Would it be very unpythonic to use this setitem function to overcome the list comprehension limitation?

Question

>>> a=range(5)
>>> [a[i] for i in range(0,len(a),2)] ## list comprehension for side effects
[0, 2, 4]
>>> a
[0, 1, 2, 3, 4]
>>> [a[i]=3 for i in range(0,len(a),2)] ## try to do assignment
SyntaxError: invalid syntax
>>> def setitem(listtochange,n,value):  ## function to overcome limitation
    listtochange[n]=value
    return value

>>> [setitem(a,i,'x') for i in range(0,len(a),2)] ## proving the function
['x', 'x', 'x']
>>> a 
['x', 1, 'x', 3, 'x']   # We did assignment anyway

Answer 1

For my timing mentioned (see also the Improving pure Python prime sieve by recurrence formula) from time import clock

def rwh_primes1(n):
    # https://stackoverflow.com/questions/2068372/fastest-way-to-list-all-primes-below-n-in-python/3035188#3035188
    """ Returns  a list of primes < n """
    sieve = [True] * (n//2)
    for i in xrange(3,int(n**0.5)+1,2):
        if sieve[i//2]:
            sieve[i*i//2::i] = [False] * ((n-i*i-1)//(2*i)+1)
    return [2] + [2*i+1 for i in xrange(1,n/2) if sieve[i]]

def rwh_primes_tjv(n):
    # recurrence formula for length by amount1 and amount2 tjv
    """ Returns  a list of primes < n """
    sieve = [True] * (n//2)
    amount1 = n-10
    amount2 = 6

    for i in range(3,int(n**0.5)+1,2):
        if sieve[i//2]:
             ## can you make recurrence formula for whole reciprocal?
            sieve[i*i//2::i] = [False] * (amount1//amount2+1)
        amount1-=4*i+4
        amount2+=4

    return [2] + [2*i+1 for i in xrange(1,n//2) if sieve[i]]

def rwh_primes_len(n):
    """ Returns  a list of primes < n """
    sieve = [True] * (n//2)

    for i in range(3,int(n**0.5)+1,2):
        if sieve[i//2]:
            sieve[i*i//2::i] = [False] * len(sieve[i*i//2::i])

    return [2] + [2*i+1 for i in xrange(1,n//2) if sieve[i]]

def rwh_primes_any(n):
    """ Returns  a list of primes < n """
    halfn=n//2
    sieve = [True] * (halfn)

    for i in range(3,int(n**0.5)+1,2):
        if sieve[i//2]:
            any(sieve.__setitem__(item,False) for item in range(i*i//2,halfn,i))

    return [2] + [2*i+1 for i in xrange(1,n//2) if sieve[i]]


if __name__ == "__main__":
    n = 1000000

    print("rwh sieve1")
    t=clock()
    r=rwh_primes1(n)
    print("Length %i,  %s ms" %(len(r),1000*(clock()-t)))

    print("rwh sieve with recurrence formula")
    t=clock()
    r=rwh_primes_tjv(n)
    print("Length %i,  %s ms" %(len(r),1000*(clock()-t)))

    print("rwh sieve with len function")
    t=clock()
    r=rwh_primes_len(n)
    print("Length %i,  %s ms" %(len(r),1000*(clock()-t)))

    print("rwh sieve with any with side effects")
    t=clock()
    r=rwh_primes_any(n)
    print("Length %i,  %s ms" %(len(r),1000*(clock()-t)))
    raw_input('Ready')

""" Output:
rwh sieve1
Length 78498,  213.199442946 ms
rwh sieve with recurrence formula
Length 78498,  218.34143725 ms
rwh sieve with len function
Length 78498,  257.80008353 ms
rwh sieve with any with side effects
Length 78498,  829.977273648 ms
Ready
"""

Length function and all with setitem are not satisfactory alternatives, but the timings are here to demonstrate it.

rwh sieve with len function Length 78498, 257.80008353 ms

rwh sieve with any with side effects Length 78498, 829.977273648 ms

Answer 2

Don't use list comprehensions to perform side-effects - that is not Pythonic. Use an explicit loop instead:

for i in range(0,len(a),2):
    a[i] = 3

Apart the side-effects in list comprehensions being surprising and unexpected, you are constructing a result list that you never use which is wasteful and completely unnecessary here.

Answer 3

Yes. And I recommend using

a[::2] = ['x'] * len(a[::2])

instead.

---

Edit:

Microbenchmarks for Python 2.6:

~:249$ python2.6 -m timeit -s 'a = range(2000)' 'a[::2] = [8] * len(a[::2])'
10000 loops, best of 3: 26.2 usec per loop

~:250$ python2.6 -m timeit -s 'a = range(2000)' 'a[::2] = [8] * (len(a)/2)'
10000 loops, best of 3: 19.6 usec per loop

~:251$ python2.6 -m timeit -s 'a = range(2000)' 'for i in xrange(0,len(a),2): a[i] = 8'
10000 loops, best of 3: 92.1 usec per loop

~:252$ python2.6 -m timeit -s 'a = range(2000)
> def assign(x,i,v):x[i]=v;return v' '[assign(a,i,8) for i in xrange(0, len(a), 2)]'
1000 loops, best of 3: 336 usec per loop

Python 3.1:

~:253$ python3.1 -m timeit -s 'a = list(range(2000))' 'a[::2] = [8] * len(a[::2])'
100000 loops, best of 3: 19.8 usec per loop

~:254$ python3.1 -m timeit -s 'a = list(range(2000))' 'a[::2] = [8] * (len(a)//2)'
100000 loops, best of 3: 13.4 usec per loop

~:255$ python3.1 -m timeit -s 'a = list(range(2000))' 'for i in range(0,len(a),2): a[i] = 8'
10000 loops, best of 3: 119 usec per loop

~:256$ python3.1 -m timeit -s 'a = list(range(2000))
> def assign(x,i,v):x[i]=v;return v' '[assign(a,i,8) for i in range(0, len(a), 2)]'
1000 loops, best of 3: 361 usec per loop

Answer 4

You can also use list.__setitem__

a = range(5)
[a.__setitem__(i,"x") for i in range(0,len(a),2)]

Or if you want to avoid the contruction of an intermediate list:

any(a.__setitem__(i,"x") for i in range(0,len(a),2))

But assignment in list comprehensions is indeed unpythonic.