Using bokeh's Arrow to draw VectorField?

So I found the drawing of the vectorfield arrows in the bokeh backend to be a bit awkward (mostly in looks and lack of styling options) and tried to replace it by bokeh’s Arrow through a CompositeElementPlot. Problem is (which is maybe why you didn’t do that in the first place…) that bokeh’s Figure does not expose Arrow, which is an annotation, not a marker or glyph (if I understand correctly), and hence drawing arrows in holoviews has to pass by style options, not data (again, if my understanding is correct).

Question is, do you think that could be addressed on the Holoviews side, or would one have to make a PR on the bokeh side? I briefly skimmed some related discussions on here from when you were implementing VectorField, but that’s a while ago and I couldn’t find any information on this in particular.

[philippjfr]

Here’s how to use a vectorized Arrow:

p = figure(x_range=(0, 10), y_range=(0, 10))

cds = ColumnDataSource(data=dict(x_start=[0,1, 2], y_start=[0,1, 2], x_end=[1,3, 5], y_end=[0,5, 8]))
p.add_layout(Arrow(end=OpenHead(), source=cds, x_start='x_start', y_start='y_start', x_end='x_end', y_end='y_end'))

pn.Column(p)

[badiku]

like this? seems no problem:

p = figure(x_range=(0, 10), y_range=(0, 10))
cds = ColumnDataSource(data=dict(x_start=[0,1, 2], y_start=[0,1, 2], x_end=[1,3, 5], y_end=[0,5, 8], line_width=[16,5, 8], color=['red','blue','yellow']))
p.add_layout(Arrow(end=OpenHead(), source=cds, line_color='color', line_width='line_width', x_start='x_start', y_start='y_start', x_end='x_end', y_end='y_end'))
show(p)