WIP: Fonts

[gustaphe]

I started fiddling with a redefinition of the fonts system (as discussed in #3463)

It's currently in a state where it compiles and plots, but I've only done just enough adjustment in the gr backend to get it in "working" order. But I think it illustrates the idea.

Maybe the gains aren't that big, and there's probably quite a bit of work left, but the shortening of the attribute list feels pretty good.

plot(randn(10); font=(:green, 45.0))

[isentropic]

Perhaps I'm a little late to this discussion, but what are the clear advantages of separating font attrs into a new named tuple therefore breaking a lot of existing recipes (perhaps most of them)? This is a massive breaking change, but what are the clear advantages here?

Additionally, this has to be implemented for all backends or else other backends would just stop working

[gustaphe]

Not at all, it's an infant idea, but I wanted to see what it would look like.

Advantages I see:

• Shortening attribute list -- each text object currently has 6 attributes by a specific pattern (unless control of any of them is missing): *family, *pointsize, *rotation, *color, *halign and *valign. All of these attributes are already encapsulated in the Font type
• Guaranteed uniformity -- those text objects that currently are missing certain controls won't, and any text objects added through a scheme like this won't either
• Clarifying and simplifying font dependencies -- instead of specifying where xguidefontfamily, xguidefontsize... are inherited from separately, one simply specifies that guidefont is font, but with a specific pointsize, etc.

That said, it is a breaking change, and quite a bit of details to implement. It might be possible to implement it in a non-breaking fashion, where font=(45.0, "Computer Modern") is a convenience for fontrotation=45.0, fontfamily="Computer Modern", or vice versa.

[BeastyBlacksmith]

You may also look at #2854 where I wrote a macro to traverse into any Plot specific types (including Font) and creating the attributes automatically (incl. aliases).
I think it could also work non-breaking if we can make sure, that aliases work for all kinds of recipes.