Bokeh class definition for `_ModelInDocument` has either moved or been deprecated, import needs to be updated

In Bokeh 12.5dev14 release, the class _ModelInDocument is no longer defined in bokeh.model. This is required for datashader.bokeh_ext.InteractiveImage. Here is the traceback:

---------------------------------------------------------------------------
ImportError                               Traceback (most recent call last)
<ipython-input-7-9d4688ff35b4> in <module>()
      1 import datashader
----> 2 from datashader.bokeh_ext import InteractiveImage

/Users/schowell/data/code/datashader/datashader/bokeh_ext.py in <module>()
      8 from bokeh.io import notebook_div
      9 from bokeh.models import CustomJS, ColumnDataSource, Square, HoverTool, GlyphRenderer
---> 10 from bokeh.model import _ModelInDocument as add_to_document
     11 from bokeh.io import _CommsHandle
     12 from bokeh.util.notebook import get_comms

ImportError: cannot import name '_ModelInDocument'

There is a function definition for _ModelInDocument in bokeh/embed.py but it is different code. I am not sure where this got moved to but it breaks the bokeh extension for interactive datashader images. @bokeh/core was this class definition moved or deprecated?

[StevenCHowell]

Bokeh has been updated to 0.12.5, while datashader remains at vanilla 0.4.0 and therefore none of the examples can be used.

@sbanjac you can install specific versions of packages, e.g., conda install bokeh=0.12.4. You should have a look at creating conda environments. I just created a conda environment with datashader, Bokeh 0.12.4, and Jupyter, which I could use to generate bokeh/datashader plots. You would create the environment using

$> conda create -y -n ds python=3.5 datashader bokeh=0.12.4 jupyter

then activate the environment using

$> source activate ds

This will allow you to now use datashader with bokeh.

[astrojuanlu]

You are right, there is no conda-forge package and I was running datashader 0.4.0 from the main conda channels. Sorry for the noise and thanks!