BiasedNFWPhaseSpace crashes when populating halo catalog

[johannesulf]

The following simple code crashes with BiasedNFWPhaseSpace but works fine with NFWPhaseSpace. It's probably related to the fact that BiasedNFWPhaseSpace creates more lookup tables.

from halotools.sim_manager import FakeSim
from halotools.empirical_models import HodModelFactory
from halotools.empirical_models import TrivialPhaseSpace, Zheng07Cens
from halotools.empirical_models import BiasedNFWPhaseSpace, Zheng07Sats

halocat = FakeSim()

cens_occ_model =  Zheng07Cens()
cens_prof_model = TrivialPhaseSpace()

sats_occ_model =  Zheng07Sats()
sats_prof_model = BiasedNFWPhaseSpace()

model = HodModelFactory(
    centrals_occupation=cens_occ_model,
    centrals_profile=cens_prof_model,
    satellites_occupation=sats_occ_model,
    satellites_profile=sats_prof_model)

model.populate_mock(halocat = halocat)

Traceback (most recent call last):
  File "test.py", line 20, in <module>
    model.populate_mock(halocat = halocat)
  File "/home/research/Projects/halotools/env/lib/python3.7/site-packages/halotools/empirical_models/factories/hod_model_factory.py", line 1209, in populate_mock
    ModelFactory.populate_mock(self, halocat, **kwargs)
  File "/home/research/Projects/halotools/env/lib/python3.7/site-packages/halotools/empirical_models/factories/model_factory_template.py", line 236, in populate_mock
    self.mock.populate(**mockpop_kwargs)
  File "/home/research/Projects/halotools/env/lib/python3.7/site-packages/halotools/empirical_models/factories/hod_mock_factory.py", line 346, in populate
    func(table=self.galaxy_table[gal_type_slice], seed=seed, **d)
  File "/home/research/Projects/halotools/env/lib/python3.7/site-packages/halotools/empirical_models/factories/model_factory_template.py", line 292, in decorated_func
    return func(*args, **kwargs)
  File "/home/research/Projects/halotools/env/lib/python3.7/site-packages/halotools/empirical_models/phase_space_models/analytic_models/satellites/nfw/nfw_phase_space.py", line 110, in assign_phase_space
    MonteCarloGalProf.mc_pos(self, table=table, seed=seed)
  File "/home/research/Projects/halotools/env/lib/python3.7/site-packages/halotools/empirical_models/phase_space_models/analytic_models/monte_carlo_helpers.py", line 439, in mc_pos
    x, y, z = self.mc_halo_centric_pos(*profile_params, **kwargs)
  File "/home/research/Projects/halotools/env/lib/python3.7/site-packages/halotools/empirical_models/phase_space_models/analytic_models/satellites/nfw/nfw_phase_space.py", line 718, in mc_halo_centric_pos
    return MonteCarloGalProf.mc_halo_centric_pos(self, *concentration_array, **kwargs)
  File "/home/research/Projects/halotools/env/lib/python3.7/site-packages/halotools/empirical_models/phase_space_models/analytic_models/monte_carlo_helpers.py", line 354, in mc_halo_centric_pos
    x, y, z = self.mc_solid_sphere(*profile_params, **kwargs)
  File "/home/research/Projects/halotools/env/lib/python3.7/site-packages/halotools/empirical_models/phase_space_models/analytic_models/satellites/nfw/nfw_phase_space.py", line 684, in mc_solid_sphere
    return MonteCarloGalProf.mc_solid_sphere(self, *concentration_array, **kwargs)
  File "/home/research/Projects/halotools/env/lib/python3.7/site-packages/halotools/empirical_models/phase_space_models/analytic_models/monte_carlo_helpers.py", line 299, in mc_solid_sphere
    *profile_params, seed=seed)
  File "/home/research/Projects/halotools/env/lib/python3.7/site-packages/halotools/empirical_models/phase_space_models/analytic_models/monte_carlo_helpers.py", line 213, in _mc_dimensionless_radial_distance
    rad_prof_func_table_indices.flatten())
  File "/home/research/Projects/halotools/env/lib/python3.7/site-packages/halotools/empirical_models/model_helpers.py", line 267, in call_func_table
    out[ix] = f(abscissa[ix])
IndexError: index 1061093 is out of bounds for axis 0 with size 81875

[aphearin]

Hmmm, when I run the same code using an actual catalog, rather than FakeSim, no exception is raised. So this looks like it might be a subtle bug. Have you noticed this come up in other real halo catalogs?

This is closely related to #920. I have some prototype code I'm turning into a PR in which the entire digitization of profiles is eliminated. It's just taking some time because there is kind of a lot of machinery to scrap.

[johannesulf]

Yes, I had the same bug in an actual halo catalog.

[johannesulf]

This bug was most likely introduced by cddfd5f. The crash only occurs for BiasedNFWPhaseSpace where the lookup table has two dimensions (conc, conc_gal_bias) instead of one (conc) like NFWPhaseSpace.

The fix is probably as simple as...
self.rad_prof_func_table_indices[np.array(digitized_param_list, dtype='intp')] -> self.rad_prof_func_table_indices[tuple(digitized_param_list)]