Install

Download Anaconda and install (tested with Anaconda3-2022.05-Linux-x86_64.sh)

Configure .condarc

Create the file with a text editor if it doesn't exist or just run conda config to create an empty file.

/home/<user>/.condarc

Content:

auto_activate_base: false

channels:
  - defaults
  - conda-forge
  - fbriol   # FES tide https://github.com/CNES/aviso-fes

# https://docs.conda.io/projects/conda/en/latest/user-guide/configuration/use-condarc.html#ssl-verification-ssl-verify
ssl_verify: false

# https://docs.conda.io/projects/conda/en/latest/user-guide/configuration/use-condarc.html#show-channel-urls-show-channel-urls
show_channel_urls: True

If bebind a proxy (usual in corporations), it's possible to set proxy variables in the .condarc file like this. However, the mamba package manager seems to ignore these definitions and conda itself ignores the no_proxy variable as reported here. Due to this, is recommended to only set the proxy variables in the ~/.bashrc file, e.g.:

export {all_proxy,ALL_PROXY,http_proxy,HTTP_PROXY,https_proxy,HTTPS_PROXY,ftp_proxy,FTP_PROXY}=http://myuser:[email protected]:8080
export {no_proxy,NO_PROXY}=localhost,127.0.0.1,my-company.com

New envs

NOTE: mamba is a reimplementation of the conda package manager in C++, so it is much faster when solving dependencies. It is worth installing it in the base env and just replacing the conda with mamba command when creating a new env (mamba create -n <env_name> ...) or installing a new package (mamba install <pkg>). I could not find a single reason to not use mamba instead of conda when creating a env or installing a package. But in case of a problem with any of the commands below, conda can be used instead. mamba has some dependencies, so it can take conda a few minutes (~5) to install it.

conda install -y -n base mamba

After installing mamba it is best to change the base environment (env) as little as possible and always create a new env when trying new packages. The new env can be created "by hand", e.g.:

mamba create -n <env_name> python=3.10 ipython spyder mypy tqdm pytest python-magic pandas pandas-stubs pyjanitor missingno xarray dask netcdf4 cfgrib zarr rasterio rioxarray cf_xarray eccodes cdsapi pygrib astropy metpy metar seawater gsw pyfes pyinterp universal_pathlib unidecode humanize tabulate termcolor aniso8601 gooey bs4 tenacity cachetools pipreqs pikepdf cx_oracle flask flask-restful celery cartopy cartopy_offlinedata owslib seaborn windrose plotly python-kaleido folium ipyleaflet cmocean colorcet cmasher python-docx xlsxwriter xlrd openpyxl geopy alphashape descartes gitpython git lftp awscli nco cdo ncview

or imported from text files, e.g.:

txt with exact copy (versions) of packages:

• conda list --explicit > spec-file.txt
• conda create --name <env_name> --file spec-file.txt

yml with same packages (but not necessarly same versions):

• mamba env export --no-build > environment.yml
• mamba env create -f environment.yml but remembering to edit the name and prefix variables at the first and last lines to the desired values.

NOTE: even using mamba an env with so many packages can take several (~5) minutes to solve all the dependencies, best to go grab a ☕.

After creation, the new env must be activated, e.g.:

conda activate <env_name>

cx_oracle:

To connect to Oracle databases using the cx_oracle library, the user also needs the Oracle library files.

Dowload and extract:

• instantclient-basic-linux.x64-21.1.0.0.0.zip
• instantclient-sqlplus-linux.x64-21.1.0.0.0.zip

Set the paths:

export LD_LIBRARY_PATH=/some/dir/instantclient_21_1:$LD_LIBRARY_PATH
export PATH=/some/dir/instantclient_21_1:$PATH
export ORACLE_HOME=/some/dir/instantclient_21_1

env to test CF Conventions

mamba create -n cfconv cfchecker compliance-checker

env to run QGIS

mamba create -n qgis python=3.8 qgis jupyter qtconsole matplotlib numpy pandas xarray netCDF4

Others

List the available envs:

conda env list

Remove an env:

conda remove --name <env_name> --all

pycodestyle

Ignore some errors and warnings in the code style:

/home/<user>/.config/pycodestyle

Content:

[pycodestyle]
ignore = E501, E722, W503, W605