> On 12 Dec 2016, at 14:32, Sven Van Caekenberghe <s...@stfx.eu> wrote:
>
>
>> On 12 Dec 2016, at 18:29, Marcus Denker <marcus.den...@inria.fr> wrote:
>>
>>
>>> On 1 Dec 2016, at 11:41, Christophe Demarey <christophe.dema...@inria.fr>
>>> wrote:
>>>
>>>
>>>> Le 1 déc. 2016 à 15:04, Marcus Denker <marcus.den...@inria.fr> a écrit :
>>>>
>>>> The idea that a set is always created with a minimal size is because we
>>>> always add at least some elements
>>>
>>> not always. You know in some cases 80% of the time, your instances will
>>> have an empty Set and with some values for the remaining 20%.
>>> If you check in Moose, you should find a lot this pattern.
>>
>> It would be interesting to play around with these things… the problem is
>> that they are very context dependent, so there is lots
>> of questions what to do (more a research topic than something one can solve
>> now…).
>>
>> What we should do:
>>
>> -> the size should be 5, like it is for Dictionaries We need to unify the
>> two version of the method that calculate the initial size.
>>
Issue tracker:
https://pharo.fogbugz.com/f/cases/19458/unify-sizeFor-for-HashedCollection-and-Set
>> What we could do:
>>
>> -> provide a version of “new” that actually makes explicity a small array of
>> size 1. For the case where the user knows that
>> the Set will be empty most of the time.
>
> There could be an #empty class method for that.
> And one could overwrite the #with: class method to create a singleton.
https://pharo.fogbugz.com/f/cases/19459/add-empty-instance-creation-method
<https://pharo.fogbugz.com/f/cases/19459/add-empty-instance-creation-method>
(just the issues, no code)
Marcus
