Why evaluating boolean objects takes time in python

Question

I compared these two code snippets with a module timeitand realized that the second is a bit faster:

~$ python -m timeit —setup "l=[1, 2];k=1" "l[k==1]"
10000000 loops, best of 3: 0.0414 usec per loop
~$ python -m timeit —setup "l=[1, 2];k=1" "l[0 if k==1 else 1]"
10000000 loops, best of 3: 0.0372 usec per loop

Since the logic is the same, I thought that evaluating Boolean objects takes longer than integer equivalence ( True== 1 and False== 0), so I came up with the following test, and it turns out that I was correct:

~$ python -m timeit —setup "l=range(1000)" "l[False]"
10000000 loops, best of 3: 0.0411 usec per loop
~$ python -m timeit —setup "l=range(1000)" "l[False]"
10000000 loops, best of 3: 0.0394 usec per loop
~$ python -m timeit —setup "l=range(1000)" "l[False]"
10000000 loops, best of 3: 0.0416 usec per loop
~$ python -m timeit —setup "l=range(1000)" "l[True]"
10000000 loops, best of 3: 0.0428 usec per loop
~$ python -m timeit —setup "l=range(1000)" "l[True]"
10000000 loops, best of 3: 0.0394 usec per loop
~$ python -m timeit —setup "l=range(1000)" "l[True]"
10000000 loops, best of 3: 0.0393 usec per loop
~$ 
~$
~$ python -m timeit —setup "l=range(1000)" "l[0]"
10000000 loops, best of 3: 0.0232 usec per loop
~$ python -m timeit —setup "l=range(1000)" "l[0]"
10000000 loops, best of 3: 0.0232 usec per loop
~$ python -m timeit —setup "l=range(1000)" "l[0]"
10000000 loops, best of 3: 0.0232 usec per loop
~$ python -m timeit —setup "l=range(1000)" "l[1]"
10000000 loops, best of 3: 0.0232 usec per loop
~$ python -m timeit —setup "l=range(1000)" "l[1]"
10000000 loops, best of 3: 0.0232 usec per loop
~$ python -m timeit —setup "l=range(1000)" "l[1]"
10000000 loops, best of 3: 0.0232 usec per loop

But I do not know what is the reason. I mean, why is evaluation Trueand Falsetaking longer? I also noticed another mysterious thing during the test. In the first part, the standards have a variation of the results, and the numbers for the second are stable.

Answer 1

l[k==1] l[0 if k==1 else 1] . , , , . , , :

>>> timeit.timeit('l[k==1]', 'l=[1,2];k=1', number=100000000)
10.782931089401245
>>> timeit.timeit('l[0 if k==1 else 1]', 'l=[1,2];k=1', number=100000000)
11.140317916870117

l[0 if k==1 else 1] , , l[k==1] BINARY_SUBSCR:

TARGET_NOARG(BINARY_SUBSCR)
{
    w = POP();
    v = TOP();
    if (PyList_CheckExact(v) && PyInt_CheckExact(w)) {
        /* INLINE: list[int] */
        Py_ssize_t i = PyInt_AsSsize_t(w);
        if (i < 0)
            i += PyList_GET_SIZE(v);
        if (i >= 0 && i < PyList_GET_SIZE(v)) {
            x = PyList_GET_ITEM(v, i);
            Py_INCREF(x);
        }
        else
            goto slow_get;
    }
    else
      slow_get:
        x = PyObject_GetItem(v, w);

, Python 2, True , 1 - LOAD_CONST. LOAD_CONST co_consts, .

Answer 2

boolean integer division . () . :

python2

~$ python2 -m timeit --setup "l=range(1000)" "l[False]"
10000000 loops, best of 3: 0.0366 usec per loop
~$ python2 -m timeit --setup "l=range(1000)" "l[False]"
10000000 loops, best of 3: 0.0332 usec per loop
~$ python2 -m timeit --setup "l=range(1000)" "l[1]"
10000000 loops, best of 3: 0.0193 usec per loop
~$ python2 -m timeit --setup "l=range(1000)" "l[1]"
100000000 loops, best of 3: 0.0194 usec per loop
~$ python2 -m timeit --setup "l=range(1000)" "l[1]"
100000000 loops, best of 3: 0.0195 usec per loop
~$ python2 -m timeit --setup "l=range(1000)" "l[0]"
100000000 loops, best of 3: 0.0196 usec per loop

Python 3

~$ python -m timeit --setup "l=range(1000)" "l[0]"
10000000 loops, best of 3: 0.0712 usec per loop
~$ python -m timeit --setup "l=range(1000)" "l[0]"
10000000 loops, best of 3: 0.072 usec per loop
~$ python -m timeit --setup "l=range(1000)" "l[0]"
10000000 loops, best of 3: 0.0719 usec per loop
~$ python -m timeit --setup "l=range(1000)" "l[False]"
10000000 loops, best of 3: 0.082 usec per loop
~$ python -m timeit --setup "l=range(1000)" "l[False]"
10000000 loops, best of 3: 0.0821 usec per loop

: Python, Python. - . , 0 1 , 4 ... ( , , )