Plots.jl: Way to pass keywords to the backend

[BeastyBlacksmith]

It would be nice to have a way to pass backend-specific keywords.
One idea is, to have a backend_options keyword that takes a Dict or NamedTuple that gets passed as keywords to the backend.

[mkborregaard]

Do you feel like experimenting with a PR for one of the backends? It's been suggested several times.

[BeastyBlacksmith]

I did take a quick look on it, but the backend <-> frontend interface is not as clear-cut, as i had imagined and I get lost in the Plots pipeline quite quickly. So at least till the beginning of next year I won't have time.

[mkborregaard]

Have a look how this keyword arg was added and passed through to PyPlot https://github.com/JuliaPlots/Plots.jl/pull/1009/files

[BeastyBlacksmith]

Actually I see, that extra_kwargs does already exist, it is "just" not passed at the moment.
Edit: I mean it is passed as it should be available to the backend code, but i thought there would be some way to pass that directly to a backend plot interface function which handles it.

[BeastyBlacksmith]

Maybe it is at the wrong spot though, since it belongs to the Plot object, but as far as I can see, the backends operate on the Subplot level?

[mkborregaard]

I'm not sure what you mean by that?

[BeastyBlacksmith]

Let me put it as a question: How would a backend query the plotattributes?

[mkborregaard]

The backend processes the Plot object and builds up a series_list containing the keywords passed to the backend. So I guess the extra_kw args should just be appended to the series_list without any transformations.

[BeastyBlacksmith]

A bit of a memo to myself :

• extra_kwargs needs to be actually be populated
• this requires different handling of unsupported kwargs, like only checking the Plots specific kwargs and put the rest in extra_kwargs

[isentropic]

perhaps:
plot(x1, x2, backendkeyword="parameter=value")
aliases:
plot(x1, x2, bk="parameter=value")
plot(x1, x2, backend_keywords="parameter=value, parameter2=value2")

There would be some effort parsing "parameter=value, parameter2=value2", but should be fine I think, I can try implementing this with pyplot

[BeastyBlacksmith]

I think it would make most sense if you were checkout #2200 and see, if you can make it work for pyplot too.
I don't think we need special parsing here.