A Tausworthe random number generator
terchestor
terchestor at gmail.com
Tue Jan 21 09:08:52 PST 2014
On Tuesday, 21 January 2014 at 12:32:43 UTC, bearophile wrote:
> terchestor:
>
>> https://github.com/Terchestor/dlang/tree/Tausworthe-class
>>
>
>> /++
>> §§§ Tausworthe
>
> What's the purpose of those §? They add noise. If they aren't
> necessary I suggest to remove them.
>
Well, a fantasy, but I wont keep it!
>
>> class Tausworthe ( I, F ) if ( __traits(isIntegral, I ) &&
>> __traits(isFloating, F ) )
>
> I suggest to not put spaces around ( ). And if you want also
> remove the space after the class/functions name.
That's an old usage of mine.
> Perhaps you want to use a struct? The decision is important.
> Are your methods all virtual?
Not a struc, I need a class.
>
>> {
>> alias I integral_t;
>> alias F floating_t;
>
> Today the syntax "alias Name = Foo;" is preferred.
I knew not. Do ypu mean: alias integral_t = I;
> In D code the use of "_t" as suffix for types is less common.
> And perhaps it's better to use a starting upper case letter for
> their name.
I forgot the Capital letter, but I like the trailing _t, like
size_t.
>
>> enum tauswortheConstFloat : floating_t
>> {
>> tau0 = 4294967296.0f
>
> Class/struct/Enum/Union names should start with an upper case
> letter.
>
>
>> tau1 = 4294967294,
>
> Perhaps that literal can end with a U.
>
>
>> ~this( )
>> { }
>
> No need for this.
Of course!
>
>> void seed( integral_t seed_=1 )
>> ...
>> auto s = seed_;
>
> In general annotate with enum/const/immutable all variables
> that don't need to mutate.
I'd better put an assertion (or enforce?) to be sure seed won't
be null, and give constantness to seed.
>> integral_t fseed( const integral_t s_ )
>> {
>> return ( tauswortheConst.lgc * s_ ) &
>> tauswortheConst.mask;
>> }
>
> That's probably better annotated as pure nothrow.
OK.
>
>> }//§§==========END_CLASS::Tausworthe==========
>
> I suggest to reduce the amount of === in this code.
OK.
>
>> auto ?tausworthe0 = new Tausworthe!(ulong, double);
>
> While D supports Unicode identifiers, they are a often a pain
> in the ass (this is a personal opinion), so perhaps you want to
> kill them all.
Not sure. I'm experimenting a kind of tagging to easily
distinguish nested functions (with greek phi) or temporaries.
Perhaps not a so good idea, but i'll see if enhances the
readability later.
>
>> //§§ __COPYRIGHT_NOTICE__
>
> I think there's a ddoc field syntax for the copyright in the
> module ddoc.
Yes, but it's a work on progress...
>
Thank you so much for your invaluable help.
More information about the Digitalmars-d
mailing list