add option to pass a list of selectors to dataset.iterate function

[CamphynR]

adding option to pass several selectors to dataset.iterate function

[fschlueter]

Looks good to me. The only thing is that in principle the typing for this argument has to be adapted. Maybe @cozzyd knows that to write there. But I am also happy to ignore that.

[cozzyd]

You should be able to change the type to a Union of the existing type and a Sequence type, I think? Otherwise people won't know how to use this correctly by looking at the type signature.

While I think this is a fine way to do it, I wonder if a more flexible approach would be to provide a helper functor that looks something like:

class SelectorAll:
  def __init__(self, selectors : Sequence[Callable[EventInfo]]):
    self._selectors = selectors
  def __call__(self, info : EventInfo):
    return numpy.all([selector(info) for selector in _selectors])

Then you could also define SelectorAny, SelectorNoneOf, SelectorOnlyOneOf, SelectorAtLeast2, etc.

Then you would be able to use it as: iterate( ... , selector = SelectorAll((A,B,C)))

[fschlueter]

change the type to a Union of the existing type and a Sequence of the existing type

@CamphynR can you try this. I hope this does not become unreadable

[fschlueter]

@cozzyd interesting! I kind of like the idea but I wonder if we should keep it as simple as possible. I guess with your class one could still pass single functions/lambdas but no lists anymore.

I would like for Ruben to go a head with the PR so he can continue working. But we can think of improving this. Implementing selectors as a class could us also allow to implement some nice logging of the filtering.

[fschlueter]

We could also think about filtering base of waveforms ?

[cozzyd]

What I mean is the interface already supports passing a class (as long as the class is a functor and has the correct call method) without any changes, but that still remains true even with the changes and passing a sequence can certainly be a bit more ergonomic.