Default to branch name when no dictionary present

[RogerJanusiak]

When building a query like this:

query = ds.Select(lambda e: {
    "track_eta": e.id_track_eta,
    "track_phi": e.id_track_phi,
})
data = query.value()

I think that it could be nice to make it so that it could be written like this:

query = ds.Select(lambda e: {
     e.id_track_eta,
    e.id_track_phi,
})
data = query.value()

and it would use id_track_eta/phi by default as the name of the dictionary objects in the awkward array.

[RogerJanusiak]

If we don't change it to match what is above, at least improving the errors would be nice!

[masonproffitt]

It's an interesting idea, but I can see a couple problems. The syntax in the second example is for a Python set, which doesn't have keys, so I think returning a dict-type object would be surprising to most users. Also, what happens if one or more of the elements is an expression more complicated than just an attribute (e.g., e.id_track_pt / 1000)?

The use case pattern in your first example is common enough that I agree there should be a more convenient syntax, but I'm not sure exactly what it should look like. Perhaps a new FuncADL operation type would be appropriate--something like SelectAttributes that would take a list of attribute names as an argument.

If we don't change it to match what is above, at least improving the errors would be nice!

Can you give an example of the errors you're referring to, and what would you suggest as the new error text?