Investigate decorator functionality

[Kitchi]

It would be cool to wrap Python functions to create the DAG layers, something like

from dagger import dagcorator

# define some submit_list with variables to iterate over

@dagcorator.add_to_layer(layer_name = 'A', submit_vars = submit_list)
def func1:
    # parse args, etc.
    return

@dagcorator.add_to_layer(layer_name='B', submit_vars = submit_list_2)
def func2():
    # parse args etc.
    return 


if __name__ == '__main__':
    func1()
    func2()

Need to think through the usage patterns a little more, I'm not clear about how exactly to proceed with this at the moment.

[Kitchi]

The caveats of using a decorator-like pattern are :

1. The functions need to be actually called in order for the decorator functions to be activated.
2. I'm not sure how to register a dag name, directory etc. unless it's done globally at the top. I'm not too thrilled about enforcing the use of global declarations.
3. Does dagcorator.add_to_layer need to be a @classmethod? Unless we use a global we cannot instantiate an object at that point.

I need to look more into typical decorator usage patterns to understand the best way to do this.

[Kitchi]

Even if we don't specify explicit dependencies, internally when the decorated functions are called we can define an order of execution that determines the parent/child relationship.

[Kitchi]

Commit 52699b4 has some initial functionality for this, that has not been tested yet.