Why manual string reverse is worse than slice reverse in Python 2.7? What is the algorithm being used in Slice?

Question

Below the performance difference between Slice and manual reverse operation. If this is the case, What is the reason for that?

timeit.timeit("a[::-1]","a=[1,2,3,4,5,6]",number=100)
6.054327968740836e-05

timeit.timeit("[a[i] for i in range(len(a)-1,-1,-1)]","a=[1,2,3,4,5,6]",number=100)
0.0003132152330920235

Answer 1

Here's the bytecode

from dis import dis
a = [1,2,3,4,5,6]

def func1():
    a[::-1]

def func2():
    [a[i] for i in range(len(a)-1,-1,-1)]

def func3():
    reversed(a)

In the second method, you're finding the length, creating a copy with range and creating the variable i.

Can also use reversed to create an iterable.

Answer 2

The slice notation for reversing a list drops down into C, which is considerably faster than a pure python implementation of a reverse. For instance, in the pure python approach the python interpreter must read, decode, and execute each instruction in the byte code, whereas the C call will be executing natively and suffer no such penalty. This penalty also extends to things such as method lookups when indexing an item and so forth, whereas in the C call there is no method, just address arithmetic. So efficient is the C implementation that it doesn't even bother with a specialised reversed slice function, and still beats the pure python implementation. Rather it creates a copy of the slice and the reverses the slice in place (done else where).

List slice code for cpython:

static PyObject *
list_slice(PyListObject *a, Py_ssize_t ilow, Py_ssize_t ihigh)
{
    PyListObject *np;
    PyObject **src, **dest;
    Py_ssize_t i, len;
    if (ilow < 0)
        ilow = 0;
    else if (ilow > Py_SIZE(a))
        ilow = Py_SIZE(a);
    if (ihigh < ilow)
        ihigh = ilow;
    else if (ihigh > Py_SIZE(a))
        ihigh = Py_SIZE(a);
    len = ihigh - ilow;
    np = (PyListObject *) PyList_New(len);
    if (np == NULL)
        return NULL;

    src = a->ob_item + ilow;
    dest = np->ob_item;
    for (i = 0; i < len; i++) {
        PyObject *v = src[i];
        Py_INCREF(v);
        dest[i] = v;
    }
    return (PyObject *)np;
}