axvline, axhline, shade_between

[arsenovic]

i use these functions frequently to highlight regions of interest, bounds, etc.

are there any plans to support these in mpl_exporter/plotly/mpld3. im not familiar enough with the architecture to know where exactly the information is being lost.

@theengineear

[theengineear]

Yup. Plotly has been mostly trying to deal with visualizing actual data, but having objects on a plot can definitely assist in understanding the data. This will soon be added on the Plotly side.

We're actually going to be helping out with some matplotlib development for 1.5.x to refactor how figure representations are exported there. Currently, it's impossible for us to know whether a line in a matplotlib plot is data or extra information. However, once this refactor is done, the high-level plot semantics will be included and we'll be able to properly handle this conversion.

I'll leave this open, but this probably won't see much improvement for a month or so while this project gets up an running. I'll definitely post something here when these things are better handled.

[arsenovic]

nice. thanks

[bryevdv]

@theengineear please ping me (@bryevdv) or @damianavila if we can help out.

[theengineear]

@bryevdv, will do! I expect to start working on this project end of this week or early next.

[jakevdp]

The problem with axhline and axvline is that they involve blended transforms: that is, one direction follows the data, while one direction is attached to the screen coordinates. Without a full-featured transformation representation, they will be very difficult to add to mplexporter or the derived projects. Plotly might be able to do this (I'm not sure), but I don't see mpld3 getting there any time soon, unless someone has some significant time to invest in it.

When you say "shade_between", do you actually mean "fill_between"? If so, that's already supported in mpld3.

[aflaxman]

How about sticking a “x follows data” flag of some sort in the json, and then mpld3 can continue to do it wrong until someone gets around to using it?

[theengineear]

@jakevdp, is the problem in mplexporter from matplotlib in that you can't run something like contains_branch on the transformation? I.e., is it that it's (1) not possible at all to know what the blended transformation is, or (2) that there's no usable subset of blended transformations that the current projects could leverage?

Right now Plotly can anchor to x in one dimension and what we call paper in another. However, like I said, there's not yet support for arbitrary objects (e.g., a horizontal line).

[danielballan]

Related: MEP 25

[jakevdp]

I ran into some problems with contains_branch earlier - it only works for instantiated transforms, not transform classes. This means you can't just do contains_branch(BlendedTransform): you actually have to check for an instantiated BlendedTransform, but if you had access to that, you wouldn't have to check the branch for it!

Anyway, there's nothing fundamentally that keeps you from getting mixed transform information from matplotlib: it's just that we haven't implemented that yet. Implementing that would take some careful thought regarding the design, though, so that we don't code ourselves into a corner later on.