprint(baskets)
[[10, 20, 30, 40, 50, 60, 70],
 [11, 21, 31, 41, 51, 61, 71],
 [12, 22, 32, 42, 52, 62, 72],
 [13, 23, 33, 43, 53, 63, 73],
 [14, 24, 34, 44, 54, 64, 74],
 [15, 25, 35, 45, 55, 65],
 [16, 26, 36, 46, 56, 66],
 [17, 27, 37, 47, 57, 67],
 [18, 28, 38, 48, 58, 68],
 [19, 29, 39, 49, 59, 69]]

To productionize this solution, we write a function, and provide the type annotations:

from itertools import cycle
from typing import List, Any

def cycle_baskets(items: List[Any], maxbaskets: int) -> List[List[Any]]:
    baskets = [[] for _ in range(min(maxbaskets, len(items)))]
    for item, basket in zip(items, cycle(baskets)):
        basket.append(item)
    return baskets

In the above, we take our list of items, and the max number of baskets. We create a list of empty lists, in which to append each element, in a round-robin style.

Slices

Another elegant solution is to use slices - specifically the less-commonly used step argument to slices. i.e.:

start = 0
stop = None
step = number_of_baskets

first_basket = items[start:stop:step]

This is especially elegant in that slices don't care how long the data are - the result, our first basket, is only as long as it needs to be. We'll only need to increment the starting point for each basket.

In fact this could be a one-liner, but we'll go multiline for readability and to avoid an overlong line of code:

from typing import List, Any

def slice_baskets(items: List[Any], maxbaskets: int) -> List[List[Any]]:
    n_baskets = min(maxbaskets, len(items))
    return [items[i::n_baskets] for i in range(n_baskets)]

And islice from the itertools module will provide a lazily iterating approach, like that which was originally asked for in the question.

I don't expect most use-cases to benefit very much, as the original data is already fully materialized in a list, but for large datasets, it could save nearly half the memory usage.

from itertools import islice
from typing import List, Any, Generator
    
def yield_islice_baskets(items: List[Any], maxbaskets: int) -> Generator[List[Any], None, None]:
    n_baskets = min(maxbaskets, len(items))
    for i in range(n_baskets):
        yield islice(items, i, None, n_baskets)

View results with:

from pprint import pprint

items = list(range(10, 75))
pprint(cycle_baskets(items, 10))
pprint(slice_baskets(items, 10))
pprint([list(s) for s in yield_islice_baskets(items, 10)])

Updated prior solutions

Here's another balanced solution, adapted from a function I've used in production in the past, that uses the modulo operator:

def baskets_from(items, maxbaskets=25):
    baskets = [[] for _ in range(maxbaskets)]
    for i, item in enumerate(items):
        baskets[i % maxbaskets].append(item)
    return filter(None, baskets) 

And I created a generator that does the same if you put it into a list:

def iter_baskets_from(items, maxbaskets=3):
    '''generates evenly balanced baskets from indexable iterable'''
    item_count = len(items)
    baskets = min(item_count, maxbaskets)
    for x_i in range(baskets):
        yield [items[y_i] for y_i in range(x_i, item_count, baskets)]
    

And finally, since I see that all of the above functions return elements in a contiguous order (as they were given):

def iter_baskets_contiguous(items, maxbaskets=3, item_count=None):
    '''
    generates balanced baskets from iterable, contiguous contents
    provide item_count if providing a iterator that doesn't support len()
    '''
    item_count = item_count or len(items)
    baskets = min(item_count, maxbaskets)
    items = iter(items)
    floor = item_count // baskets 
    ceiling = floor + 1
    stepdown = item_count % baskets
    for x_i in range(baskets):
        length = ceiling if x_i < stepdown else floor
        yield [items.next() for _ in range(length)]

