Conda recipe for Cylc 8.

e.g.

cd recipes
conda skeleton pypi isodatetime
conda build isodatetime
anaconda login
anaconda upload ~/Development/python/anaconda3/conda-bld/linux-x86/isodatetime*.tar.bz2
conda build purge 

Example .condarc:

channels:
  - defaults
  - conda-forge
  - kinow
ssl_verify: true
auto_activate_base: false
anaconda_upload: false

If not loading Anaconda by default, due to conda/conda issue #7980, you will need to source something similar to:

#!/bin/bash

PATH=~/Development/python/anaconda3/bin:$PATH
. ~/Development/python/anaconda3/etc/profile.d/conda.sh

The work directory of a conda environment can be accessed via an environment variable, allowing us thus to build up the Cylc UI directory as follows: ${CONDA_PREFIX}/work/cylc-ui.

You need to enable the kinow channel for this test to work (see .condarc file above).

Then create a new Conda environment, e.g. cylc1.

• conda create -n cylc1
• conda activate cylc1

Now you only need to install the placeholder package "cylc". You can refer to the recipes/cylc/meta.yaml for its contents. But this is basically installing conda packages for metomi-isodatetime, cylc-flow, cylc-uiserver, and unzipping cylc-ui into the conda environment work directory ($PREFIX conda build env var).

• conda install cylc (you can wait for the confirmation question, where you have a chance to see all the packages being installed, including a few cylc* and metomi*)

At this point you can test your environment if you would like, with some commands like:

• cylc --version
• cylc run --no-detach five (or any other suite you use for testing)
• cylc-uiserver --help
• isodatetime

Finally, if you found nothing wrong with your environment, start the hub with:

• jupyterhub --JupyterHub.spawner_class="jupyterhub.spawner.LocalProcessSpawner" --Spawner.args="['-s', '${CONDA_PREFIX}/work/cylc-ui']" --Spawner.cmd="cylc-uiserver" --JupyterHub.logo_file="${CONDA_PREFIX}/work/cylc-ui/img/logo.png"

Voilà 🎉. Go to http://localhost:8000, log in with your local user credentials, and enjoy Cylc 8 alpha!

You can now remove the environment if you would like.

• conda deactivate
• conda env remove -n cylc1

ps: we could use the jupyterhub_config.py shipped with cylc-uiserver, but that file is not included in the wheel file, so conda install cylc-uiserver does not download it. We could still download from GitHub, but that would be over-complicating for this quick experiment IMHO.

• https://github.com/bioconda/bioconda-recipes/tree/master/recipes
• https://medium.com/@giswqs/building-a-conda-package-and-uploading-it-to-anaconda-cloud-6a3abd1c5c52
• https://docs.conda.io/projects/conda-build/en/latest/resources/define-metadata.html