Not able to get eta from MomentumArray4D

[alpetukhov]

Dear vector developers!

I'm trying to read the .root file containing the TLorentzVectors with uproot and want to compute the pseudorapidity of those vectors. For that I'm creating an awkward array from the components of the TLorentzVector, with the approach similar to this StackOverflow answer.

import uproot
import vector

f = uproot.open('file.root:tree')
arrays = f.arrays(["truth_p4"])
truth_lorenz = vector.zip(
    {
        'px': arrays['truth_p4']['fP']['fX'],
        'py': arrays['truth_p4']['fP']['fY'],
        'pz': arrays['truth_p4']['fP']['fZ'],
        'E': arrays['truth_p4']['fE'],
    }
)

But if I try to access the eta field of this array

truth_lorenz.eta

I get the following error

 File "reproducer.py", line 15, in <module>
    truth_lorenz.eta
  File "/home/alex/trt_pid_venv/lib/python3.8/site-packages/vector/_methods.py", line 3300, in eta
    return eta.dispatch(self)
  File "/home/alex/trt_pid_venv/lib/python3.8/site-packages/vector/_compute/spatial/eta.py", line 90, in dispatch
    function(v.lib, *v.azimuthal.elements, *v.longitudinal.elements),
  File "/home/alex/trt_pid_venv/lib/python3.8/site-packages/vector/_compute/spatial/eta.py", line 33, in xy_z
    return lib.nan_to_num(
  File "<__array_function__ internals>", line 5, in nan_to_num
  File "/home/alex/trt_pid_venv/lib/python3.8/site-packages/awkward/highlevel.py", line 1436, in __array_function__
    return ak._connect._numpy.array_function(func, types, args, kwargs)
  File "/home/alex/trt_pid_venv/lib/python3.8/site-packages/awkward/_connect/_numpy.py", line 28, in array_function
    return function(*args, **kwargs)
  File "/home/alex/trt_pid_venv/lib/python3.8/site-packages/awkward/operations/structure.py", line 4244, in nan_to_num
    out = ak._util.recursively_apply(
  File "/home/alex/trt_pid_venv/lib/python3.8/site-packages/awkward/_util.py", line 1208, in recursively_apply
    return transform(layout, 1, user)
  File "/home/alex/trt_pid_venv/lib/python3.8/site-packages/awkward/_util.py", line 1201, in transform
    return custom()
  File "/home/alex/trt_pid_venv/lib/python3.8/site-packages/awkward/operations/structure.py", line 4234, in <lambda>
    nplike.nan_to_num(
  File "/home/alex/trt_pid_venv/lib/python3.8/site-packages/awkward/nplike.py", line 294, in nan_to_num
    return self._module.nan_to_num(*args, **kwargs)
  File "<__array_function__ internals>", line 5, in nan_to_num
  File "/home/alex/trt_pid_venv/lib/python3.8/site-packages/numpy/lib/type_check.py", line 481, in nan_to_num
    _nx.copyto(d, nan, where=idx_nan)
  File "<__array_function__ internals>", line 5, in copyto
TypeError: no implementation found for 'numpy.copyto' on types that implement __array_function__: [<class 'numpy.ndarray'>, <class 'awkward.highlevel.Array'>]

I'm able to access fields p and pt and I also get no errors if I try to access every value individually

for tl in truth_lorenz:
    print(tl.eta)

I was able to reproduce it with the following code, so I think it is not to deal with uproot itself

import vector

p4 = vector.awk(
    [
        [{'px':1, 'py':1, 'pz':1, 'E':1}]
    ]
)
p4.eta

What could've caused this issue? Is there a way around it in the vector package? Or is there maybe a better way to read the TLorentzVectors if I only need p, pT and eta to avoid it completelty?

My versions are

• python: 3.8.10
• vector: 1.1.1
• awkward: 1.4.0
• uproot: 4.0.11

Best reagrds,
Aleksandr

[henryiii]

Could you try awkward 1.4.2? This was a bug with an attempt to support zero copy arrays that failed, and was promptly removed in 1.4.2.

[alpetukhov]

Yes, switching to 1.5.0 (closest available with pip) solved this, thank you!
P.S. Oh my, I'm THIS much versions behind the latest one, better look into if it's worth updating :)

[henryiii]

Ahh, I actually meant 2.4.2. I thought you were on 2.4.0. Good that that solved it though. :)