CustomJSTransform

[birdsarah]

Adding new transforms is easy, but people should be able to self serve something like this:

from bokeh.models import CustomJSTransform

transform = CustomJSTransform.from_coffeescript(
    compute="""return 2*x""",
    v_compute="""
        vals = []
        for x in xs:
            vals.push(2*x)
        return vals
    """
)

.....

p.circle(x='x', y='y', size={field: 'x', transform: transform})

Just like #4659, user would know that single x is passed to compute and array xs is passed to v_compute. User would specify a return value and would also be able to pass in args - like for CustomJS.

[birdsarah]

ping @hafen

[bryevdv]

This seems low-effort / high value so scheduling it for 0.12.5

[souravsingh]

I am interested in working on the issue. How do I start?

[bryevdv]

HI @souravsingh it's going to be very similar to the CustomJS callback. I'd first suggest studying that, and it's JS implementation as well:

https://github.com/bokeh/bokeh/blob/master/bokeh/models/callbacks.py
https://github.com/bokeh/bokeh/blob/master/bokehjs/src/coffee/models/callbacks/customjs.coffee

It's probably the case that alot of that code can be re-used.

Then the first step would be to define what the python class that will represent a CustomJSTransform should look like. It will be a subclass of Transform with some extra properties to represent the JS code for the transform.

[datnamer]

Can we do this with pyscript also?