Add fftfreq (open question(s))

[skeydan]

Hi @dfalbel, here's my next try at fftfreq() :-).

I still have two questions (one of which is a problem, the other one just something to clarify at this point).

The problem: I see that I'm handling the out tensor incorrectly (this is evident from the test in wrapers.R already).
I was looking for examples for how to do this in the code base, but didn't find anything that matched closely enough, so what I did was following what is seen in test-storage.R ... Probably this could not even work in this case, since here, a function call is involved ...
In reality, I do not even see why that option would be useful (to be able to pass in a tensor here), so if you like we could also just remove that argument ... (I would still be interested in knowing whether it'd be feasible in principle).

The question: How does gen-namespace-examples.R work? I see it used also for function which, in gen-namespace-docs.R, already have examples? I guess I don't need to add anything there, but if I did, how'd I generate the hash?

Thanks!!

[skeydan]

Oh wait ... (re "problem") ... started looking at another PR-to-be (for torch_where) and realized that what I need is a complementing _out function! And that exists already, and as far as I see, I can just leave it as-is ... since I wouldn't know why someone would want to export it! Right?

Well in fact there still is one thing I'll modify there: the d argument should not have a default of 1L, since it's a float (nothing to do with dimensions) ... I think it should be fine to make that change inside this PR, since I'm making the analogous modification in fftfreq ...

Going to update the PR, please let me know if you disagree :-)

[skeydan]

Me again... I think what I proposed above isn't the correct way to handle the 1L ... (shouldn't update gen-namespace.R manually anyway) ... I see this in r.R:

if (default %in% c("1", "0", "-1", "2", "-2", "100", "-100",
                     "20")) {
    return(paste0(default, "L"))
  }

What do we do about this? I'll leave it untouched for now, and just remove the case-handling code for out as I said before.

[skeydan]

OK, I think that should do it ...

[dfalbel]

Looks great! Just minor updates in the docs.

[dfalbel]

Regarding the defaults: since you already use the correct default in the wrapper function there's no need to modify the code generation or the generated function as it will never be called with its own default.

Regarding out: I think most of this out arguments are used for saving memory, by writing the results of a computation in an unused buffer, but we don't currently support those in R and if add support for them in the future it will probably be made in the C/C++ level instead.

Regarding examples: There's some functionality in here that was used to generate the docs but I would be surprised if it still works. It has been a long time I don't try it - it was mostly used to generate the first batch of functions.

[skeydan]

Many thanks for the review and the explanations, - got it!