Heatmap should look up category position when building brick data

[colinfang]

The 2nd plot should start at Team B, instead of Team A.

fig = make_subplots(rows=2, cols=1, shared_yaxes=True, shared_xaxes=True)

z = [[.1, .3, .5],  
     [1.0, .8, .6],
     [.6, .4, .2]]

x = ['Team A', 'Team B', 'Team C']
y = ['Game Three', 'Game Two', 'Game One']

fig.append_trace(go.Heatmap(z=z, x=x, y=y), 1, 1)

z = [[.3, .5],  
     [.8, .6],
     [.4, .2]]

x = ['Team B', 'Team C']
y = ['Game Three', 'Game Two', 'Game One']

fig.append_trace(go.Heatmap(z=z, x=x, y=y), 2, 1)
iplot(fig)

fig.data

[{'type': 'heatmap',
  'x': ['Team A', 'Team B', 'Team C'],
  'xaxis': 'x1',
  'y': ['Game Three', 'Game Two', 'Game One'],
  'z': [[.1, .3, .5],  
     [1.0, .8, .6],
     [.6, .4, .2]],
  'yaxis': 'y1'},
 {'type': 'heatmap',
  'x': ['Team B', 'Team C'],
  'xaxis': 'x1',
  'y': ['Game Three', 'Game Two', 'Game One'],
  'z': [[.3, .5],  
     [.8, .6],
     [.4, .2]],
  'yaxis': 'y2'}]

fig.layout

{'xaxis1': {'anchor': 'y2', 'domain': [0.0, 1.0]},
 'yaxis1': {'anchor': 'free', 'domain': [0.575, 1.0], 'position': 0.0},
 'yaxis2': {'anchor': 'x1', 'domain': [0.0, 0.425]}}

https://codepen.io/anon/pen/qqdqWq

[etpinard]

replacing

x = ['Team B', 'Team C']

with

x = [None, 'Team B', 'Team C']

should do the trick.

[colinfang]

It doesn't work in JS.

I hope this is a bug instead of expected behaviour, as in Bar plot the shared axis do align on the name.

[etpinard]

Ha I see. You're expecting each z column to shift in accordance with x.

I believe this is the result of the same underlying issue as #1097

Here's a way to work around the problem: https://codepen.io/etpinard/pen/QGbGNP

[etpinard]

@colinfang thanks for posting 🍻

Next time, to make our lives a little easier, please don't make references to the plotly.py API (or any other API library) in this repo. It can make things a little confusing on how to approach the issue. A reproducible codepen suffices.

[etpinard]

Update: solving this problem and #1097 in general is a little harder than I thought.

At the moment, the heatmap plotting routine fails when x and y are unsorted (even when x and y are numerical - see http://codepen.io/etpinard/pen/RoKGpj)

So, I think we'll need to order x and y and rearrange z in the calc step. From heatmap/plot.js this issue isn't news.

    // TODO: if there are multiple overlapping categorical heatmaps,
    // or if we allow category sorting, then the categories may not be
    // sequential... may need to reorder and/or expand z