From recipe to feedstock

What is a feedstock?

A feedstock is a code repository containing a Pangeo Forge recipe along with associated metadata files.

Directory structure

The feedstock repo should contain a subdirectory named feedstock, which contains at minimum the following three files:

.
└── feedstock
    ├── meta.yaml
    ├── recipe.py
    └── requirements.txt

The recipe.py file is the recipe itself (see Recipe Composition for how to create this file).

The other two files are described below.

meta.yaml

At minimum, this file requires a recipes section with id and object fields:

# meta.yaml

recipes:
  - id: "unique-recipe-id"
    object: "recipe_module_name:recipe_object_name"

The id field is a unique identifier for your recipe, and can be any descriptive or memorable name of your choosing.

The object field records the name of the Python module which contains your recipe and the name of the recipe (i.e., transforms) object within that module to deploy.

Note

While a recipes section is the minimum requirement for a meta.yaml, this file is also intended to contain additional provenance information about the dataset. Documentation of this aspect is pending implementation of a schema for these fields: pangeo-forge/pangeo-forge-runner#93.

# TODO: Document `dict_object` usage pattern.

requirements.txt

This file should contain the list of dependencies required to run your recipe, including pangeo-forge-recipes, and should follow pip’s requirements file format.

It is advisable for all packages listed here to be pinned to a specific version, which is beneficial for reproducible deployments. For example:

# requirements.txt

pangeo-forge-recipes==<version>