Accessing custom cpp classes with uproot

[moleluca]

Hi.
I wrote a cpp code with classes and now I want to analyze the output tree with uproot.
What is the proper way to implement this?
I use a script which is including uproot3, I don't know if this might be different than uproot5. Of course I can upgrade the code if needed.
The script (attached) gives this error:

warnings.warn(
Traceback (most recent call last):
  File "/cvmfs/sft.cern.ch/lcg/views/LCG_105/x86_64-el9-gcc13-opt/lib/python3.9/site-packages/uproot3/tree.py", line 414, in get
    return self._branchlookup[name]
KeyError: b'tracks.clusters.det'

During handling of the above exception, another exception occurred:

Traceback (most recent call last):
  File "/eos/project-w/wis-tb24/public/tb24-tracking/tools/tb24_tracking_analysis.py", line 203, in <module>
    main(args.file_path)
  File "/eos/project-w/wis-tb24/public/tb24-tracking/tools/tb24_tracking_analysis.py", line 195, in main
    event_id, trigger_timestamp, a_x, b_x, a_y, b_y, quality_flag, clusters_det, clusters_time, clusters_pos_x, clusters_pos_y, clusters_pos_z, clusters_size, clusters_charge = get_tracks_data(tree)
  File "/eos/project-w/wis-tb24/public/tb24-tracking/tools/tb24_tracking_analysis.py", line 21, in get_tracks_data
    clusters_det = tree.array("tracks.clusters.det")
  File "/cvmfs/sft.cern.ch/lcg/views/LCG_105/x86_64-el9-gcc13-opt/lib/python3.9/site-packages/uproot3/tree.py", line 525, in array
    branches = list(self._normalize_branches(branch, awkward0))
  File "/cvmfs/sft.cern.ch/lcg/views/LCG_105/x86_64-el9-gcc13-opt/lib/python3.9/site-packages/uproot3/tree.py", line 857, in _normalize_branches
    for x in self._normalize_branches([arg], awkward0):
  File "/cvmfs/sft.cern.ch/lcg/views/LCG_105/x86_64-el9-gcc13-opt/lib/python3.9/site-packages/uproot3/tree.py", line 892, in _normalize_branches
    branch = self.get(word, aliases=aliases)
  File "/cvmfs/sft.cern.ch/lcg/views/LCG_105/x86_64-el9-gcc13-opt/lib/python3.9/site-packages/uproot3/tree.py", line 416, in get
    return self._get(name, recursive, filtername, filtertitle, aliases)
  File "/cvmfs/sft.cern.ch/lcg/views/LCG_105/x86_64-el9-gcc13-opt/lib/python3.9/site-packages/uproot3/tree.py", line 409, in _get
    raise uproot3.rootio._KeyError("not found: {0}\n in file: {1}".format(repr(name), self._context.sourcepath))
_KeyError: not found: b'tracks.clusters.det'
 in file: ../../data/srs/tracks/run_36_20240922165747_events_cut10_tracks.root

I attach both the script and the input file.
Archive.zip

Any help would be much appreciated.
Thanks

[jpivarski]

Uproot can't use C++ class definitions because it doesn't compile C++ code. ROOT does this either by running Cling to compile a set of C++ class definitions or by loading a library of precompiled code (.so on Linux, .dylib on MacOS, .dll on Windows). But Uproot is only Python.

If you've defined a method det on tracks.clusters that finds the clusters associated with a given track, then this matching would have to be reimplemented in Python (possibly vectorized by NumPy or Awkward Array, so that it's fast). However, that would be much easier to do if the data stored in the file were arrays or std::vectors of integer indexes, rather than full C++ objects.

>>> tree["events/tracks/tracks.clusters"].typename
'std::vector<Tracking::Cluster>[]'
>>> tree["events/tracks/tracks.clusters"].interpretation
AsObjects(AsArray(True, False, AsVector(False, Model_Tracking_3a3a_Cluster), ()))

Instead of std::vector<Tracking::Cluster>[] (which has been serialized in a format Uproot doesn't recognize, #38), if you had std::vector<std::vector<int>> where the integers are positions of the clusters in their arrays, then it would be easier to write that logic as an Awkward Array slice on the Python side.

[moleluca]

Hi. Thank you for the prompt and detailed response. Best regards

…

On Tue, 24 Sept 2024 at 17:26, Jim Pivarski ***@***.***> wrote: Uproot can't use C++ class definitions because it doesn't compile C++ code. ROOT does this either by running Cling to compile a set of C++ class definitions or by loading a library of precompiled code (.so on Linux, .dylib on MacOS, .dll on Windows). But Uproot is only Python. If you've defined a method det on tracks.clusters that finds the clusters associated with a given track, then this matching would have to be reimplemented in Python (possibly vectorized by NumPy or Awkward Array, so that it's fast). However, that would be much easier to do if the data stored in the file were arrays or std::vectors of integer indexes, rather than full C++ objects. >>> tree["events/tracks/tracks.clusters"].typename'std::vector<Tracking::Cluster>[]'>>> tree["events/tracks/tracks.clusters"].interpretationAsObjects(AsArray(True, False, AsVector(False, Model_Tracking_3a3a_Cluster), ())) Instead of std::vector<Tracking::Cluster>[] (which has been serialized in a format Uproot doesn't recognize, #38 <#38>), if you had std::vector<std::vector<int>> where the integers are positions of the clusters in their arrays, then it would be easier to write that logic as an Awkward Array slice <https://awkward-array.org/doc/main/reference/generated/ak.Array.html#ak.Array.__getitem__> on the Python side. — Reply to this email directly, view it on GitHub <#1301 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/ALEDZGCKJ3XXFYOQTK3WD6TZYGAEDAVCNFSM6AAAAABOYOB47WVHI2DSMVQWIX3LMV43URDJONRXK43TNFXW4Q3PNVWWK3TUHMYTANZUGA3DQNI> . You are receiving this because you authored the thread.Message ID: ***@***.***>