Add Julia equivalent of "nargout" to docs

[lsorber]

What are the best ways of translating MATLAB functions to Julia of which the output and its computations are based on nargout?

In my case, I would like to know the best workaround for a hypothetical optimization suite written in Julia in which the user is asked to supply a function which computes an objective function value, and as an optional second output a derivative. The advantage of a nargout approach is that the user need only write one function and, importantly, that in many cases some of the computation between the objective function and its derivative may be shared.

[lindahua]

I think Julia is not going to take the MATLAB way, that is to let the function to decide the behavior based on the number of outputs arguments. And as far as I can tell, there is no way to tell "the number of outputs" within a function.

[lsorber]

@lindahua So it seems. However, I think it would be a good idea to add a section to the docs with the Julian way of handling such situations. For example, in my case I am interested in the best solution to the optimization suite problem I described above.

[timholy]

The standard solution is to pass additional outputs as pre-allocated inputs, and have your function modify the inputs in-place.

FYI: both Optim and NLOpt handle this precise issue (optionally asking for the gradient) in just this way. NLOpt's approach is slightly better: rather than passing nothing to indicate you don't want the gradient, from a type-consistency perspective it's better to pass an empty gradient vector. The optimization function can pre-construct an empty vector so that you don't have the overhead of doing so on each iteration.

[timholy]

From the perspective of the docs, I think we need a Julia <--> Matlab page somewhere. Want to start one?

[lsorber]

@timholy Thanks, that's the answer I was looking for. I would love to get involved more actively with Julia, but time is a continuously depleted resource of mine.

[timholy]

time is a continuously depleted resource of mine

join the crowd :-)

[JeffBezanson]

Of course there is the list at http://docs.julialang.org/en/latest/manual/getting-started/#noteworthy-differences-from-matlab but maybe we're at the point where a list like this no longer cuts it.

[timholy]

It's certainly useful, and fairly complete (although this particular issue doesn't appear). Someone recently pointed out http://wiki.scipy.org/NumPy_for_Matlab_Users which has some attractive features.

[ViralBShah]

Updated URL: http://docs.julialang.org/en/latest/manual/noteworthy-differences/#noteworthy-differences-from-matlab

[tkelman]

backported (squashed along with the formatting fix) in 4baf8e4