Output

To test them out:

print(baskets_from(range(6), 8))
print(list(iter_baskets_from(range(6), 8)))
print(list(iter_baskets_contiguous(range(6), 8)))
print(baskets_from(range(22), 8))
print(list(iter_baskets_from(range(22), 8)))
print(list(iter_baskets_contiguous(range(22), 8)))
print(baskets_from('ABCDEFG', 3))
print(list(iter_baskets_from('ABCDEFG', 3)))
print(list(iter_baskets_contiguous('ABCDEFG', 3)))
print(baskets_from(range(26), 5))
print(list(iter_baskets_from(range(26), 5)))
print(list(iter_baskets_contiguous(range(26), 5)))

Which prints out:

[[0], [1], [2], [3], [4], [5]]
[[0], [1], [2], [3], [4], [5]]
[[0], [1], [2], [3], [4], [5]]
[[0, 8, 16], [1, 9, 17], [2, 10, 18], [3, 11, 19], [4, 12, 20], [5, 13, 21], [6, 14], [7, 15]]
[[0, 8, 16], [1, 9, 17], [2, 10, 18], [3, 11, 19], [4, 12, 20], [5, 13, 21], [6, 14], [7, 15]]
[[0, 1, 2], [3, 4, 5], [6, 7, 8], [9, 10, 11], [12, 13, 14], [15, 16, 17], [18, 19], [20, 21]]
[['A', 'D', 'G'], ['B', 'E'], ['C', 'F']]
[['A', 'D', 'G'], ['B', 'E'], ['C', 'F']]
[['A', 'B', 'C'], ['D', 'E'], ['F', 'G']]
[[0, 5, 10, 15, 20, 25], [1, 6, 11, 16, 21], [2, 7, 12, 17, 22], [3, 8, 13, 18, 23], [4, 9, 14, 19, 24]]
[[0, 5, 10, 15, 20, 25], [1, 6, 11, 16, 21], [2, 7, 12, 17, 22], [3, 8, 13, 18, 23], [4, 9, 14, 19, 24]]
[[0, 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]]

Notice that the contiguous generator provide chunks in the same length patterns as the other two, but the items are all in order, and they are as evenly divided as one may divide a list of discrete elements.

Answered   2023-09-20 20:21:08

def chunk(input, size):
    return map(None, *([iter(input)] * size))

Answered   2023-09-20 20:21:08

  • Doesn't work in Python 3.8, is that for 2.x? - anyone
  • For Python 3.x: return map(lambda *x: x, *([iter(input)] * size)). Yet it drops tail of the list if it cannot be divided in the equal chunks - anyone

If you know list size:

def SplitList(mylist, chunk_size):
    return [mylist[offs:offs+chunk_size] for offs in range(0, len(mylist), chunk_size)]

If you don't (an iterator):

def IterChunks(sequence, chunk_size):
    res = []
    for item in sequence:
        res.append(item)
        if len(res) >= chunk_size:
            yield res
            res = []
    if res:
        yield res  # yield the last, incomplete, portion

In the latter case, it can be rephrased in a more beautiful way if you can be sure that the sequence always contains a whole number of chunks of given size (i.e. there is no incomplete last chunk).

Answered   2023-09-20 20:21:08

I saw the most awesome Python-ish answer in a duplicate of this question:

from itertools import zip_longest

a = range(1, 16)
i = iter(a)
r = list(zip_longest(i, i, i))
>>> print(r)
[(1, 2, 3), (4, 5, 6), (7, 8, 9), (10, 11, 12), (13, 14, 15)]

You can create n-tuple for any n. If a = range(1, 15), then the result will be:

[(1, 2, 3), (4, 5, 6), (7, 8, 9), (10, 11, 12), (13, 14, None)]

If the list is divided evenly, then you can replace zip_longest with zip, otherwise the triplet (13, 14, None) would be lost. Python 3 is used above. For Python 2, use izip_longest.

