Python Fastest way to find Indexes of item in list

Use list.index(elem, start)! That uses a for loop in C (see it's implementation list_index_impl function in the source of CPython's listobject.c). Avoid looping through all the elements in Python, it is slower than in C.

def index_finder(lst, item):
    """A generator function, if you might not need all the indices"""
    start = 0
    while True:
        try:
            start = lst.index(item, start)
            yield start
            start += 1
        except ValueError:
            break

import array
def index_find_all(lst, item, results=None):
    """If you want all the indices.
    Pass results=[] if you explicitly need a list,
    or anything that can .append(..)"""
    if results is None:
        length = len(lst)
        results = array.array('B') if length <= 2**8 else array.array('H') if length <= 2**16 else array.array('L') if length <= 2**32 else array.array('Q')
    start = 0
    while True:
        try:
            start = lst.index(item, start)
            results.append(start)
            start += 1
        except ValueError:
            return results

# Usage example
l = [1, 2, 3, 4, 5, 6, 7, 8] * 32

print(*index_finder(l, 1))
print(*index_find_all(l, 1))

def find(target, myList):
    for i in range(len(myList)):
        if myList[i] == target:
            yield i

def find_with_list(myList, target):
     inds = []
     for i in range(len(myList)):
         if myList[i] == target:
             inds += i,
     return inds


In [8]: x = range(50)*200
In [9]: %timeit [i for i,j in enumerate(x) if j == 3]
1000 loops, best of 3: 598 us per loop

In [10]: %timeit list(find(3,x))
1000 loops, best of 3: 607 us per loop
In [11]: %timeit find(3,x)
1000000 loops, best of 3: 375 ns per loop

In [55]: %timeit find_with_list(x,3)
1000 loops, best of 3: 618 us per loop

Assuming you want a list as your output: All options seemed exhibit similar time performance for my test with the list comprehension being the fastest (barely).

And if you're cool with returning a generator, it's way faster than the other approaches. Thought it doesn't account for actually iterating over the indices, nor does it store them, so the inds cannot be iterated over a second time.


Simply create a dictionary of item->index from the list of items using zip like so:

items_as_dict = dict(zip(list_of_items,range(0,len(list_of_items))))
index = items_as_dict(item)