Plotly density_contour with log fails when contour lines span negative values

Question

I have this code

import pandas as pd
import plotly.express as px

data_df=pd.read_csv('my_file.csv')

fig=px.scatter(
    data_df,
    x="Income",
    y="Revenue factor",
    log_y=True,
    marginal_x="violin",
    marginal_y="violin",
    )

fig.show()

with this data

"Revenue factor","Income"
5.6,10538.024
5.8,10768.521
8.5,7951.655
7.3,9927.96
12.4,9154.182
22.9,9303.803
21.3,8742.111
6.2,8719.435
13.5,9104.053
6.5,10193.054
15.6,9402.679
3.8,10566.417
5,9792.008
7,11425.052
10.2,8284.642
4.6,10038.097
14,10201.633
6,9682.332
10.3,8784.033
7.3,9067.933
8.3,10074.925
5,10820.537
15.3,8456.473
14.9,8883.006
5.9,12202.999
14,8575.485
16.1,9881.27
17.9,9759.649
6.6,10005.727
15.3,10237.755
5.1,9844.915
10.5,11673.672
8.62,10565.106
8.62,11057.078
11.7,10543.177
5.8,10729.003
9.8,11830.967
13.2,9315.292
2.6,11895.213
8.15,8306.931
21.7,7595.944
11.6,9333.204
19.6,7798.425
16.6,9498.777
9.35,8924.454
6.99,9787.287
4.31,10272.085
8.99,8121.588
10,9754.302
5.6,12085.931
7.4,9459.656
5.9,11703.813
21.9,11286.706
5.6,12151.199
5.1,11643.14
14.2,9952.406
6,10882.89
4,9745.805
8.9,8081.441
6.9,8943.834
2.7,12849.368
7.6,9669.406
12.1,8649.289
10.8,10571.735
13.2,9499.685
19.5,8271.722
14.5,9184.476
11.09,10500.022
7.12,10669.6
4.28,12404.747
11.44,10194.588
8.51,11920.15
10.91,9688.818
9.2,10361.284
18.6,8933.945
8.2,9414.157
16.3,9730.937
6.1,11125.076
9.3,10445.249
12.4,9597.09
12,8426.657
9,11216.912
10.5,10178.949
6,11366.618
2.8,10997.343
2.8,11964.936
10.8,11190.276
20.7,9591.806
4.9,11799.541
8.5,10342.593
6.7,12571.491
8.7,11705.411
15,9961.351
10.2,10329.058
13.5,11870.332
13.6,10501.098
11.1,9791.326
10.4,9483.818
5.7,11337.888
13.7,8817.618
13.43,9054.572
14.54,10975.658
14.04,9583.998
16.23,9731.019
18.82,9221.659
8.42,11174.223
10,8836.596
19.39,10512.63
14.94,10773.545
7.33,9137.491
4.29,14067.74
3.84,11106.177
11.08,9485.621
24.72,8491.606
13.1,10059.06
12.9,11210.718
8.8,11038.553
32.4,6994.603
14.6,11286.141
3.9,12330.234
13.6,10402.071

which nicely produces this plot

Now I would like to use px.density_contour so I just change

import pandas as pd
import plotly.express as px

data_df=pd.read_csv('my_file.csv')

fig=px.density_contour(
    data_df,
    x="Income",
    y="Revenue factor",
    log_y=True,
    marginal_x="violin",
    marginal_y="violin",
    )

fig.show()

and I get this:

which is clearly wrong. The problem is in the log scale and the contour lines spanning negative values, as seen in this plot produced just by removing log_y=True:

Is this a bug? Limitation of Plotly? Any solution? Note that none of my points is negative.

If I have time, I'll take a closer look into this but as you mentioned, it's probably an unintended consequence of the contour lines spanning negative values. If this is indeed what is happening and what causes the entire contour plot to fail to render, a "naive" way to correct this would be to produce `NaN` for portions of the contour line where you are evaluating log(y) for y<= 0 — Derek O, May 05 '22 at 17:16
@user171780 I am running into the exact same problem. Would you mind opening your example here as an issue at https://github.com/plotly/plotly.py/issues AFAICS there are no related issues reported yet. — roble, Jun 23 '22 at 12:55

score 0 · Answer 1 · answered Jun 10 '23 at 23:33

I just ran into the same trouble.

I put together a workaround that has two parts.

PART 1:
Replace the value to plot logarithmically with the log of the value (manually do the logarithm) and set log_x/log_y for that axis to False.

Just to be clear, for you, the workaround would be

set log_y=False
replace the data in the first column with its logarithm

Your data would change from

"Revenue factor","Income"
5.6,10538.024
5.8,10768.521
8.5,7951.655
7.3,9927.96
12.4,9154.182
22.9,9303.803
...

to

"Revenue factor","Income"
0.748,10538.024
0.763,10768.521
0.929,7951.655
0.863,9927.96
1.093,9154.182
1.359,9303.803
...

PART 2:
If you want to put custom labels on that axis to match taking the logarithm, you can consult
https://plotly.com/python/tick-formatting
or
Plotly: How to customize xaxis tick labels using plotly express?

An alternative workaround might be to customize how plotly.express does contours, if that can keep the contours from going into the region with negative values, but I have not attempted that.

Plotly density_contour with log fails when contour lines span negative values

1 Answers1