Parallel categories not recognizing dimension columns

[robroc]

Data frame with two categorical columns and one numerical column being passed to px.parallel_categories. Only one column gets visualized, no matter what is passed to dimensions: a single column, two columns, or no columns.

Data types changed to str and category, but no change.

Data:

Code:

px.parallel_categories(occupations_by_year, dimensions=['Main_jobs', 'Election year'], color = 'count')

Result:

Python 3.6
plotly==3.10.0
plotly-express==0.3.0

[nicolaskruchten]

How many unique values of main_jobs are there?

[robroc]

About 30 unique values. I was hoping to make each year a node in the chart and the jobs the bands between them, a bit like a Sankey. Is this possible?

…

On Thu, Jun 6, 2019, 4:48 PM Nicolas Kruchten ***@***.***> wrote: How many unique values of main_jobs are there? — You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub <https://github.com/plotly/plotly_express/issues/98?email_source=notifications&email_token=AAJJTOMGAKJHJALXRSHQCGTPZFZYBA5CNFSM4HVHS2J2YY3PNVWWK3TUL52HS4DFVREXG43VMVBW63LNMVXHJKTDN5WW2ZLOORPWSZGODXEDXCQ#issuecomment-499661706>, or mute the thread <https://github.com/notifications/unsubscribe-auth/AAJJTONC4GQ72KJHUBPNAZLPZFZYBANCNFSM4HVHS2JQ> .

[nicolaskruchten]

Yes, but I'll have to generalize the cutoff heuristic, which right now caps the number of values at 20 IIRC. This a little blunt for cases like this where you only have 2 dimensions. It's reasonably if you had e.g. 10 dimensions because in that case you might have up to 20^10 combinations :)

[jacobbaron]

I have encountered this issue as well. I have four columns, one of which has ~30 categories. There are about 1100 unique combinations in the data, so not ridiculous, but the column with the 30 categories is left out of the plot.

[jasonsross]

I 2nd the desire to set the cutoff of number of categories. Like these charts but I've hit this roadblock and I think I was just barely over the cutoff limit. I'm ok to group my categories down to reduce complexity, but wish to have a little more flexibility in doing so.

[nicolaskruchten]

I'll get this fixed in the next version :) I'll move this issue to the plotly.py repo to get it scheduled.

[fjprobos]

Same issue over here. It is an easy way to look for colinearity among categorical features. Please get it fixed.

unique transmission_model : 19
unique model_family: 24

Python 3.5.6
Plotly 4.4.1