Build new image

[alimanfoo]

This issue collects together some requests for additional features in the datalab image:

• Add support for docker (i.e., ability to run the docker command)
• Add the gist command (#81)
• Add support for git bash-completion
• Add jupyterlab extensions if possible: jupyterlab-toc, jupyterlab-flake8, jupyterlab-variableInspector (#82)
• Upgrade the conda environment to use pinned environment file from malariagen/binder v2.5.1
• Investigate whether jupyterlab can be used without needing to downgrade (#80) - maybe this also needs an upgrade to the jupyterhub package?

cc @slejdops

[alimanfoo]

cc @leehart

[alimanfoo]

cc @hardingnj

[alimanfoo]

Hi @slejdops, I just tried installing the binder v2.5.1 environment locally and it solves and installs fine, no conflicts. I did:

$ git clone git@github.com:malariagen/binder.git
$ cd binder
$ git checkout v2.5.1
$ ./install-conda.sh

Is there something different about how you are trying to create the environment when you build the image?

[slejdops]

conda cannot resolve environment while building the datalab docker image based on jupyter/base-notebook:python-3.7.6

possibly related to https://github.com/conda/conda/issues/9268

[alimanfoo]

Hi @slejdops, is this the current datalab image dockerfile?

FWIW I think the approach of trying to install into the base environment is likely to cause problems. Can we create a new environment instead? That's what we always do when using malariagen/binder locally. If we create a new (fresh) environment then we know we can get a perfect reproduction of the pinned environment, no conflicts.

[slejdops]

A new image based on malariagen/binder v2.5.1 deployed to the live cluster. cc @hardingnj cc @leehart