DRY when adding sub-dispatch nodes about schedula HOT 11 CLOSED

I prefer to play with the object type then add two extra kw...

example:

inputs=('a', 'b') 

is equivalent to

inputs={'a':'a','b':'b'}

from schedula.

And can you combine both?

from schedula.

what do you mean?

from schedula.

Specify both auto-mapped names and explicitly mapped ones, in the same add_dispatcher() invocation?

from schedula.

I want to avoid extra kws. I can build an utility like:

def auto_dict(*args, **kw):
    return sh.combine_dict(kw, base={k: k for k in args}) 

what do you think?

from schedula.

In practical terms, what you suggest achieves the same, but with more code/dependencies.

What's the reason not to add new kwds?

from schedula.

I have to add for both inputs and outputs kw, while the utl is solving both (so less code).

from schedula.

Well, that's a valid reason.

Let's compare the two:

from utils import auto_dict

add_dispatcher(
               dsp, 
               auto_dict('a', 'b', c='c_renamed'), 
               auto_dict('out_a', 'out_b', out_c='out_c_renamed'))

vs

add_dispatcher(
               dsp, 
               {'b': 'out_b_renamed'}, 
               {'out_b': 'out_b_renamed'}, 
               inp_names=['a', 'b'], 
               out_names=['a_out', 'b_out']))

I like the use of python's kwd syntax.
Note though that most of the time, at least in co2mpas code, it would be only the last 2 kwds that are used - no that many renamed params.

But with an intermediate function you loose the opportunity to invoke sub-dispatchers directly from configuration text-files, at least not without requiring the use of that helper function, in which case, why not add the kwds anyway?

In any case, auto_dict is totally opaque name to tell what it does, and hard to remember with all other auto-magic dict functions.
I would call it identical_kv_dict or even shorter idkvmap.

from schedula.

Of course the optimal would be, since most of the times only the identical kvs are used, add a new method that has the inp_names/out_names kwds as the first args of the new methods, and move inputs/outputs args at the end.
So in a nutshell, reverse the meaning of those two.

That's refactoring for DRY.

from schedula.

Something like this?

add_dispatcher(
               dsp, 
               ['a', 'b', {'c'='c_renamed'}], 
               ('out_a', 'out_b', dict(out_c='out_c_renamed'))

And these should be invalid:

add_dispatcher(
               dsp, 
               'abc',  # valid: a sequence 
               ('out_a', 'out_b', dict(out_c='out_c_renamed'), 'd')  # INVALID: trailing element

from schedula.

Implementation example:

d.add_dispatcher(dsp, inputs=('a', {'b': 'c'}, 'd', {'e': 'f', 'g': 'h'})

equivalent:

d.add_dispatcher(dsp, inputs={'a': 'a', 'b': 'c', 'd': 'd', 'e': 'f', 'g': 'h'})

from schedula.