Want to write TTree with an Awkward record array, rather than a dict of non-record arrays

[laf070810]

I'm a newbie to uproot and encountered some problems when trying to use uproot. Sorry for possible misunderstandings of the usage, though I think my expectation might be reasonable.

Create a jagged array

import uproot
import awkward as ak
array = ak.Array({'a': [[1.1, 2.2, 3.3], [], [4.4, 5.5]], 'b': [[1.1, 2.2, 3.3], [], [4.4, 5.5]], 'c': [1, 2, 3]})

and save it to a ROOT file

with uproot.recreate('data.root') as output_file:
    output_file['tree'] = array

I expect that the saved tree has 3 branches

name | typename | interpretation
---------------------+--------------------------+------------------------------- a | double[] | AsJagged(AsDtype('>f8')) b | double[] | AsJagged(AsDtype('>f8')) c | int64_t | AsDtype('>i8')

but actually I got errors

TypeError: fields of a record must be NumPy types, though the record itself may be in a jagged array

field 'a' has type var * float64

This situation easily happens when reading a TTree with variable length arrays and then saving it back after some manipulations.

[masonproffitt]

I'm running into this error as well (in ServiceX). I'd expect this to do the same thing as output_file['tree'] = {'a': ak.Array([[1.1, 2.2, 3.3], [], [4.4, 5.5]]), 'b': ak.Array([[1.1, 2.2, 3.3], [], [4.4, 5.5]]), 'c': ak.Array([1, 2, 3])} (which does work).

[jpivarski]

Noted: this is an important user-expectation. I think I've answered this question on Gitter, too.