Loading Tasks#

Graph generation involves creating tasks for each kind. Kinds are ordered to satisfy kind-dependencies, and then the Loader specified in kind.yml is used to obtain the initial set of tasks for that kind.

In short, loaders are responsible for turning the kind.yml file into an initial set of tasks that can then be passed down into the transforms.

What is a Loader?#

Loaders are Python functions with the following signature:

def loader(cls, kind, path, config, parameters, loaded_tasks):

kind is the name of the kind.

path is the path to the configuration directory for the kind. This can be used to load extra data, templates, etc.

config is a dict containing the contents of the kind.yml file.

parameters give details on which to base the task generation. See Parameters for details.

loaded_tasks contains all tasks that have already been loaded (which is guaranteed to contain all kinds of tasks listed in the kind-dependencies key).

The return value is a list of inputs to the transforms listed in the kind’s transforms property. The specific format for the input depends on the first transform - whatever it expects.

Available Loaders#

Taskgraph provides two loaders, although in truth they are virtually identical:

The Transform Loader is a barebones loader.

The Default Loader does everything that the Transform Loader does, and additionally ensures that the Run and Task transforms are used on every task it loads.

Projects may optionally provide their own loaders that support different behavior.