{Independent,}Sample trait for StandardNormal or rename to Ziggurat

[SuperFluffy]

All structs in rand::distributions implement the Sample and IndependentSample trait, which gives us the ability to call fn ind_sample<R: Rng>(&self, &mut R) -> Support, with Support the value of our random number.

However, StandardNormal, also in rand::distributions, does not implement the aforementioned traits, but Rand instead, which allows to call fn rand<R: Rng>(rng: &mut R) -> Self, where self will be StandardNormal(x), with x the value of our random number.

I believe this to be an inconsistency. Either StandardNormal should be moved out of rand::distributions to make it explicit that it behaves in a completely different way, or the Sample and IndependentSample traits should be implemented for it.

Alternatively, I propose StandardNormal to be renamed to ZIGNOR (or Ziggurat, I guess).

[bluss]

StandardNormal is indeed not a distribution, just a cute newtype for a float with a different Rand implementation. Just like Closed01.

[dhardy]

Is there a good rationale for keeping the "standard distributions" like StandardNormal at all?

[pitdicker]

I agree StandardNormal is a strange fit in distributions. It seems more like a part of the implementation of Normal and LogNormal that got accidentally exposed.

@dhardy changed it to a simple (private) function in dhardy@a10251b.

[pitdicker]

@dhardy changed it to a simple (private) function in dhardy/rand@a10251b.

Now landed as part of #256.

[pitdicker]

Thanks for closing all the bugs 😄. Do you think I could get permission to also do so?

[dhardy]

Thanks! I don't actually have permission to do that; @alexcrichton?

[alexcrichton]

Sure! @dhardy I've made you an admin for the repo and invited @pitdicker as a collaborator

[pitdicker]

@alexcrichton Thank you!