Answered   2023-09-20 20:21:08

Here's the one liner:

[AA[i:i+SS] for i in range(len(AA))[::SS]]

Details. AA is array, SS is chunk size. For example:

>>> AA=range(10,21);SS=3
>>> [AA[i:i+SS] for i in range(len(AA))[::SS]]
[[10, 11, 12], [13, 14, 15], [16, 17, 18], [19, 20]]
# or [range(10, 13), range(13, 16), range(16, 19), range(19, 21)] in py3

To expand the ranges in py3 do

(py3) >>> [list(AA[i:i+SS]) for i in range(len(AA))[::SS]]
[[10, 11, 12], [13, 14, 15], [16, 17, 18], [19, 20]]

Answered   2023-09-20 20:21:08

With Assignment Expressions in Python 3.8 it becomes quite nice:

import itertools

def batch(iterable, size):
    it = iter(iterable)
    while item := list(itertools.islice(it, size)):
        yield item

This works on an arbitrary iterable, not just a list.

>>> import pprint
>>> pprint.pprint(list(batch(range(75), 10)))
[[0, 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]]

UPDATE

Starting with Python 3.12, this exact implementation is available as itertools.batched

Answered   2023-09-20 20:21:08

If you had a chunk size of 3 for example, you could do:

zip(*[iterable[i::3] for i in range(3)]) 

source: http://code.activestate.com/recipes/303060-group-a-list-into-sequential-n-tuples/

I would use this when my chunk size is fixed number I can type, e.g. '3', and would never change.

Answered   2023-09-20 20:21:08

The toolz library has the partition function for this:

from toolz.itertoolz.core import partition

list(partition(2, [1, 2, 3, 4]))
[(1, 2), (3, 4)]

Answered   2023-09-20 20:21:08

I was curious about the performance of different approaches and here it is:

Tested on Python 3.5.1

import time
batch_size = 7
arr_len = 298937

#---------slice-------------

print("\r\nslice")
start = time.time()
arr = [i for i in range(0, arr_len)]
while True:
    if not arr:
        break

    tmp = arr[0:batch_size]
    arr = arr[batch_size:-1]
print(time.time() - start)

#-----------index-----------

print("\r\nindex")
arr = [i for i in range(0, arr_len)]
start = time.time()
for i in range(0, round(len(arr) / batch_size + 1)):
    tmp = arr[batch_size * i : batch_size * (i + 1)]
print(time.time() - start)

#----------batches 1------------

def batch(iterable, n=1):
    l = len(iterable)
    for ndx in range(0, l, n):
        yield iterable[ndx:min(ndx + n, l)]

print("\r\nbatches 1")
arr = [i for i in range(0, arr_len)]
start = time.time()
for x in batch(arr, batch_size):
    tmp = x
print(time.time() - start)

#----------batches 2------------

from itertools import islice, chain

def batch(iterable, size):
    sourceiter = iter(iterable)
    while True:
        batchiter = islice(sourceiter, size)
        yield chain([next(batchiter)], batchiter)


print("\r\nbatches 2")
arr = [i for i in range(0, arr_len)]
start = time.time()
for x in batch(arr, batch_size):
    tmp = x
print(time.time() - start)

#---------chunks-------------
def chunks(l, n):
    """Yield successive n-sized chunks from l."""
    for i in range(0, len(l), n):
        yield l[i:i + n]
print("\r\nchunks")
arr = [i for i in range(0, arr_len)]
start = time.time()
for x in chunks(arr, batch_size):
    tmp = x
print(time.time() - start)

#-----------grouper-----------

from itertools import zip_longest # for Python 3.x
#from six.moves import zip_longest # for both (uses the six compat library)

def grouper(iterable, n, padvalue=None):
    "grouper(3, 'abcdefg', 'x') --> ('a','b','c'), ('d','e','f'), ('g','x','x')"
    return zip_longest(*[iter(iterable)]*n, fillvalue=padvalue)

arr = [i for i in range(0, arr_len)]
print("\r\ngrouper")
start = time.time()
for x in grouper(arr, batch_size):
    tmp = x
print(time.time() - start)

Results:

slice
31.18285083770752

index
0.02184295654296875

batches 1
0.03503894805908203

batches 2
0.22681021690368652

chunks
0.019841909408569336

grouper
0.006506919860839844

Answered   2023-09-20 20:21:08

You may also use get_chunks function of utilspie library as:

>>> from utilspie import iterutils
>>> a = [1, 2, 3, 4, 5, 6, 7, 8, 9]

>>> list(iterutils.get_chunks(a, 5))
[[1, 2, 3, 4, 5], [6, 7, 8, 9]]

You can install utilspie via pip:

sudo pip install utilspie

Disclaimer: I am the creator of utilspie library.

