search

idontgetoutmuch / random

Random number library
Other
3 stars 2 forks source link

Add NFData #170

Closed lehins closed 4 years ago

lehins commented 4 years ago

Add NFData instances for STGenM, IOGenM, AtomicGenM and StateGenM

IORef and STRef have NFData instances. There is no reason not to derive them for mutable generators as well. Very useful for benchmarks and such.

Shimuuar commented 4 years ago

Here is rather short discussion on whether NFData instances for IO/STRef should exist: https://github.com/haskell/deepseq/pull/47

I think that these instances shouldn't exists since they only reduce reference to WHNF.

lehins commented 4 years ago

Thanks @Shimuuar for the link. Why everything drags for so long in Haskell? haskell/deepseq#16 has been discussed for 4 years now and still nothing happens. Just rip the bandage, remove the instances, and be done with it. Longer the inaction, the more people get to rely on those instances, then harder it becomes to remove them and more code gets broken in the end.

Considering the unknown future of NFData instances for IORef and STRef, I prefer not to deal with this issue here in random. Closing.

Shimuuar commented 4 years ago

Why everything drags for so long in Haskell?

Laziness?

lehins commented 4 years ago

@Shimuuar I don't think it is laziness, I think it is inability to make an executive decision. There is not much work involved in solving an issue like that.

Shimuuar commented 4 years ago

I just couldn't resist joking about haskell's model of evaluation. I do agree that this sort of indecision seems to be endemic in haskell ecosystem.

idontgetoutmuch commented 4 years ago

@Shimuuar I don't think it is laziness, I think it is inability to make an executive decision. There is not much work involved in solving an issue like that.

There are too many clever people in Haskell mostly with valid reasons for what they suggest. I don't know what the answer is. random itself could have been improved incrementally but this sort of discussion held back any such improvements.