Sunburst should take grandparents info

[rotemgb]

Sunburst should take grandparents info (as many as one want), to avoid identical ids passed to Sunburst, especially when working on categorical data.

For example, taking the sunburst plot example from the website:

If we would like to add gender on the second circle, and some of the salesperson initials would have been the same, some of the sunburst ids would also be the same, and it would be possible to plot.

This problem happens every time you have categorial data.

Thank you,

[emmanuelle]

Thank you for noticing this @rotemgb. I just hit exactly the same bug :-). We will update the doc.

[tayyabvohra]

@emmanuelle @rotemgb I am having this issue can you please resolve this , the issue is similar to this one..
#2285

[rotemgb]

As far as I know, this issue hasn't been resolved. I don't think @emmanuelle completely understood the issue. The problem is not with documentation, but the class of sunburst need to be modified to give flexibility when working on categorical data.

@tayyabvohra As for a work around the issue, I created a small loop that add to each category a space in the beginning or the end. for example, the string 'coffee' and 'coffee ' (with a space in the end) are not the same. That way sunburst "think" that they are different and manage to link each one of them to different children. This solution works well with small number of categories, but when you have too many, the plot will be a little messy, as the names of the labels can't be modified (as far as I know), and you will have the spaces you added in the plot.

I think 2 things should be added to sunburst:
**Give the users the flexibility to link not just the parents, but also the grandparents. That will solve the categorical data with sunburst issue.
**Add the flexibility to change the labels on the plot. meaning, sunburst shouldn't take automatically the name of the parents and children as the labels, and there should be a way for us to modify it.

@tayyabvohra hope that helps :)

[tayyabvohra]

@rotemgb can you share your code here for this problem..!

[nicolaskruchten]

So to be clear: id must be unique per segment and must have a unique parent. The calling code probably should indeed prepend the ids of the parents to the ids of the children, recursively, but this is not something that can be done internally using the id/parent/label API.

That said, we added the path API to px.sunburst to handle doing this automatically for tabular data. Check it out here: https://plot.ly/python/sunburst-charts/#sunburst-of-a-rectangular-dataframe-with-plotlyexpress

@emmanuelle do you think you could you see if you can make the docs clearer/more explicit about how unique IDs must be?

[nicolaskruchten]

Re the two proposed solutions:

Give the users the flexibility to link not just the parents, but also the grandparents. That will solve the categorical data with sunburst issue.

This is already possibly from outside Plotly: IDs should be prepended with parent information, recursively, and we will make this clearer in examples like https://plot.ly/python/sunburst-charts/#sunburst-with-repeated-labels

Add the flexibility to change the labels on the plot. meaning, sunburst shouldn't take automatically the name of the parents and children as the labels, and there should be a way for us to modify it.

This is already not the case: id and label are different fields per node. id defaults to label if not provided but both can be provided so a node can have a label="Female" for display, a parent="a-b" and an id="a-b-c". Another node can then be a child of this one with another label (even if that label is the same as some other sector :) and a parent="a-b-c" and an id="a-b-c-d".

We should make this distinction between id and label clearer in the docs as well (and repeat this clarification in the treemap docs).

Thanks for the feedback :)

[tayyabvohra]

Can you provide any code for this problem , I cannot find any solution related to the problem on the OP questions chart, although this problem can be solved with different type of sunburst chart provided in the documentation @nicolaskruchten

[rotemgb]

@tayyabvohra you can find my code with the work-around the problem in here https://www.kaggle.com/rotemgb/money-talks-gender-disparities-in-tech , but after seeing this solution by @nicolaskruchten using path, I think would solve your problem.

That said, we added the path API to px.sunburst to handle doing this automatically for tabular data. Check it out here: https://plot.ly/python/sunburst-charts/#sunburst-of-a-rectangular-dataframe-with-plotlyexpress

I don't have a code for that, because I haven't used it (I'm not sure this option existed when I used it..), but follow the code examples in this documentation.

[rotemgb]

@nicolaskruchten you all did a great adding the path functionality! That solves all problems with categorical data.

About the documentation, maybe adding an example on adding labels to your demo function def build_hierarchical_dataframe in https://plot.ly/python/sunburst-charts/#sunburst-of-a-rectangular-dataframe-with-plotlyexpress might help.

Thank you!