Implementation of numpy's meshgrid function for ndarray

[jreniel]

I have made an attempt to try to replicate the core functionality of numpy's meshgrid function, using ndarray.
My attempt, which seems to be yielding identical results to the numpy version is hosted here.

My hope is that this function, or an equivalent, makes it to ndarray, since I feel like meshgrid is one of those core numpy functions that we tend to use repeatedly in my field, and it would be an invaluable addition to the ndarray crate.

To be clear, I'm asking to not necessarily use the same function that I've posted here, I'll be happy too if the ndarray devs implement a better solution. All I would like to see is to have this function be a part of the ndarray collection of functions.

[ivnsch]

Any updates on this? I also need meshgrid in rust.

[akern40]

@jreniel, thank you for the reference implementation! @ivnsch and jreniel, do your applications require mesh grid with an arbitrary (i.e., run-time) number of arrays? This dictates whether we can maintain the dimensionality of the output arrays in the type system, or whether we have to resort to ArrayD like jreniel's approach. Let me know; I have a working implementation that works for up to six meshgrid inputs and avoids heap allocations, and I think I can make it work for dynamic-dimensional arrays if needed.

[ivnsch]

@akern40 sounds great! my current use case requires only a fixed number of arrays (3).

[jreniel]

Hi @akern40 , sounds like great news that you are working on this!

I work with ocean and atmospheric spatial data, We often use 2D meshgrids for representing surface data, such as building rasters for bathymetry or other types of data, but the need to represent 3D water/atmos column data is also present and common.
What I can think of is that it would be nice to have both meshgrid2D and meshgrid3D versions. ND is probably an overkill, but who knows if anyone would need that in practice? (maybe some mathematicians involved in higher dimensional computations/numerics?)

Thanks again, I'm excited to know that the Rust language just keeps growing! =)

[akern40]

Thanks to you both for the quick responses! I'll see if I can get a PR open to add this over the weekend.

[akern40]

Alright, draft is up; I'd written it to be compatible with the array reference implementation in #1440, so we need to get that merged in first. But if you need it immediately, you can use the branch in #1477; it should otherwise pass all CI/CD.