Installing dev versions with Conda?

[jmr]

Hello,

Is there documentation on how to install development versions of bokeh on Windows with Conda?

On a fresh install of Miniconda3-py39_4.12.0-Windows-x86_64.exe, in an anaconda console running as administrator

conda search --override-channels -c "bokeh/label/dev" bokeh

finds all the 3.0.0 dev packages, which is OK

conda install -c "bokeh/label/dev" bokeh

offers to install 2.4.3

conda install -c "bokeh/label/dev" bokeh=3.0.0.doesnotexist

correctly return a PackagesNotFoundError

conda install -c "bokeh/label/dev" bokeh=3.0.0.dev20

fails with the following output:

Collecting package metadata (current_repodata.json): done
Solving environment: failed with initial frozen solve. Retrying with flexible solve.
Solving environment: failed with repodata from current_repodata.json, will retry with next repodata source.
Collecting package metadata (repodata.json): done
Solving environment: failed with initial frozen solve. Retrying with flexible solve.
Solving environment: -
Found conflicts! Looking for incompatible packages.
This can take several minutes.  Press CTRL-C to abort.
failed

UnsatisfiableError: The following specifications were found to be incompatible with each other:

Output in format: Requested package -> Available versions

I’m able to install by forcing

conda install https://anaconda.org/bokeh/bokeh/3.0.0.dev20/download/noarch/bokeh-3.0.0.dev20-py_0.tar.bz2

but in that case, my code fails due to unmet dependencies - but those can be fixed with

conda install -c conda-forge xyzservices

Still - what is the correct way to install?

Thanks

[Bryan]

Not sure what is going on for you locally, but things work the way they have always worked for me in a fresh environment:

testdev ❯ conda install -c bokeh/c/dev bokeh
Collecting package metadata (current_repodata.json): done
Solving environment: done

## Package Plan ##

  environment location: /Users/bryan/anaconda/envs/testdev

  added / updated specs:
    - bokeh


The following packages will be downloaded:

    package                    |            build
    ---------------------------|-----------------
    bokeh-3.0.0.dev20          |             py_0        11.8 MB  bokeh/c/dev
    contourpy-1.0.5            |  py310haf03e11_0         204 KB
    libwebp-1.2.4              |       h56c3ce4_0          75 KB
    libwebp-base-1.2.4         |       hca72f7f_0         316 KB
    pandas-1.4.4               |  py310he9d5cce_0         9.7 MB
    xyzservices-2022.9.0       |     pyhd8ed1ab_0          38 KB  conda-forge
    ------------------------------------------------------------
                                           Total:        22.1 MB

AFAIK bokeh/c/dev and bokeh/label/dev are equivalent.

If you are re-using an old env perhaps that is the issue, with some existing package causing a conflict (e.g. holoviews limits < 3.0 I think)

[jmr]

Thanks. I think having conda-forge enabled is the key here. I just discovered its existence installing xyzservices.

[Bryan]

Oh, that’s unfortunate. I had thought xyzservices was in defaults but I guess in fact it is not. I have floated the suggestion of adding it via some internal channels, not sure how long that might take to actually get added, though.

[Bryan]

I’m told it is being added, so hopefully fairly soon.

[Ian_Thomas]

It is in progress: Add xyzservices 2022.9.0 by skupr-anaconda · Pull Request #2 · AnacondaRecipes/xyzservices-feedstock · GitHub

[Bryan]

This is done:

The following packages will be UPDATED:

  certifi                          2022.6.15-py39hecd8cb5_0 --> 2022.9.24-py39hecd8cb5_0
  typing-extensions  conda-forge/noarch::typing-extensions~ --> pkgs/main/osx-64::typing-extensions-4.3.0-py39hecd8cb5_0
  xyzservices        conda-forge/noarch::xyzservices-2022.~ --> pkgs/main/osx-64::xyzservices-2022.9.0-py39hecd8cb5_0

[jmr]

That was fast! Fixed indeed:

• Uninstall miniconda, reinstall
• conda install -c “bokeh/label/dev” bokeh=3.0.0.dev20

And it works. Thanks!