Answered   2023-09-20 20:21:08

I like the Python doc's version proposed by tzot and J.F.Sebastian a lot, but it has two shortcomings:

  • it is not very explicit
  • I usually don't want a fill value in the last chunk

I'm using this one a lot in my code:

from itertools import islice

def chunks(n, iterable):
    iterable = iter(iterable)
    while True:
        yield tuple(islice(iterable, n)) or iterable.next()

UPDATE: A lazy chunks version:

from itertools import chain, islice

def chunks(n, iterable):
   iterable = iter(iterable)
   while True:
       yield chain([next(iterable)], islice(iterable, n-1))

Answered   2023-09-20 20:21:08

code:

def split_list(the_list, chunk_size):
    result_list = []
    while the_list:
        result_list.append(the_list[:chunk_size])
        the_list = the_list[chunk_size:]
    return result_list

a_list = [1, 2, 3, 4, 5, 6, 7, 8, 9, 10]

print split_list(a_list, 3)

result:

[[1, 2, 3], [4, 5, 6], [7, 8, 9], [10]]

Answered   2023-09-20 20:21:08

heh, one line version

In [48]: chunk = lambda ulist, step:  map(lambda i: ulist[i:i+step],  xrange(0, len(ulist), step))

In [49]: chunk(range(1,100), 10)
Out[49]: 
[[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, 79, 80],
 [81, 82, 83, 84, 85, 86, 87, 88, 89, 90],
 [91, 92, 93, 94, 95, 96, 97, 98, 99]]

Answered   2023-09-20 20:21:08

  • Please, use "def chunk" instead of "chunk = lambda". It works the same. One line. Same features. MUCH easier to the n00bz to read and understand. - anyone
  • The function object resulting from def chunk instead of chunk=lambda has .__name__ attribute 'chunk' instead of '<lambda>'. The specific name is more useful in tracebacks. - anyone

Another more explicit version.

def chunkList(initialList, chunkSize):
    """
    This function chunks a list into sub lists 
    that have a length equals to chunkSize.

    Example:
    lst = [3, 4, 9, 7, 1, 1, 2, 3]
    print(chunkList(lst, 3)) 
    returns
    [[3, 4, 9], [7, 1, 1], [2, 3]]
    """
    finalList = []
    for i in range(0, len(initialList), chunkSize):
        finalList.append(initialList[i:i+chunkSize])
    return finalList

Answered   2023-09-20 20:21:08

At this point, I think we need a recursive generator, just in case...

In python 2:

def chunks(li, n):
    if li == []:
        return
    yield li[:n]
    for e in chunks(li[n:], n):
        yield e

In python 3:

def chunks(li, n):
    if li == []:
        return
    yield li[:n]
    yield from chunks(li[n:], n)

Also, in case of massive Alien invasion, a decorated recursive generator might become handy:

def dec(gen):
    def new_gen(li, n):
        for e in gen(li, n):
            if e == []:
                return
            yield e
    return new_gen

@dec
def chunks(li, n):
    yield li[:n]
    for e in chunks(li[n:], n):
        yield e

Answered   2023-09-20 20:21:08

Without calling len() which is good for large lists:

def splitter(l, n):
    i = 0
    chunk = l[:n]
    while chunk:
        yield chunk
        i += n
        chunk = l[i:i+n]

And this is for iterables:

def isplitter(l, n):
    l = iter(l)
    chunk = list(islice(l, n))
    while chunk:
        yield chunk
        chunk = list(islice(l, n))

