BUG: accessing data in re-loaded sphere object errors

[chrishavlin]

Bug report

Only took a quick look based on a question on slack, but I'm pretty sure this is a bug:

from yt.testing import fake_amr_ds
import yt
ds = fake_amr_ds()
sp = ds.sphere(ds.domain_center, ds.quan(0.5,'code_length'))
fi = 'tmpfile.h5'
sp.save_as_dataset(fi)
ds_in = yt.load(fi)
ds_in.all_data()

Actual outcome

---------------------------------------------------------------------------
ValueError                                Traceback (most recent call last)
Input In [16], in <cell line: 8>()
      6 sp.save_as_dataset(fi)
      7 ds_in = yt.load(fi)
----> 8 ds_in.all_data()

File ~/src/yt_general/yt/yt/data_objects/static_output.py:1194, in Dataset.all_data(self, find_max, **kwargs)
   1189 def all_data(self, find_max=False, **kwargs):
   1190     """
   1191     all_data is a wrapper to the Region object for creating a region
   1192     which covers the entire simulation domain.
   1193     """
-> 1194     self.index
   1195     if find_max:
   1196         c = self.find_max("density")[1]

File ~/src/yt_general/yt/yt/data_objects/static_output.py:606, in Dataset.index(self)
    603 @property
    604 def index(self):
    605     if self._instantiated_index is None:
--> 606         self._instantiated_index = self._index_class(
    607             self, dataset_type=self.dataset_type
    608         )
    609         # Now we do things that we need an instantiated index for
    610         # ...first off, we create our field_info now.
    611         oldsettings = np.geterr()

File ~/src/yt_general/yt/yt/geometry/particle_geometry_handler.py:72, in ParticleIndex.__init__(self, ds, dataset_type)
     70 self.dataset = weakref.proxy(ds)
     71 self.float_type = np.float64
---> 72 super().__init__(ds, dataset_type)
     73 self._initialize_index()

File ~/src/yt_general/yt/yt/geometry/geometry_handler.py:47, in Index.__init__(self, ds, dataset_type)
     44 # Note that this falls under the "geometry" object since it's
     45 # potentially quite expensive, and should be done with the indexing.
     46 mylog.debug("Detecting fields.")
---> 47 self._detect_output_fields()

File ~/src/yt_general/yt/yt/geometry/particle_geometry_handler.py:360, in ParticleIndex._detect_output_fields(self)
    358 dsl = []
    359 units = {}
--> 360 pcounts = self._get_particle_type_counts()
    361 field_cache = {}
    362 for dom in self.data_files:

File ~/src/yt_general/yt/yt/geometry/particle_geometry_handler.py:86, in ParticleIndex._get_particle_type_counts(self)
     84 def _get_particle_type_counts(self):
     85     result = collections.defaultdict(lambda: 0)
---> 86     for df in self.data_files:
     87         for k in df.total_particles.keys():
     88             result[k] += df.total_particles[k]

File ~/src/yt_general/yt/yt/geometry/particle_geometry_handler.py:106, in ParticleIndex.data_files(self)
    103 if self._data_files is not None:
    104     return self._data_files
--> 106 self._setup_filenames()
    107 return self._data_files

File ~/src/yt_general/yt/yt/geometry/particle_geometry_handler.py:140, in ParticleIndex._setup_filenames(self)
    138 try:
    139     _filename = template % {"num": i}
--> 140     df = cls(self.dataset, self.io, _filename, fi, (start, end))
    141 except FileNotFoundError:
    142     mylog.warning(
    143         "Failed to load '%s' (missing file or directory)", _filename
    144     )

File ~/src/yt_general/yt/yt/frontends/ytdata/data_structures.py:233, in YTDataHDF5File.__init__(self, ds, io, filename, file_id, range)
    230 with h5py.File(filename, mode="r") as f:
    231     self.header = {field: parse_h5_attr(f, field) for field in f.attrs.keys()}
--> 233 super().__init__(ds, io, filename, file_id, range)

File ~/src/yt_general/yt/yt/data_objects/static_output.py:2081, in ParticleFile.__init__(self, ds, io, filename, file_id, range)
   2079 if self.start is None:
   2080     self.start = 0
-> 2081 self.end = max(self.total_particles.values()) + self.start

ValueError: max() arg is an empty sequence

Expected outcome

get all the data!

Version Information
yt '4.4.dev0'

[chrishavlin]

Note, providing the fields arguments seems to avoid the problem:

from yt.testing import fake_amr_ds
import yt
ds = fake_amr_ds()
sp = ds.sphere(ds.domain_center, ds.quan(0.5,'code_length'))
fi = 'tmpfile.h5'
sp.save_as_dataset(fi, fields=ds.field_list)
ds_in = yt.load(fi)
ad = ds_in.all_data()
ad[('grid', 'Density')]