iterating bit-by-bit across int?

[Matthew Wilson]

I'm playing around with genetic algorithms and I want to write a
function that mutates an integer by iterating across the bits, and about
1 in 10 times, it should switch a zero to a one, or a one to a zero.

I'm not sure how many bits are inside a python integer. The library
reference says at least 32.

I'm thinking about writing a function that eats integers and poops out
lists of bools; and then I can iterate through that, and change values
in there. But before I do that, does anyone have a better idea?

[Peter Hansen]

Anton Vredegoor wrote:

la*****@my-deja.com (John Ladasky) wrote:

I'm thinking about writing a function that eats integers and poops out
lists of bools; and then I can iterate through that, and change values
in there. But before I do that, does anyone have a better idea?

Using your approach, you would first need to disassemble the integer,
then reassemble it. You can cut this bitwise cranking in half.
Define an integer, in which 10% of the bits is a "1". Then do an
exclusive-or operation between this integer and the one you wish to
mutate.

This creates a new problem which is interesting in itself. How to
create this integer?

I would think the simplest approach is (similar to a suggestion
already made about the original problem) to predefine a set of bitmasks,
each with only one bit set, then simply select one or more at random
and add (or OR) them together, then XOR the result with the original.

You could either iterate over the list of bitmasks, checking a
random result at each step to decide whether to include that bitmask,
or you could just random.shuffle( ) the list, then select a random
number of entries from the start of the list. Using the new sum()
(or a filter() with operator.add) would make the whole thing as
simple as a couple of function calls:

genome = 12345678
numMutations = random.randint( 0, 4)
mutant = genome ^ sum(random.shuf fle(bitmasks)[:numMutations])

# done!!

The algorithm used to select the number of mutations to make could
of course be more sophisticated than the above...

-Peter

[Tim Roberts]

Brian Kelley <bk*****@wi.mit .edu> wrote:

Paul Rubin wrote:

Long ints can have as many bits as you want.

Such as -1L which has an infinite number of bits.

Oh come on, that's just silly. If that were true, I could not type this at
a Python prompt because it would cause an overflow:

i = -1L

-1L, in fact, has exactly the same number of bits as 1L and, probably,
1048576L.
--
- Tim Roberts, ti**@probo.com
Providenza & Boekelheide, Inc.

[Brian Kelley]

Tim Roberts wrote:

Brian Kelley <bk*****@wi.mit .edu> wrote:

Paul Rubin wrote:

Long ints can have as many bits as you want.

Such as -1L which has an infinite number of bits.

Oh come on, that's just silly. If that were true, I could not type this at
a Python prompt because it would cause an overflow:

i = -1L

try

i = -1L
while i:
i = i >> 1

and then repeat the statement :) My only point is that you have to be
careful with long ints.