add non-array version for hline and vline

[daschw]

Not sure if we want this but this would be an easy workaround only for the hline/vline shorthands for #2129 and #2575.

allows e.g.

plt = plot(rand(10))
hline!(0.5)
vline!(plt, 7)

I am really sceptical about this, because

plot(1, st = :hline)

does something different from

hline(1)

[BeastyBlacksmith]

I don't think

plot(1, st = :hline)

is often seen in the wild. I would be in favor of this change.

[PhilipVinc]

I'm nobody,
but I'd love this change. It always bites me that it does not work.

[daschw]

@mkborregaard You seemed to have some doubts about this in #2575. What's your opinion on this?

[mkborregaard]

I agree with your scepticism. I generally don't like when we make workarounds that distort our ability to reason about things, so plot(4, st = :hline) should be equivalent to hline(4).
If we really want this, I think we should deprecate the plot(4) functionality, which is poorly documented anyway. I tend to use it actually, to create a plot with 4 subplots and plot to them iteratively, but I'd be OK with changing the interface to plot(num_subplots = 4), as it is actually a perversion of Plots logic itself that plot(4) passes a positional parameter that isn't the data. Breaking of course, but seems harmless.

[daschw]

I'd also prefer a more general approach without special-casing different series types. However, hline and vline are special cases anyway regarding the positional input arguments. hline(x, y) and vline(x, y) would both ignore x and for vline y would correspond to the x axis.

I tend to use it actually, to create a plot with 4 subplots and plot to them iteratively, but I'd be OK with changing the interface to plot(num_subplots = 4)

Can't you just do plot(layout = 4) or do you want to push! to initialized series?

[isentropic]

Perhaps array versions could be named vlines, hlines?

[mkborregaard]

Can't you just do plot(layout = 4) or do you want to push! to initialized series?

🤯

I guess we really could just deprecate plot(4) then. But would you then make passing a scalar automatically wrap that in a 1-lenght array? It just feels so... R-like.

[yha]

I'm in favor of deprecating the current meaning of plot(::Int) and promoting scalars to 1-length vectors regardless of series type. I think this can be done pretty easily in a recipe.
I think this is a very natural interpretation in scatter(x,y) as well as hline/vline, and it seems in line with the Plots approach of providing a flexible intuitive interface with many ways of doing the same thing. Seems harmless to me, especially compared to other ways Plots already tries to guess the user's intention, like the different handling of vectors-of-vectors versus vectors-of-tuples.

I tend to use it actually, to create a plot with 4 subplots and plot to them iteratively, but I'd be OK with changing the interface to plot(num_subplots = 4)

But plot(4) currently creates 4 series in one subplot, not 4 subplots...