undocumented method is quest of comment
Hi guys I would love to have a better method comment for hashMultiply | low | low := self bitAnd: 16383. ^(16r260D * low + ((16r260D * (self bitShift: -14) + (16r0065 * low) bitAnd: 16383) * 16384)) bitAnd: 16r0FFFFFFF Some love? Stef
It's not documented because hashing is complex... The goal is to shuffle the receiver bits and distribute them over the available 28bits used for hash codes. why and how it does so is a tough subject, Andres Valloud wrote a whole book on it ;) The basic operation which is emulated is relatively simple, but implementation is overly complex, because it tries to avoid creating LargeIntegers in 32bits VM *16384 is <=> bitShift: 14 bitAnd 16383 is <=> isolating lowest 14 bits... While at it, one should also rewrite the method contents in order to use the new primitive available and profit by more efficient 64bit version for 64bit VM... Fortunately, one can still pick the methods from Squeak :) then forget the original authors because it really does not matter, does it? 2017-11-16 22:25 GMT+01:00 Stephane Ducasse <stepharo.self@gmail.com>:
Hi guys
I would love to have a better method comment for
hashMultiply | low | low := self bitAnd: 16383. ^(16r260D * low + ((16r260D * (self bitShift: -14) + (16r0065 * low) bitAnd: 16383) * 16384)) bitAnd: 16r0FFFFFFF
Some love?
Stef
https://github.com/pharo-project/pharo/pull/487 2017-11-16 23:32 GMT+01:00 Nicolas Cellier < nicolas.cellier.aka.nice@gmail.com>:
It's not documented because hashing is complex... The goal is to shuffle the receiver bits and distribute them over the available 28bits used for hash codes. why and how it does so is a tough subject, Andres Valloud wrote a whole book on it ;)
The basic operation which is emulated is relatively simple, but implementation is overly complex, because it tries to avoid creating LargeIntegers in 32bits VM
*16384 is <=> bitShift: 14 bitAnd 16383 is <=> isolating lowest 14 bits...
While at it, one should also rewrite the method contents in order to use the new primitive available and profit by more efficient 64bit version for 64bit VM...
Fortunately, one can still pick the methods from Squeak :) then forget the original authors because it really does not matter, does it?
2017-11-16 22:25 GMT+01:00 Stephane Ducasse <stepharo.self@gmail.com>:
Hi guys
I would love to have a better method comment for
hashMultiply | low | low := self bitAnd: 16383. ^(16r260D * low + ((16r260D * (self bitShift: -14) + (16r0065 * low) bitAnd: 16383) * 16384)) bitAnd: 16r0FFFFFFF
Some love?
Stef
One little action is worth a thousand words :) 2017-11-17 11:51 GMT+01:00 Pavel Krivanek <pavel.krivanek@gmail.com>:
https://github.com/pharo-project/pharo/pull/487
2017-11-16 23:32 GMT+01:00 Nicolas Cellier <nicolas.cellier.aka.nice@ gmail.com>:
It's not documented because hashing is complex... The goal is to shuffle the receiver bits and distribute them over the available 28bits used for hash codes. why and how it does so is a tough subject, Andres Valloud wrote a whole book on it ;)
The basic operation which is emulated is relatively simple, but implementation is overly complex, because it tries to avoid creating LargeIntegers in 32bits VM
*16384 is <=> bitShift: 14 bitAnd 16383 is <=> isolating lowest 14 bits...
While at it, one should also rewrite the method contents in order to use the new primitive available and profit by more efficient 64bit version for 64bit VM...
Fortunately, one can still pick the methods from Squeak :) then forget the original authors because it really does not matter, does it?
2017-11-16 22:25 GMT+01:00 Stephane Ducasse <stepharo.self@gmail.com>:
Hi guys
I would love to have a better method comment for
hashMultiply | low | low := self bitAnd: 16383. ^(16r260D * low + ((16r260D * (self bitShift: -14) + (16r0065 * low) bitAnd: 16383) * 16384)) bitAnd: 16r0FFFFFFF
Some love?
Stef
Indeed no little action is worth a thousand words. Everybody plants its own seed. Tx nicolas and Pavel. Stef On Fri, Nov 17, 2017 at 3:00 PM, Nicolas Cellier <nicolas.cellier.aka.nice@gmail.com> wrote:
One little action is worth a thousand words :)
2017-11-17 11:51 GMT+01:00 Pavel Krivanek <pavel.krivanek@gmail.com>:
https://github.com/pharo-project/pharo/pull/487
2017-11-16 23:32 GMT+01:00 Nicolas Cellier <nicolas.cellier.aka.nice@gmail.com>:
It's not documented because hashing is complex... The goal is to shuffle the receiver bits and distribute them over the available 28bits used for hash codes. why and how it does so is a tough subject, Andres Valloud wrote a whole book on it ;)
The basic operation which is emulated is relatively simple, but implementation is overly complex, because it tries to avoid creating LargeIntegers in 32bits VM
*16384 is <=> bitShift: 14 bitAnd 16383 is <=> isolating lowest 14 bits...
While at it, one should also rewrite the method contents in order to use the new primitive available and profit by more efficient 64bit version for 64bit VM...
Fortunately, one can still pick the methods from Squeak :) then forget the original authors because it really does not matter, does it?
2017-11-16 22:25 GMT+01:00 Stephane Ducasse <stepharo.self@gmail.com>:
Hi guys
I would love to have a better method comment for
hashMultiply | low | low := self bitAnd: 16383. ^(16r260D * low + ((16r260D * (self bitShift: -14) + (16r0065 * low) bitAnd: 16383) * 16384)) bitAnd: 16r0FFFFFFF
Some love?
Stef
participants (3)
-
Nicolas Cellier -
Pavel Krivanek -
Stephane Ducasse