interaction between tooltips and gridplots

I tried to build a grid plot involving tooltips. The result was that, if the tooltips display was not contained in the figure associated to it, it was truncated by the figure to the right and below as in the following example. However, it does not get truncated by figures above or to the left:

Screenshot

This one is truncated

This one is not:

Code

import numpy
from bokeh.plotting import figure, show, output_file
from bokeh.plotting import figure, output_file, show, ColumnDataSource
from bokeh.sampledata.iris import flowers
from bokeh.io import gridplot
from bokeh.models import HoverTool

df = flowers.copy()
df['cat_col'] = numpy.round(flowers.petal_width * 2)

colormap = {'setosa': 'red', 'versicolor': 'green', 'virginica': 'blue'}
colors = [colormap[x] for x in df['species']]

result = []
for species in df['species'].unique():
    sub_res = []
    for cat in df['cat_col'].unique():
        sub_df = df[(df['species'] == species) & (df['cat_col'] == cat)]
        colors = [colormap[x] for x in sub_df['species']]
        hoverdata = ['species:{} - cat:{}'.format(species, cat) for x in sub_df['species']]
        source = ColumnDataSource(data=dict(hovercol=hoverdata))

        hover = HoverTool(tooltips=[("Hover col", "@hovercol"),])
        p = figure(tools=[hover])
        p.circle(sub_df["petal_length"], sub_df["petal_width"],
                 color=colors, fill_alpha=0.2, size=10, source=source)
        sub_res.append(p)
    result.append(sub_res)


output_file("iris.html", title="iris.py example")

show(gridplot(result, plot_width=200, plot_height=200))

Bokeh version

Python version : 3.4.3 |Continuum Analytics, Inc.| (default, Dec 1 2015, 11:59:33) [MSC v.1600 32 bit (Intel)] IPython version : 4.0.0 Bokeh version : 0.12.1

[birdsarah]

Another option which doesn’t fix it, but may help, would be to implement hover in a fixed position, like on this leaflet demo: http://leafletjs.com/examples/choropleth.html

[bryevdv]

My guess is, if Bokeh on the Python side is upgraded to 2.0.0, then BokehJS on the JS side should also be 2.0.0 to be able to parse the JSON but I could not find a corresponding version for BokehJS to test

Yes versions definitely JS/Python need to match, in general. Not just between major releases, either. Are you intending to hard code a fixed BokehJS version but accept JSON output from arbitrary Python Bokeh versions? I would not regard that as reliable.

Do you plan to also release a BokehJS 2.0.0rc1 on npm for downstream projects to test Bokeh embedding ?

I had not planned on it. Until now, we only publish full releases to NPM. This is certainly something we can consider changing in the future, especially if there is increased interest in the NPM package. Given the compressed schedule to 2.0 this week I am not sure it’s worth it for me to scramble to do something different in the middle of release week. But we can have a very quick 2.0.1 if there are any follow-on issues.

Do you eventually have a migration guide to 2.0.0 ?

Each release has its own (if any) migration notes:

https://docs.bokeh.org/en/dev/docs/releases.html

That said, while we do our best, we only put in what we think to put in based on the use cases we know about, so there may be things that are missed for some people.