TypeError: Model_ROOT_3a3a_RNTuple.keys() got an unexpected keyword argument 'filter_branch'
Opened this issue · 2 comments
Trying https://github.com/iris-hep/calver-coffea-agc-demo with one of RNTuple file (v1) - to reproduce it, change in cell 5
ttbar_file = "https://xrootd-local.unl.edu:1094//store/user/AGC/rntuple/rntuple-v1-128MiB/nanoAOD/ST_s-channel_4f_InclusiveDecays_13TeV-amcatnlo-pythia8/cmsopendata2015_single_top_s_chan_19394_PU25nsData2015v1_76X_mcRun2_asymptotic_v12-v1_00000_0000.root"
we see next error:
---------------------------------------------------------------------------
TypeError Traceback (most recent call last)
Cell In[5], line 3
1 ttbar_file = "https://xrootd-local.unl.edu:1094//store/user/AGC/rntuple/rntuple-v1-128MiB/nanoAOD/ST_s-channel_4f_InclusiveDecays_13TeV-amcatnlo-pythia8/cmsopendata2015_single_top_s_chan_19394_PU25nsData2015v1_76X_mcRun2_asymptotic_v12-v1_00000_0000.root"
----> 3 events = NanoEventsFactory.from_root({ttbar_file: "Events"}, schemaclass=NanoAODSchema).events()
5 # create the task graph to build a histogram
6 reconstructed_top_mass = calculate_trijet_mass(events)
File /usr/local/lib/python3.10/site-packages/coffea/nanoevents/factory.py:684, in NanoEventsFactory.events(self)
682 """Build events"""
683 if self._is_dask:
--> 684 events = self._mapping(form_mapping=self._schema)
685 report = None
686 if isinstance(events, tuple):
File /usr/local/lib/python3.10/site-packages/uproot/_dask.py:296, in dask(files, filter_name, filter_typename, filter_branch, recursive, full_paths, step_size, steps_per_file, library, ak_add_doc, custom_classes, allow_missing, open_files, form_mapping, allow_read_errors_with_report, known_base_form, decompression_executor, interpretation_executor, **options)
277 return _get_dak_array(
278 files,
279 filter_name,
(...)
293 interpretation_executor,
294 )
295 else:
--> 296 return _get_dak_array_delay_open(
297 files,
298 filter_name,
299 filter_typename,
300 filter_branch,
301 recursive,
302 full_paths,
303 custom_classes,
304 allow_missing,
305 real_options,
306 interp_options,
307 form_mapping,
308 steps_per_file,
309 allow_read_errors_with_report,
310 known_base_form,
311 decompression_executor,
312 interpretation_executor,
313 )
314 else:
315 raise NotImplementedError()
File /usr/local/lib/python3.10/site-packages/uproot/_dask.py:1598, in _get_dak_array_delay_open(files, filter_name, filter_typename, filter_branch, recursive, full_paths, custom_classes, allow_missing, real_options, interp_options, form_mapping, steps_per_file, allow_read_errors_with_report, known_base_form, decompression_executor, interpretation_executor)
1594 else:
1595 obj = uproot._util.regularize_object_path(
1596 ffile_path, fobject_path, custom_classes, allow_missing, real_options
1597 )
-> 1598 common_keys = obj.keys(
1599 recursive=recursive,
1600 filter_name=filter_name,
1601 filter_typename=filter_typename,
1602 filter_branch=filter_branch,
1603 full_paths=full_paths,
1604 ignore_duplicates=True,
1605 )
1606 base_form = _get_ttree_form(
1607 awkward, obj, common_keys, interp_options.get("ak_add_doc")
1608 )
1610 divisions = [0]
TypeError: Model_ROOT_3a3a_RNTuple.keys() got an unexpected keyword argument 'filter_branch'
cc: @alexander-held
[cms-jovyan@jupyter-oksana-2eshadura-40cern-2ech calver-coffea-agc-demo]$ pip list | grep uproot
uproot 5.5.1
@ariostas, it sounds like the TTree (HasBranches) keys method was copied without enough modification. HasBranches.keys has a filter_branches to select branches, but maybe the equivalent is called "fields" for RNTuple? Or it doesn't exist (filter_branches takes a function on TBranch objects, which maybe aren't objects for fields?)?
Yeah, the filter_branches keyword argument doesn't make sense for RNTuples, but it would make sense to have a filter_fields now that there are field objects.
The current implementation of keys is not fully-featured yet. Here's how it looks.
uproot5/src/uproot/models/RNTuple.py
Lines 203 to 211 in 8a71b73
The filter_typename, recursive, and full_paths don't do anything yet. I'll have to check how recursive, and full_paths work for TTrees because I'm not sure how they translate to RNTuples.