Implement * (or other unicode symbol) as in xarray about dimensionaldata.jl HOT 14 CLOSED

That is interesting. Shouldn't be too hard to implement, feel free to have a go.

Just remember that most of the core tools are defined in primitives.jl, you shouldn't have to do much yourself.

from dimensionaldata.jl.

I'm wondering if this is compatible with current behaviour of * on arrays. Will there be any circumstances where * will currently work (ie not error because of size mismatch etc) but this change will produce a different result?

If so we may need to add a method for this.

from dimensionaldata.jl.

I have concluded that using the symbol * is not a good idea, as it will naturally try to compare itself with the behavior of pure arrays, which I don't think makes much sense. I will use the symbol \otimes instead.

from dimensionaldata.jl.

Would it be possible to generalise this a little more? We could write this so you can us +,-,/,|,& etc as well using the same methods. Something like:

somemethod(f, A, B)

Where f can be *, or anything else that accepts two args. But I'm not sure what to call the method...

from dimensionaldata.jl.

dimensionwise ?

from dimensionaldata.jl.

That gets the meaning across really well. dimwise would be shorter but maybe less clear, I'm not sure.

from dimensionaldata.jl.

Hi, I was going to start working on that today, but I saw that there is #48 and #49 , which are not directly relevant, but still good to know and should be taken into account to write the most clear source code.

So I'll wait for them to be done and then I'll start doing this next week. In the meantime I'll start a Documentation PR that builds more proper docs based on Documenter.

from dimensionaldata.jl.

nice, I've solved this. Implemented it and tested it in my code and it works. Will PR it once #48 and #50 are merged, because I don't want to be constantly doing rebasing stuff.

from dimensionaldata.jl.

@Datseris, when you say A * B, is that python syntax or have you translated? I like the idea of being able to use xarray's A * B (I'd call this "broadcasting"), but I'm not as much a fan of their A @ B.

from dimensionaldata.jl.

A*B is Python syntax, but as I said elsewhere it does radically different thing that what * does here now, after Raf merged the * PR. Once you are done with #48 we will get it here as well, but not with *, with dimewise instead.

from dimensionaldata.jl.

Ah, I had thought you were thinking of matrix multiplication. I think xarray's A * B would make more sense as A .* B in Julia. So, something like @dimwise A .* B for automatic alignment, since it's difficult to pass .* as an argument to a function.

from dimensionaldata.jl.

Ah, I had thought you were thinking of matrix multiplication. I think xarray's A * B would make more sense as A .* B in Julia. So, something like @dimwise A .* B for automatic alignment, since it's difficult to pass .* as an argument to a function.

I like just passing * and avoiding the macro. Did you have a go at this @Datseris?

from dimensionaldata.jl.

Yeah I have this solved in my own code. I have implemented the dimwise function as originally described, it works for any binary f: dimwise(f, A, B), where f = *, -, +, ...

from dimensionaldata.jl.

dimwise was implemented in #151

from dimensionaldata.jl.