pyimport syntax: from <module> import <symbol>

[cncastillo]

Maybe in @pyimport macros you could add some other sintaxis, for example if the user dont need to load all the package and only one function or more of it:

@pyimport from numpy import zeros, array, fromtxt
@pyimport from mayavi import mlab
@pyimport from scipy.ndimage import gaussian_filter as filter

[stevengj]

I would prefer to use the Julian syntax here (when in Rome...): @pyimport numpy: zeros, array, fromtxt etcetera.

[hammer]

+1 been laboring with multi-line import statements due to lack of this feature.

[stevengj]

I'm a little worried about namespace conflicts. If I implement @pyimport numpy: zeros for example, by setting const zeros = pyimport("numpy")[:zeros], then I will overwrite the built-in Julia zeros function. It seems safer just to require you to call numpy.zeros(...).

@hammer, can you give a concrete example where this would be important to you?

[marius311]

Obviously this is 2 years old, but maybe something has changed since then that warrants adding this? I think it'd be a great feature. I don't have concrete examples to add, just to say that obviously it would allow writing more concise easier-to-read code in many cases.

As for overwriting, once I've written @pyimport numpy: zeros doesn't that mean I want the overwriting to happen? Its not like it'd be unexpected in some way.

[stevengj]

Honestly, I'm thinking of getting rid of @pyimport completely, and telling people to just do numpy = pyimport("numpy"). The reason is that it is confusing that you can use . to get the members of a @pyimport module, but for everything else you need foo[:bar].

You can still just do z = pyimport("numpy")[:zeros], so a Pythonic syntax doesn't actually give you any new functionality.

[cstjean]

I would prefer to use the Julian syntax here (when in Rome...): @pyimport numpy: zeros, array, fromtxt etcetera.

I implemented it in ScikitLearn:

using ScikitLearn.Skcore: @pyimport2
@pyimport2 scipy.ndimage: (gaussian_filter, fourier_uniform)

but I struggled to get the hygiene right. I can make a PR for it if there is interest.

Honestly, I'm thinking of getting rid of @pyimport completely, and telling people to just do numpy = pyimport("numpy").

-1 for me, that would get tiring.

[marius311]

Thanks guys, yea @pyimport2 is exactly what I'm looking for, helpful to know that's there.

Re: removing @pyimport completely, its true its slightly inconsistent that you get to use the dot notation for the first level, but honestly as a user that was pretty easy to figure out from skimming your docs and trying a few simple examples. For me it'd seem to be sacrificing too much convenience in the name of consistency. My own +1, as you could probably guess, would instead be incorporating the @pyimport2 functionality into PyCall.