Writing pandas DataFrame with Uproot5, var * var * float64

[brightneedle]

I have a dataframe where one or more columns are lists of 1D arrays, eg:

   run_number  event_number  pass_vbf_sel  ...                                        bjetTrigSFs      truth_mhh ntag
0      310691          1593         False  ...  ((0.9997807, 0.9997807, 0.9997807, 0.9997807, ...  719639.263731    3
1      303208          1353          True  ...  ((0.9832633, 0.98921436, 0.97731936, 0.9911188...  401586.494818    2
2      307259           419          True  ...  ((0.9763006, 0.98883635, 0.9638313, 0.98624736...  714343.701542    4
3      300863          1293         False  ...  ((0.99956995, 0.99956995, 0.99956995, 0.999569...  492971.115823    3
4      307259           873         False  ...  ((0.9992055, 0.9992055, 0.9992055, 0.9992055, ...  301056.351272    2

When I try and write the dataframe to .root using uproot.recreate, I get the following error:

TypeError: cannot write Awkward Array type to ROOT file:

    var * var * float64

Is there a way to get round this? I have to be able to keep the same structure as the dataframe, i.e. making N separate 1-D arrays won't work for my application

[jpivarski]

This is a data type that Uproot can't write. (The disk-format for var * var * float64 is completely different from the format for var * float64, and var * float64 can be vectorized, whereas var * var * float64 can't.)