Skip to content

Latest commit

 

History

History
135 lines (105 loc) · 4.51 KB

README.md

File metadata and controls

135 lines (105 loc) · 4.51 KB
Raw

conda

This explains some things about how to utilize conda to manage python environments. To learn much more details about conda, we suggest visiting the conda documentation website.

A YAML specification file is to configure a python environment. We have prepared a few specification files, please see the conda directory. The YAML specification file is also platform dependent. In most cases, the linux-64 will work. For older systems, we have provided a YAML file for an older 64-bit glibc.

These instructions assume you have conda/miniconda installed. If conda is not installed, please consult this webpage.

If conda cannot be used, please consult the required software for packages to install.

Several basic YAML files are provided to install an appropriate python environment. To expidite the conda environment solver, a YAML_export or YAML_explicit file is also provided for quicker recovery of python environment.

Initialization:

$ cat conda/gridTools.yml
$ conda env create -f=conda/gridTools_export-linux-64.yml
$ conda env export > conda/gridTools_export-linux-64.yml
$ conda list --explicit > conda/gridTools_explicit-linux-64.txt

NOTE: If you are recreating the package lists (txt) or the export (yml) file. Do that before running the pip install for gridtools.

NOTE: The exported yml file will specify a directory path for installation. It is recommended to delete the last line in the yml file before publication.

NOTE: The explicit package listings will contain a specific OS platform target (e.g. linux-64) and may not be compatible with the intended target platform for installation of the library.

For a quicker recovery of a conda environment, use the exported YAML file:

$ conda env remove --name gridTools
$ conda env create -f=conda/gridTools_export-linux-64.yml

NOTE: In some cases, the above may fail. You then have two options. (1) Use the generic gridTools.yml file which may take several hours to resolve. (2) Create a gridTools enviornment with python and then install conda packages sequentially using the slow_bootstrap.sh script.

Alternate Install(1):

$ conda env create -f conda/gridTools.yml
$ conda activate gridTools

Alternate Install(2):

$ conda create -n gridTools python==3.7.10
$ conda activate gridTools
$ ./slow_bootstrap.sh

You can also restore a conda environment using the explicit listing of packages.

$ conda create --name gridTools --file conda/gridTools_explicit-linux-64.txt

Performance

Sometimes the conda resolution can take a very long time to run. Setting an automatic timeout is recommended if building a custom conda enviroment. This can be done using the timeout feature of UNIX time command.

In this example, the timeout is set to 5 minutes to allow resolution of the environment.

$ time timeout 5m conda env create -f conda/gridTools.yml

Environments

Current operational environment for the grid toolset: gridTools

gridTools

This is the main enviroment for utilizing the grid generation libraries.

NOTE: Avoid version 0.5.2 of xesmf. If you need to use the xgcm library, install it as a separate environment.

legacyTools

NOTE: This is a very limited environment with netcdf4 and matplotlib's basemap to review former functionality of older libraries and software.

This environment supported investigation into the following libraries:

Post install

A successful install of conda packages will make the pip install . quite easy. All the dependent packages are again referenced in the requirements.txt file for pip. At the conclusion of the install, only a minimal amount of package should be actually installed.

You should see a similar message:

Successfully built numpypi xesmf
Installing collected packages: xesmf, numpypi, gridtools
  Running setup.py develop for gridtools
  Successfully installed gridtools-0.2.0+aaf0877 numpypi-1.0.0+40b7b0d xesmf-0.5.4.dev8+g7a2830c

If you are just doing an install and do not plan to edit any code, please use: pip install .

If you are planning to edit code within gridtools, please use: pip install -e .. This will allow you to edit the gridtools library. Changes will immediately be seen when you restart the jupyter kernel and rerun all the cells or rerun the python script.