Add the mirror_defaults decorator

[idanarye]

This can solve the problem of using AUTO vs replicating arguments. We could have:

def retry(times, func, args=[], kwargs={}, acceptable=Exception, sleep=1,
          max_sleep=False, log_level=logging.DEBUG, pred=None, unacceptable=()):
    ...


@mirror_defaults(retry)
retrying(times, acceptable=AUTO, sleep=AUTO, max_sleep=AUTO, log_level=AUTO, pred=AUTO, unacceptable=AUTO):
    ...

[idanarye]

c.c. @koreno - I can't add you as a reviewer.

[koreno]

1. AUTO might be a desireable override. better make it mirror_defaults(foo, token=MIRROR) (allow using an alternate token, default to MIRROR)
2. can this be applied in many other places in easypy? wait/iter_wait? resilient/resilience? if it works then we can consider incorporating this. Otherwise I'm feeling this might be too fringe and magic... (I haven't thought it thoroughly)

[koreno]

c.c. @koreno - I can't add you as a reviewer.

@doroncohen will I be re-admitted to this repo?...

[idanarye]

• AUTO might be a desireable override. better make it mirror_defaults(foo, token=MIRROR) (allow using an alternate token, default to MIRROR)

How do you feel about mirror_defaults being the default "token" (since it's already imported)?

• can this be applied in many other places in easypy? wait/iter_wait? resilient/resilience? if it works then we can consider incorporating this. Otherwise I'm feeling this might be too fringe and magic... (I haven't thought it thoroughly)

I see no reason why it shouldn't work everywhere - though we may need to have special treatment for generators to avoid messing up reflection...

[doron-cohen]

I think that we should allow the users to use these features but using them in the repo might put off other developers from contributing.

[koreno]

I think that we should allow the users to use these features but using them in the repo might put off other developers from contributing.

I think that if we don't use an easypy feature within easypy, then it shouldn't be in easypy...
Either we sanction it and demonstrate its greatness in our code, or we deem it improper coding for some reason and therefor reject it.

• AUTO might be a desireable override. better make it mirror_defaults(foo, token=MIRROR) (allow using an alternate token, default to MIRROR)

How do you feel about mirror_defaults being the default "token" (since it's already imported)?

mm.... that feels odd and unreadable. mirror_defaults.MIRROR?

• can this be applied in many other places in easypy? wait/iter_wait? resilient/resilience? if it works then we can consider incorporating this. Otherwise I'm feeling this might be too fringe and magic... (I haven't thought it thoroughly)

I see no reason why it shouldn't work everywhere - though we may need to have special treatment for generators to avoid messing up reflection...

then I think we should implement this treatment...

[idanarye]

mm.... that feels odd and unreadable. mirror_defaults.MIRROR?

Too long. Let's just use MIRROR - one could always import it together with mirror_defaults

[idanarye]

Can we decide about the token before I start applying this wherever possible and using that token everywhere?

[YuvalEvron]

Can we decide about the token before I start applying this wherever possible and using that token everywhere?

We can also pass the names of parameters we want to mirror as *args to the decorator.

[idanarye]

Another option is to create a decorator that verifies the defaults are the same:

def foo(a=1, b=2, c=3):
    ...


@ensure_same_defaults(foo)
def bar(a=1, b=2, c=4):
    ...

This will raise an exception because bar's default for c is 4 instead of 3. It's less automatic than mirror_defaults but it has the advantage of IPython and IDEs showing the correct signature.

[koreno]

So do we still want this? or did we lose interest?

[YuvalEvron]

So do we still want this? or did we lose interest?

Looks like we can't agree on the implementation.

[idanarye]

So do we still want this? or did we lose interest?

Looks like we can't agree on the implementation.

Not on the mirror_defaults implementation, but I think we can agree on the ensure_same_defaults implementation?

[YuvalEvron]

So do we still want this? or did we lose interest?

Looks like we can't agree on the implementation.

Not on the mirror_defaults implementation, but I think we can agree on the ensure_same_defaults implementation?

👍