On Wed, Oct 19, 2011 at 08:26:18PM +0000, commits@source.squeak.org wrote:
Nicolas Cellier uploaded a new version of Kernel to project The Inbox: http://source.squeak.org/inbox/Kernel-nice.644.mcz
==================== Summary ====================
Name: Kernel-nice.644 Author: nice Time: 19 October 2011, 10:26:06.268 pm UUID: 612f5a8e-66ec-4531-86ab-ebc494038a94 Ancestors: Kernel-nice.642
Add Integer>>bitCount to count the bits set (to 1) in an Integer.
I like this.
But if #bitAt: is defined for all integers, then for consistency #bitCount should work for negative integers also. Integers don't really have bits unless you are thinking in terms of the underlying representation, which is certain to be twos complement in any use case that I can imagine. So counting bits in a negative integer should work in the same manner implied by #bitAt:.
I am surprised that a method like this does not already exist. But the method finder does not seem to find #bitCount either, so I wonder if some equivalent method already exists somewhere?
BTW, tinkerers may enjoy playing with www.squeaksource.com/TwosComplement, which demonstrates bit twiddling in twos complement register arithmetic:
16r0505 bitCount ==> 4 (16r0505 asRegister bits select: [:bit | bit value]) size ==> 4
Dave
2011/10/20 David T. Lewis lewis@mail.msen.com:
On Wed, Oct 19, 2011 at 08:26:18PM +0000, commits@source.squeak.org wrote:
Nicolas Cellier uploaded a new version of Kernel to project The Inbox: http://source.squeak.org/inbox/Kernel-nice.644.mcz
==================== Summary ====================
Name: Kernel-nice.644 Author: nice Time: 19 October 2011, 10:26:06.268 pm UUID: 612f5a8e-66ec-4531-86ab-ebc494038a94 Ancestors: Kernel-nice.642
Add Integer>>bitCount to count the bits set (to 1) in an Integer.
I like this.
But if #bitAt: is defined for all integers, then for consistency #bitCount should work for negative integers also. Integers don't really have bits unless you are thinking in terms of the underlying representation, which is certain to be twos complement in any use case that I can imagine. So counting bits in a negative integer should work in the same manner implied by #bitAt:.
Our integers are unbounded because arbitrarily large. So the numbers of bits set to 1 is infinite for a negative integer... So #bitCount will fail for the same reason as #highBit.
Or we could answer the bitCountOfMagnitude ? Or count the bit set to zero for a negative number ? (and let highBit answer position of higher 0)...
Nicolas
I am surprised that a method like this does not already exist. But the method finder does not seem to find #bitCount either, so I wonder if some equivalent method already exists somewhere?
BTW, tinkerers may enjoy playing with www.squeaksource.com/TwosComplement, which demonstrates bit twiddling in twos complement register arithmetic:
16r0505 bitCount ==> 4 (16r0505 asRegister bits select: [:bit | bit value]) size ==> 4
Dave
On Thu, Oct 20, 2011 at 04:53:16PM +0200, Nicolas Cellier wrote:
2011/10/20 David T. Lewis lewis@mail.msen.com:
On Wed, Oct 19, 2011 at 08:26:18PM +0000, commits@source.squeak.org wrote:
Nicolas Cellier uploaded a new version of Kernel to project The Inbox: http://source.squeak.org/inbox/Kernel-nice.644.mcz
==================== Summary ====================
Name: Kernel-nice.644 Author: nice Time: 19 October 2011, 10:26:06.268 pm UUID: 612f5a8e-66ec-4531-86ab-ebc494038a94 Ancestors: Kernel-nice.642
Add Integer>>bitCount to count the bits set (to 1) in an Integer.
I like this.
But if #bitAt: is defined for all integers, then for consistency #bitCount should work for negative integers also. Integers don't really have bits unless you are thinking in terms of the underlying representation, which is certain to be twos complement in any use case that I can imagine. So counting bits in a negative integer should work in the same manner implied by #bitAt:.
Our integers are unbounded because arbitrarily large. So the numbers of bits set to 1 is infinite for a negative integer... So #bitCount will fail for the same reason as #highBit.
Or we could answer the bitCountOfMagnitude ? Or count the bit set to zero for a negative number ? (and let highBit answer position of higher 0)...
You're right of course, and specifying the size does seem a bit strained now that you mention it. So your original implementation is probably best.
Dave
squeak-dev@lists.squeakfoundation.org
-
David T. Lewis -
Nicolas Cellier