Factories are task (callable) generators that take an arbitrary input and produce scheduler-compatible tasks from it.
The input can be a filesystem directory tree of executable files (being wrapped
in subprocess.call()), a tree of python modules (with tasks as python
functions), a JSON/YAML text file defining tasks, anything that can be machine
processed, ... provided there is a factory capable of processing it.
This means you can build source code, execute test suites, clean up remote databases, etc. as long as there is a factory abstracting it as standalone tasks (callables) for the scheduler.
from pexen import factory
from pprint import pprint
from tests.factory_test_data import modtree as modtree_root
f = factory.ModTreeFactory()
funcs = list(f(modtree_root))
pprint(funcs)[<function dummy1 at 0x7f9e5f908f80>,
<function dummy2 at 0x7f9e5f90f950>,
<function dummy3 at 0x7f9e5f91a050>,
<function gen_minidummy.<locals>.minidummy at 0x7f9e5f91a0e0>,
...
<function gen_minidummy.<locals>.minidummy at 0x7f9e5f91a5f0>,
<function dummy4 at 0x7f9e5f91a710>]
Here, ModTreeFactory is used to traverse a python package, its modules and
any subpackages, collecting callables tagged with @pexen.sched.meta.func_meta
decorator.
There is a semi-formal specification of the API and the behavior a factory class
should have, for anybody implementing custom factories and/or executing other
factories from within factories (such as to process a .py file as a module
while traversing a filesystem directory tree).
See the specification in a separate file.