sankey: introducing attributes (node|link).customdata

[antoinerg]

Closes#4243 by introducing attributes node.customdata and link.customdata as suggested in #4243 (comment).

[etpinard]

Nice!

I see that customdata is currently part of the sankey root-level attributes:

Maybe we should remove it, like we currently do for parcats:

plotly.js/src/traces/parcats/attributes.js

Lines 217 to 218 in 80262b5

	// Hide unsupported top-level properties from plot-schema
	customdata: undefined,

[emmanuelle]

Thank you @antoinerg for the great reactivity :-). Would it make sense to add a test where the node or link customdata has more than one dimensions, to check that it's possible to display several values from the customdata in the hovertemplate?

[antoinerg]

Maybe we should remove it, like we currently do for parcats

Done in 63ef3e9

[antoinerg]

Thank you @antoinerg for the great reactivity :-). Would it make sense to add a test where the node or link customdata has more than one dimensions, to check that it's possible to display several values from the customdata in the hovertemplate?

@emmanuelle what do you mean by more than one dimension? Right now, I made it so that if customdata is an array, accessing customdata from an hovertemplate will return the value at this point's index (ie. customdata[pointNumber]). Is that what you're after?

[emmanuelle]

@antoinerg yes. I just wanted to be sure that if customdata = [['a', 'b'], ['c', 'd']], it is possible to access either a with customdata[0] in the template, or b with customdata[1], etc.

[antoinerg]

@antoinerg yes. I just wanted to be sure that if customdata = [['a', 'b'], ['c', 'd']], it is possible to access either a with customdata[0] in the template, or b with customdata[1], etc.

Thanks for the clarification! I'm not sure this is handled properly. I'll add a test :)

[antoinerg]

@emmanuelle tested to work properly in c43873a

[alexcjohnson]

Suggested change

	'Assigns extra data each node.'
	'Assigns extra data to each node.'

[alexcjohnson]

Suggested change

	'Assigns extra data each link.'
	'Assigns extra data to each link.'

[alexcjohnson]

Looks good! 💃

[jzamalloa1]

Thank you very much all for working on this much needed feature! Do you know if this (node|customdata, link|customdata) will be pushed to the plotly sankey python library? That would be extremely useful, thanks!

[archmoj]

I think so.

[alexcjohnson]

@jzamalloa1 you mean the sankey trace type in the plotly package? Yes, the next release of the package - due out within the next week I would expect - will include this feature.

[xanderwallace85]

@jzamalloa1 you mean the sankey trace type in the plotly package? Yes, the next release of the package - due out within the next week I would expect - will include this feature.

Great! Same for the python library?

[antoinerg]

Great! Same for the python library?

Yes, the Python library will also get the improvements very shortly.

[xanderwallace85]

Great! Same for the python library?

Yes, the Python library will also get the improvements very shortly.

Any news about the updated Python library? :(

[PGrothaus]

It's in the new python release 4.6.0. Thanks a lot everyone involved, it's immensely useful!

[xanderwallace85]

It's in the new python release 4.6.0. Thanks a lot everyone involved, it's immensely useful!

Hi @PGrothaus,
I believe the problem is not completely solved in plotly. For instance, regarding the sunburst, the case to follow doesn't work properly:

customdata=np.dstack((a, b)),
hovertemplate='<br>%{customdata[0]:.3f} <br> %{customdata[1]:.3f} ',

However, if I have simply

customdata=a,
hovertemplate='<br>%{customdata:.3f} <br>',

then everything works smoothly.

[antoinerg]

@xanderwallace85 I think it would be better if you opened a separate issue for the sunburst case you're describing!

Thank you for using plotly.js :)