The functional flavour of the above:

def isplitter2(l, n):
    return takewhile(bool,
                     (tuple(islice(start, n))
                            for start in repeat(iter(l))))

OR:

def chunks_gen_sentinel(n, seq):
    continuous_slices = imap(islice, repeat(iter(seq)), repeat(0), repeat(n))
    return iter(imap(tuple, continuous_slices).next,())

OR:

def chunks_gen_filter(n, seq):
    continuous_slices = imap(islice, repeat(iter(seq)), repeat(0), repeat(n))
    return takewhile(bool,imap(tuple, continuous_slices))

Answered   2023-09-20 20:21:08

  • There is no reason to avoid len() on large lists; it's a constant-time operation. - anyone
def split_seq(seq, num_pieces):
    start = 0
    for i in xrange(num_pieces):
        stop = start + len(seq[i::num_pieces])
        yield seq[start:stop]
        start = stop

usage:

seq = [1, 2, 3, 4, 5, 6, 7, 8, 9, 10]

for seq in split_seq(seq, 3):
    print seq

Answered   2023-09-20 20:21:08

See this reference

>>> orange = range(1, 1001)
>>> otuples = list( zip(*[iter(orange)]*10))
>>> print(otuples)
[(1, 2, 3, 4, 5, 6, 7, 8, 9, 10), ... (991, 992, 993, 994, 995, 996, 997, 998, 999, 1000)]
>>> olist = [list(i) for i in otuples]
>>> print(olist)
[[1, 2, 3, 4, 5, 6, 7, 8, 9, 10], ..., [991, 992, 993, 994, 995, 996, 997, 998, 999, 1000]]
>>> 

Python3

Answered   2023-09-20 20:21:08

  • Nice, but drops elements at the end if the size does not match whole numbers of chunks, e. g. zip(*[iter(range(7))]*3) only returns [(0, 1, 2), (3, 4, 5)] and forgets the 6 from the input. - anyone
  • See this other Stack Overflow question for a detailed explanation of this technique. - anyone
def chunks(iterable,n):
    """assumes n is an integer>0
    """
    iterable=iter(iterable)
    while True:
        result=[]
        for i in range(n):
            try:
                a=next(iterable)
            except StopIteration:
                break
            else:
                result.append(a)
        if result:
            yield result
        else:
            break

g1=(i*i for i in range(10))
g2=chunks(g1,3)
print g2
'<generator object chunks at 0x0337B9B8>'
print list(g2)
'[[0, 1, 4], [9, 16, 25], [36, 49, 64], [81]]'

Answered   2023-09-20 20:21:08

Since everybody here talking about iterators. boltons has perfect method for that, called iterutils.chunked_iter.

from boltons import iterutils

list(iterutils.chunked_iter(list(range(50)), 11))

Output:

[[0, 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]]

But if you don't want to be mercy on memory, you can use old-way and store the full list in the first place with iterutils.chunked.

Answered   2023-09-20 20:21:08

Consider using matplotlib.cbook pieces

for example:

import matplotlib.cbook as cbook
segments = cbook.pieces(np.arange(20), 3)
for s in segments:
     print s

Answered   2023-09-20 20:21:08

a = [1, 2, 3, 4, 5, 6, 7, 8, 9]
CHUNK = 4
[a[i*CHUNK:(i+1)*CHUNK] for i in xrange((len(a) + CHUNK - 1) / CHUNK )]

Answered   2023-09-20 20:21:08

  • Can you explain more your answer please ? - anyone
  • Working from backwards: (len(a) + CHUNK -1) / CHUNK Gives you the number of chunks that you will end up with. Then, for each chunk at index i, we are generating a sub-array of the original array like this: a[ i * CHUNK : (i + 1) * CHUNK ] where, i * CHUNK is the index of the first element to put into the subarray, and, (i + 1) * CHUNK is 1 past the last element to put into the subarray. This solution uses list comprehension, so it might be faster for large arrays. - anyone