Getting a sublist of a Python list, with the given indices?

Using List Comprehension ,this should work –

li = [a[i] for i in b]

Testing this –

>>> a = [0,10,20,30,40,50,60]
>>> b = [0,2,4,5]
>>> li = [a[i] for i in b]
>>> li
[0, 20, 40, 50]

Something different…

>>> a = range(7)
>>> b = [0,2,4,5]
>>> import operator
>>> operator.itemgetter(*b)(a)
(0, 2, 4, 5)

The itemgetter function takes one or more keys as arguments, and returns a function which will return the items at the given keys in its argument. So in the above, we create a function which will return the items at index 0, index 2, index 4, and index 5, then apply that function to a.

It appears to be quite a bit faster than the equivalent list comprehension

In [1]: import operator

In [2]: a = range(7)

In [3]: b = [0,2,4,5]

In [4]: %timeit operator.itemgetter(*b)(a)
1000000 loops, best of 3: 388 ns per loop

In [5]: %timeit [ a[i] for i in b ]
1000000 loops, best of 3: 415 ns per loop

In [6]: f = operator.itemgetter(*b)

In [7]: %timeit f(a)
10000000 loops, best of 3: 183 ns per loop

As for why itemgetter is faster, the comprehension has to execute extra Python byte codes.

In [3]: def f(a,b): return [a[i] for i in b]

In [4]: def g(a,b): return operator.itemgetter(*b)(a)

In [5]: dis.dis(f)
  1           0 BUILD_LIST               0
              3 LOAD_FAST                1 (b)
              6 GET_ITER
        >>    7 FOR_ITER                16 (to 26)
             10 STORE_FAST               2 (i)
             13 LOAD_FAST                0 (a)
             16 LOAD_FAST                2 (i)
             19 BINARY_SUBSCR
             20 LIST_APPEND              2
             23 JUMP_ABSOLUTE            7
        >>   26 RETURN_VALUE

While itemgetter is a single call implemented in C:

In [6]: dis.dis(g)
  1           0 LOAD_GLOBAL              0 (operator)
              3 LOAD_ATTR                1 (itemgetter)
              6 LOAD_FAST                1 (b)
              9 CALL_FUNCTION_VAR        0
             12 LOAD_FAST                0 (a)
             15 CALL_FUNCTION            1
             18 RETURN_VALUE

Many of the proposed solutions will produce a KeyError if b contains an index not present in a. The following will skip invalid indices if that is desired.

>>> b = [0,2,4,5]
>>> a = [0,1,2,3,4,5,6]
>>> [x for i,x in enumerate(a) if i in b]
[0, 2, 4, 5]
>>> b = [0,2,4,500]
>>> [x for i,x in enumerate(a) if i in b]
[0, 2, 4]

enumerate produces tuples of index,value pairs. Since we have both the item and its index, we can check for the presence of the index in b

If you are a fan of functional programming, you could use map and list.__getitem__:

>>> a = [0,1,2,3,4,5,6]
>>> b = [0,2,4,5]
>>> map(a.__getitem__, b)
[0, 2, 4, 5]
>>>

The list comprehension approach is more canonical in Python though…

Yet another alternative for better performance if that is important to you- it is by no means the most Pythonic but I am pretty sure it is the most efficient:

>>> list(filter(lambda x: a.index(x) in b, a))
[0, 2, 4, 5]

Note: You do not need to convert to a list in Python 2. However you do in Python 3 onwards (if any future visitors may have a similar issue).

A bit of speed comparison for all mentioned methods and others from Python dictionary: Get list of values for list of keys:

Python 2.7.11 |Anaconda 2.4.1 (64-bit)| (default, Jan 19 2016, 12:08:31) [MSC v.1500 64 bit (AMD64)] on win32

In[2]: import numpy.random as nprnd
idx = nprnd.randint(1000, size=10000)
l = nprnd.rand(1000).tolist()
from operator import itemgetter
import operator
f = operator.itemgetter(*idx)
%timeit f(l)
%timeit list(itemgetter(*idx)(l))
%timeit [l[_] for _ in idx]  # list comprehension
%timeit map(l.__getitem__, idx)
%timeit list(l[_] for _ in idx)  # a generator expression passed to a list constructor.
%timeit map(lambda _: l[_], idx)  # using 'map'
%timeit [x for i, x in enumerate(l) if i in idx]
%timeit filter(lambda x: l.index(x) in idx, l)  # UPDATE @Kundor: work only for list with unique elements
10000 loops, best of 3: 175 µs per loop
1000 loops, best of 3: 707 µs per loop
1000 loops, best of 3: 978 µs per loop
1000 loops, best of 3: 1.03 ms per loop
1000 loops, best of 3: 1.18 ms per loop
1000 loops, best of 3: 1.86 ms per loop
100 loops, best of 3: 12.3 ms per loop
10 loops, best of 3: 21.2 ms per loop

So the fastest is f = operator.itemgetter(*idx); f(l)

Using numpy.asarray. Numpy allows getting subarray of array by list of indices.

>>> import numpy as np
>>> a = [0,10,20,30,40,50,60]
>>> b = [0,2,4,5]
>>> res = np.asarray(a)[b].tolist()
>>> res
[0, 20, 40, 50]