Problem with rng/seed of rgalsim.random.GaussianDeviate

[Hbretonniere]

While gaslim.BaseDeviate takes a random generator (which is what we want), galsim.GaussianDeviate takes a seed.

We need a method which takes a rng, but with which we can also generate vectors (sampling).

[Hbretonniere]

Actually galsim.GaussianDeviate can take a rng instead of a seed... So maybe there s no problem at all...

But Actually I don't understand why do we create again an rng inside the sample method... (gd = galsim.GaussianDeviate(rng))

Why don t we just do :

sample(rng):
   if rng=None:
      rng = galsim.GaussianDeviate(rng).  # because rng=None, it means we don t have any control on the seed of the GD
   if random nornal...:     # for the flow
      noise = np.empty(...)
      # no gd = new rng
      rng.generate(noise)     # gaussiandeviates actually has a generate method, inheritate it fron basedeviate

If you want a control on the seed, you create the rng outside the function, with :
rng = galsim.GaussianDeviate(seed), and pass the rng to the sample method.

[Hbretonniere]

I've tried it, it looks like it works. I can share you my version if you think it's the good solution.

[EiffL]

Ok nice :-) And it works with both GalSim 1.15 and 2.x?

[EiffL]

And yes, for sure, feel free to submit a PR, I'll be happy to review it