How do I lazily generate a sequence of random instances of type `T` as an `InputRange`?

On Tuesday, 26 April 2016 at 10:31:22 UTC, Nordlöw wrote:
> How do I lazily generate a sequence of random instances of type `T` as an `InputRange`?

Ahh, I found it:

        import std.range : generate, take;
        import std.random : uniform;
        auto randomSamples = generate!(() => uniform!Key).take(n);

I should have guessed that...

On Tuesday, 26 April 2016 at 10:50:27 UTC, Nordlöw wrote:
> On Tuesday, 26 April 2016 at 10:31:22 UTC, Nordlöw wrote:
>> How do I lazily generate a sequence of random instances of type `T` as an `InputRange`?
>
> Ahh, I found it:
>
>         import std.range : generate, take;
>         import std.random : uniform;
>         auto randomSamples = generate!(() => uniform!Key).take(n);
>
> I should have guessed that...

Btw if you do random generation at the moment, you should always be aware that it's super-easy to do an implicit copy if you pass around the rndGen, see:

http://dconf.org/2015/talks/wakeling.html

On Tuesday, 26 April 2016 at 11:09:42 UTC, Seb wrote:
> Btw if you do random generation at the moment, you should always be aware that it's super-easy to do an implicit copy if you pass around the rndGen

So should I pass it by ref or const ref then?

On Tuesday, 26 April 2016 at 17:38:33 UTC, Nordlöw wrote:
> On Tuesday, 26 April 2016 at 11:09:42 UTC, Seb wrote:
>> Btw if you do random generation at the moment, you should always be aware that it's super-easy to do an implicit copy if you pass around the rndGen
>
> So should I pass it by ref or const ref then?

Doh. Const ref is of no use... just by ref then, right?

On Tuesday, 26 April 2016 at 17:38:33 UTC, Nordlöw wrote:
> On Tuesday, 26 April 2016 at 11:09:42 UTC, Seb wrote:
>> Btw if you do random generation at the moment, you should always be aware that it's super-easy to do an implicit copy if you pass around the rndGen
>
> So should I pass it by ref or const ref then?

I guess both work fine, but this was just a warning for other case.
The example that @WebDrake mentioned on his slides can happen too easily - a reason why we should fix it (which is on my agenda).