On Wed, Oct 24, 2012 at 3:24 PM, Alan Gauld <alan.ga...@btinternet.com> wrote:
> On 24/10/12 18:49, eryksun wrote:
>
>> Using "in ['25', '26']" also checks a compiled tuple constant:
>>
>> >>> compile("x in ['25', '26']", '', 'eval').co_consts
>> ('25', '26', ('25', '26'))
>>
>> 3.x adds frozenset support:
>>
>> >>> compile("x in {'25', '26'}", '', 'eval').co_consts
>> ('25', '26', frozenset({'25', '26'}))
>
>
> I confess I don't know what that means!
> And if I don't I doubt if the OP will either.
> Can you explain the above in English please?
Sorry, I was showing that the compiler (at least for CPython) special
cases "in" and "not in" comparisons when the right-hand operand is a
list literal of constants. Instead of going to the trouble of building
a list every time the code runs, it uses a tuple that's stored in the
code object's co_consts attribute.
In Python 3, this optimization was extended to set literals constants.
I guess I should show that in more detail by disassembling a function
instead of manually compiling code.
First in Python 2.7.3:
>>> def func(x):
... return x in {1, 2, 3}
>>> dis.dis(func)
2 0 LOAD_FAST 0 (x)
3 LOAD_CONST 1 (1)
6 LOAD_CONST 2 (2)
9 LOAD_CONST 3 (3)
12 BUILD_SET 3
15 COMPARE_OP 6 (in)
18 RETURN_VALUE
Each time the function is called it has to build a set (BUILD_SET)
from the constants 1, 2, and 3. Compare that to Python 3.2.3:
>>> dis.dis(func)
2 0 LOAD_FAST 0 (x)
3 LOAD_CONST 4 (frozenset({1, 2, 3}))
6 COMPARE_OP 6 (in)
9 RETURN_VALUE
Here the compiler stored a frozenset in the code object. So it only
has to load the frozenset on the stack instead of building a new set
each time the function is called.
compile:
http://docs.python.org/py3k/library/functions.html#compile
frozenset:
http://docs.python.org/py3k/library/functions.html#func-frozenset
code objects:
http://docs.python.org/py3k/reference/datamodel.html#index-51
The optimization is in peephole.c, PyCode_Optimize/tuple_of_constants:
http://hg.python.org/cpython/file/3d0686d90f55/Python/peephole.c#l469
http://hg.python.org/cpython/file/3d0686d90f55/Python/peephole.c#l25
_______________________________________________
Tutor maillist - Tutor@python.org
To unsubscribe or change subscription options:
http://mail.python.org/mailman/listinfo/tutor
