ncol and nrow in Legend object

[napsternxg]

Is there a way in Bokeh to have the legend in a multi row column format like it is done in matplotlib.

In matpltolib the legend object takes as input 2 additional optinal agruments ncol and nrow which defines how many rows and column the legend object should be split in.

This is very useful when we have lots of legend items and its better to organize then into rows and columns.

Can similar api be made available in bokeh.

[bryevdv]

This does not exist currently, but there was a recent PR that added support for horizontal orientations for legends. It might be nice to support more options like rows and columns as a feature, though. Would you be interested on working on a PR to add this capability, with some help?

[napsternxg]

@bryevdv Can you point me to the PR for horizontal orientation. I can try working on this using the approach from that PR.

[bryevdv]

@napsternxg absolutely, the merged PR is here: #3731

I would say from the standpoint of the current interface, since there is already orientation now that can be horizontal or vertical, the most consistent thing would be a generic integer property grouping that would specify how many rows (when orientation is horizontal) or how many columns (when orientation is vertical). It would have a default value of 1 which would keep the current behaviors as the default. In the python code this would just be a new class attribute for Legend, something like:

    grouping = Integer(default=1, help=""" helps string here """)

The BokehJS side will be a little tricker. Basically the current code uses the orientation to compute some layout values here:

https://github.com/bokeh/bokeh/blob/master/bokehjs/src/coffee/models/annotations/legend.coffee#L40-L49

and then uses those values to actually render text and glyphs in the render function here:

https://github.com/bokeh/bokeh/blob/master/bokehjs/src/coffee/models/annotations/legend.coffee#L118-L140

Those two sections would need to be expanded to take grouping into account. Take a look at that, if it seems like something you can do/would like to do with some help, we will certainly answer any questions you have.

[guillochon]

Hi, I have a legend with many labels, and neither the default horizontal/vertical orientations work for me, I need multiple columns/rows. Has there been any progress on this?

[birdsarah]

@guillochon, we're super thin at the moment and I don't think this high on the to do list - sorry to not have more news.

[brandondube]

Any update over a year later?

[bryevdv]

@brandondube if there were any updates, they would be in this issue (or a linked PR). The story here, as is the story with most all OSS projects, is that there is far more work to do, than people to do it. In such a situation, things come down to prioritization, and therefore some things potentially taking a very long time, depending on those priorities. The only solution to speed things up is for more people to get involved. I am happy to mentor any new contributors who would be interested in working on this.

[brandondube]

Well, how do I get started? I couldn't even find the definition of Legend in the repo digging around fairly quickly in /bokeh/models.

[bryevdv]

@brandondube Awesome, first things first, Bokeh is a dual language library, with most of the actual work done in BokehJS. To work on this particular issue will require being able to modify and build BokehJS. The developers guide has a section on getting set up:

https://bokeh.pydata.org/en/latest/docs/dev_guide/setup.html

Once you can build BokehJS, you will find the python side definition of Legend here:

https://github.com/bokeh/bokeh/blob/master/bokeh/models/annotations.py#L99

Bokeh classes are almost all purely declarative, they just have typed property class attributes that define how they can be configured. For this issue, we will want to add probably a new property to specify how the legend layout is to be configured (or possibly, we might just want to expand the existing orientation property)

The BokehJS definition of Legend (where all the work happens) is here:

https://github.com/bokeh/bokeh/blob/master/bokehjs/src/lib/models/annotations/legend.ts

Once you are set up with building and running a locally modified BokehJS we can discuss particulars of how best to proceed.