Smalltalk › Pharo › Pharo Smalltalk Developers

Association hash vs equality

_‹ Previous Topic Next Topic _›

Classic

List

Threaded

2 messages Options

Jan Blizničenko

Association hash vs equality

Hello, I've been looking at implementation of Associtation and LookupKey (its
superclass).
In LookupKey, there is LookupKey>>#= and LookupKey>>#hash both implemented
to use key (compare keys and use hash of key), but in Association,
Association>>#= compares both key and value, but hash is not changed there,
so only hash of key is used, therefore it does not correspond to
implementation of method = ... shouldn't it?

To demonstrate:
(1 -> #a) = (1 -> #b).
is false but
(1 -> #a) hash = (1 -> #b) hash.
is true

I also noticed similar discussion for Doplhin smalltalk
http://forum.world.st/Association-gt-gt-hash-td3368868.html where guys state
it is all right, but I still do not see why.

Jan

--
Sent from: http://forum.world.st/Pharo-Smalltalk-Developers-f1294837.html

Guillermo Polito

Re: Association hash vs equality

Actually, #= and #hash should be defined so:

if (a=b) then a hash = b hash

if a != b, it's not important

This is required because some collections (like sets, dictionaries) will index their elements by their hash. Then, typically the collection implementation will first ask the hash, access the index with it, and retrieve the object if it is = only. Notice that several objects can have the same hash but be equals nonetheless.

On Thu, Aug 16, 2018 at 11:36 AM Jan Blizničenko <[hidden email]> wrote:

Hello, I've been looking at implementation of Associtation and LookupKey (its
superclass).
In LookupKey, there is LookupKey>>#= and LookupKey>>#hash both implemented
to use key (compare keys and use hash of key), but in Association,
Association>>#= compares both key and value, but hash is not changed there,
so only hash of key is used, therefore it does not correspond to
implementation of method = ... shouldn't it?

To demonstrate:
(1 -> #a) = (1 -> #b).
is false but
(1 -> #a) hash = (1 -> #b) hash.
is true

I also noticed similar discussion for Doplhin smalltalk
http://forum.world.st/Association-gt-gt-hash-td3368868.html where guys state
it is all right, but I still do not see why.

Jan

--
Sent from: http://forum.world.st/Pharo-Smalltalk-Developers-f1294837.html

Guille Polito

Research Engineer

Centre de Recherche en Informatique, Signal et Automatique de Lille

CRIStAL - UMR 9189

French National Center for Scientific Research - http://www.cnrs.fr

Web: http://guillep.github.io

Phone: +33 06 52 70 66 13