Still wrong return values when masking (see #50) about awkward-0.x HOT 4 CLOSED

I think empty = array[False] is a typo (and I don't know what it's supposed to be).

Other than that, everything else looks correct.

from awkward-0.x.

Oh, now I see the array defined above it (I thought this was supposed to be a function call— sorry!).

Still looking at it, I don't see any errors. This is how content and flatten() are supposed to act.

Slicing a jagged array does minimal work: it replaces the starts and stops arrays, but doesn't touch the content. This avoids having to recursively make changes to all the contents, which might be a wide Table. In a very common case, you'll be applying cuts on an array of events. You don't want it to have to modify every field of every particle— you want it to quickly mark some events as bad.

This is why there's an important difference between content and flatten(). If you want to logically flatten the data, so that only the desired values exist, use flatten().

from awkward-0.x.

Another thing to notice is that your first example used a flat boolean array as a mask, which masks out the top level of structure. The second example used a jagged boolean mask, which has a different meaning: it filters at the depth of its own jaggedness.

Removing inner values ("particles"):

array[array != 1]

Removing outer values ("events"):

array[array.min() != 1]

(Look at the mask arrays directly to see the difference.)

from awkward-0.x.

I think this should be made clear in the uproot README. Currently it only shows jagged_array.content to access the underlying ndarray when this may not always be semantically the right function to use.

EDIT:
Also here on this README. It's not clear that this is not doing what you expect.

from awkward-0.x.