TypeError while reading branch containing vector of custom objects containing TArrayF

[renyhp]

uproot version: 5.6.8

Here is the test file: test.root

And here is the simple code that breaks:

>>> uproot.open('test.root')['Test/Event/fTracks/fTracks.fPositions'].arrays()
Traceback (most recent call last):
  File "<stdin>", line 1, in <module>
  File "lib/python3.10/site-packages/uproot/behaviors/TBranch.py", line 874, in arrays
    return self._eager_arrays(
  File "lib/python3.10/site-packages/uproot/behaviors/TBranch.py", line 1135, in _eager_arrays
    return self.parent._eager_arrays(
  File "lib/python3.10/site-packages/uproot/behaviors/TBranch.py", line 1198, in _eager_arrays
    _ranges_or_baskets_to_arrays(
  File "lib/python3.10/site-packages/uproot/behaviors/TBranch.py", line 3513, in _ranges_or_baskets_to_arrays
    uproot.source.futures.delayed_raise(*obj)
  File "lib/python3.10/site-packages/uproot/source/futures.py", line 38, in delayed_raise
    raise exception_value.with_traceback(traceback)
  File "lib/python3.10/site-packages/uproot/behaviors/TBranch.py", line 3449, in basket_to_array
    basket_array = interpretation.basket_array(
  File "lib/python3.10/site-packages/uproot/interpretation/objects.py", line 175, in basket_array
    output = self.basket_array_forth(
  File "lib/python3.10/site-packages/uproot/interpretation/objects.py", line 261, in basket_array_forth
    output = self._discover_forth(
  File "lib/python3.10/site-packages/uproot/interpretation/objects.py", line 421, in _discover_forth
    output[i] = self._model.read(
  File "lib/python3.10/site-packages/uproot/containers.py", line 765, in read
    while cursor.displacement(start_cursor) < num_bytes:
TypeError: '<' not supported between instances of 'int' and 'NoneType'

Thanks!

PS: Here is the ROOT code to make the test file, if that's useful:

class MyTrack {
public:
  MyTrack(int n) {fPositions.Set(n);};
private:
  TArrayF fPositions;
};


class MyEvent {
public:
  void ClearTracks() { fTracks.clear(); };
  void AddTrack(MyTrack track) { fTracks.push_back(track); };
  
private:
  std::vector<MyTrack> fTracks;
};


void test() {
  auto f = new TFile("test.root", "RECREATE");
  auto t = new TTree("Test", "Test");
  auto ev = new MyEvent();
  t->Branch("Event", &ev);
  ev->AddTrack(MyTrack(3));
  t->Fill();
  t->Write();
}

[ariostas]

Thanks for the report!

I looked into this and the reason for the error is that Uproot can't figure out what's the number of bytes of the TArrayF. I tried bypassing that check to see what happens, but it just entirely fails to read the TArrayF. I'm not very familiar with this part of Uproot, so I didn't find an obvious fix for this and I'm not sure how much work it would take to fix it.

The source of this seems to be the TArrayF nested inside an std::vector. If you have a plain TArrayF or you switch the TArrayF to an std::vector then it works.

Unfortunately, Uproot has limitations on what kind of TTrees it can read, so this one might just be one of those. We're switching focus to RNTuple, since it that case we can actually guarantee that any kind of data will be readable with Uproot. If you have control over the code that is writing the data, I would encourage you to try using RNTuple. Otherwise, I think maybe pyroot would be the way to go for now. See @mrzimu's comment below.

[mrzimu]

You could try uproot-custom (an extension of uproot) if the code that writes the data cannot be modified easily:

>>> import uproot
>>> import uproot_custom as uc
>>> uc.AsCustom.target_branches.add("/Test:Event/fTracks/fTracks.fPositions")
>>> f = uproot.open("test.root")
>>> f["Test/Event/fTracks/fTracks.fPositions"].array()
<Array [[[1, 2, 3]]] type='1 * var * var * float32'>

(You need to run pip install uproot-custom to install the extension)

But as @ariostas said, the RNTuple is a better choide to store data.

[ariostas]

Thank you, @mrzimu! That's definitely the best way to ready it if the writing part can't be modified easily.