jupyterlab-dotscience plugin
A JupyterLab extension which enables data & model versioning and summary statistics tracking.
See dotscience for more details.
Deployment
This plugin is in two parts - the frontend and the backend. The frontend is deployed to npm, the backend to pypi.
To deploy the components automatically you should generally just follow:
pip3 install bump2version # if you don't have this installed already
./scripts/release.sh $VERSION
This will release itself using drone, then later renovate should pick up the change from npm/pypi and suggest a pull request on the jupyterlab-tensorflow repo. You can override this by doing that PR manually.
If you for any reason need to manually deploy to only npm and pypi, you need to have the following variables in your environment:
| Name | Example | Description |
|---|---|---|
PYPI_USER |
fred | the username we use to access pypi. Should be in gitlab-ci. |
PYPI_PASSWORD |
passw0rd | the password for pypi |
NPM_TOKEN |
something-separated-by-dashes | the token for npm. Get it from gitlab-ci or set up your own account and ping Charlotte to add you |
CI_COMMIT_TAG |
0.0.5 | the version to use when releasing the npm package. This should match to a tag in git - if you don't match them then the versions may differ between what's released on pypi and what's released on npm, as the pypi release process pulls the tag from git. Also note this must be a semantic version, otherwise the release will fail :/ |
then run
./shipit-pypi.sh -u $PYPI_USER -p $PYPI_PASSWORD
./shipit-npm.sh $NPM_TOKEN
You will then need to trigger the docker image repo manually if necessary (see .gitlab-ci.yml)
Run Jupyter lab on your host
source ~/miniconda/bin/activate jupyterlab-ext
jupyter lab --watch
installing backend
To install the backend, assuming you've followed the conda based setup xkcd example:
conda install pip
pip install -e jupyterlab_dotscience_backend
jupyter serverextension enable --py jupyterlab_dotscience_backend --sys-prefix
installing frontend
jupyter labextension install jupyterlab_dotscience --no-build
Summary statistics
For python applications look at the dotscience-python library. For other languages we do not yet have a library, but this specification should give you some idea what is expected.
