Add support in make_subplots for mutiple axes

[marcotama]

Hi, this is a feature request, not a bug, so I do not have a minimum example.

My use case is that I have been using make_suplots (to generate a quite simple 2x2 layout), and I would now introduce a second y axis in each subplot, in order to obtain the same effect described here: https://plot.ly/python/multiple-axes/

I could not find a way to do this through make_subplots, but I seem to understand that it is possible by setting the layout. My issue with this approach is that I am afraid to lose some "magic" that is happening in make_subplots.

Can you advise how to proceed? Is it something you are willing to implement?

[jonmmease]

Hi @marcotama,

Yeah, multiple y-axes in subplots is something we've been thinking about for version 4, but we haven't settled on an API yet. My current thinking is to add a new option to elements of the specs argument to make_subplots. Something like

from plotly.tools import make_subplots
fig = make_subplots(
    2, 2, specs=[[{'secondary_y': True}, {'secondary_y': True}],
                 [{'secondary_y': True}, {'secondary_y': True}]])

fig.add_trace(go.Scatter(...), row=1, col=1)
fig.add_trace(go.Scatter(...), row=1, col=1, secondary_y=True)

It's admittedly a little clunky, but it would get the job done without too much added API complexity. What do you think? cc @nicolaskruchten @chriddyp

[marcotama]

Hi @jonmmease,

It looks simple enough to use.
My only doubt is, does that mean if I pass 'secondary_y': True, the y axis "counter" increments by 2 instead of 1? Would it be possible to return, or have a helper function to retrieve, the axis names? Something like

from plotly.tools import make_subplots
fig, ax_info = make_subplots(
    2, 2, specs=[[{'secondary_y': True}, {}],
                 [{'secondary_x': True}, {'secondary_y': True}]])
print(ax_info)
>>> [[('x','y','y2'),('x2','y3')],[('x3','x4','y4'),('x5','y5','y6')]]

I know this can be figured out manually, but I fear it is fertile ground for bugs, especially in more complex setups.

Edit: funnily enough, I made a mistake with counting axes while writing the above example.

[jonmmease]

We would at least add the second y-axis label to the fig.print_grid() output. Would that be enough or do you have use cases in mind where you want programmatic access to the y-axis labels?

This info is all in the internal fig._grid_ref property, but it's not directly exposed in the API right now except through fig.print_grid.

[marcotama]

I think print_grid is good enough. Thanks!

[jonmmease]

Hi @marcotama,

I'd appreciate your feedback on #1564 if you have a chance to take a look and give it a spin. Installation instructions at #1564 (comment)

[marcotama]

Hi @jonmmease , I left a comment on the PR page. Apologies for the late reply, I have been busy!

[jonmmease]

Done in #1564