how to add new field in a 'zip' jagged array

[ZhenxuanZhang-Jensen]

Hi experts, I want to add a new field in an already zipped jagged array. For example, if I zip 4D info into a muons object, then I can call pt,eta,phi,charge like this: muons.Muon.pt. However, if I want to add a new field such as 2*pt into this muons object, then I can't do this with muons.Muon['2pt'] = 2 * arrays['Muon_pt"]. Is there anything I misunderstand or how can I add a new field in this jagged array? Could you please help me? thanks

muons = ak.zip({
    "pt": arrays["Muon_pt"],
    "eta": arrays["Muon_eta"],
    "phi": arrays["Muon_phi"],
    "charge": arrays["Muon_charge"],
})

[agoose77]

The syntax you are describing (muons.Muon['2pt'] = 2 * arrays['Muon_pt"]) is not quite right for assigning to an array. It also doesn't match up with the muons array that you've defined above, so I will assume it's a typing mistake?

If you have an array muons defined as

muons = ak.zip({
    "pt": arrays["Muon_pt"],
    "eta": arrays["Muon_eta"],
    "phi": arrays["Muon_phi"],
    "charge": arrays["Muon_charge"],
})

then you can assign a new/existing field with

muons['pt'] = ...

where ... is a Python expression.

This might fail if the expression cannot be broadcast against the existing array.

[jpivarski]

Just to keep everything cross-linked, the question was also asked on https://stackoverflow.com/q/74846508/1623645.

[ZhenxuanZhang-Jensen]

Hi Angus, actually, I did type wrong, but not the muons.Muon, is the 'charge' array should change to 'mass' array. Since we can separate the 'object' and 'kinematic variables' automatically with an underline when we zip 4D info with awkward(not sure why yet). for example:

with uproot.open(file, timeout = 1800) as f:
 tree = f["Events"]
muons = tree.arrays(['Muon_pt','Muon_eta','Muon_phi','Muon_mass'], library = "ak", how = "zip")

then the muons can call the pt,eta,phi,mass in this way automatically: muons.Muon.pt, muons.Muon.eta... For this reason, I don't know how to add a new field in this muons.Muon. But I found that if I assign it to a new value and reassign it to muons.Muon which can work like:

Muon = muons.Muon
Muon['pt2'] = ...
muons.Muon = Muon
muons.Muon.pt2

In this way, I can add this pt2 field in the muons.Muon.

[agoose77]

@ZhenxuanZhang-Jensen Oh, I learned something about uproot's zip argument from this — it removes the common prefix of like-named ragged branches.

Back to your question.

You cannot set the value of a field with array.field = value. In fact, if you already have a field with the name field, it will raise an exception:

>>> import uproot
>>> import skhep_testdata
>>> import awkward as ak
>>> arrays = uproot.open(skhep_testdata.data_path("uproot-HZZ.root"))["events"].arrays(how="zip")
>>> arrays.Photon = 2
---------------------------------------------------------------------------
AttributeError                            Traceback (most recent call last)
Cell In[29], line 1
----> 1 arrays.Photon = 2

File ~/Git/awkward-1.0/src/awkward/highlevel.py:1160, in Array.__setattr__(self, name, value)
   1158     super().__setattr__(name, value)
   1159 elif name in self._layout.fields:
-> 1160     raise ak._errors.wrap_error(
   1161         AttributeError(
   1162             "fields cannot be set as attributes. use #__setitem__ or #ak.with_field"
   1163         )
   1164     )
   1165 else:
   1166     raise ak._errors.wrap_error(
   1167         AttributeError(
   1168             "only private attributes (started with an underscore) can be set on arrays"
   1169         )
   1170     )

AttributeError: fields cannot be set as attributes. use #__setitem__ or #ak.with_field

The proper syntax for setting a field is to use the subscript operator, e.g. array['x'] = .... However, for nested fields, e.g. arrays.Photon.Px, you need to use multiple strings

arrays['Photon', 'Px'] = ...

or tuple of strings--

arrays[('Photon', 'Px')] = ...

You can't use arrays.Photon['Px'] = ... — see the user guide for details as to why this doesn't work.

So, if uproot has given you an array with a type like

2421 * {
    Muon: var * {
        pt: float32,
        eta: float32,
        phi: float32,
        mass: float32
    }
}

So, you just need to write

muons['Muon', 'pt2'] = 2 * muons.Muon.pt

[ZhenxuanZhang-Jensen]

Thank you so much! The answer looks very nice to me.