r""" Module containing the `~halotools.mock_observables.npairs_jackknife_3d` function
used to estimate errors in the `~halotools.mock_observables.tpcf` function.
"""
from __future__ import absolute_import, division, print_function, unicode_literals
import numpy as np
import multiprocessing
from functools import partial
from warnings import warn
from .rectangular_mesh import RectangularDoubleMesh
from .mesh_helpers import _set_approximate_cell_sizes, _cell1_parallelization_indices
from .cpairs import npairs_jackknife_3d_engine
from .npairs_3d import _npairs_3d_process_args
from ...custom_exceptions import HalotoolsError
__author__ = ('Andrew Hearin', 'Duncan Campbell')
__all__ = ('npairs_jackknife_3d', )
# cbx_aph: jtags aren't optional here -- the assertion `"jtags1 must be >= 1"` down below fails
# if you don't pass them. I don't think that they should be optional and should probably be
# brought before period. I've done this, but this is a breaking API change so just letting
# you know :) The other option is to keep them as kwargs and just remove the `optional` tag
# in the docs.
# Similarly N_samples isn't optional
[docs]
def npairs_jackknife_3d(sample1, sample2, rbins, jtags1, jtags2, N_samples,
period=None, weights1=None, weights2=None, num_threads=1,
approx_cell1_size=None, approx_cell2_size=None):
r"""
Pair counter used to make jackknife error estimates of real-space pair counter
`~halotools.mock_observables.pair_counters.npairs`.
Parameters
----------
sample1 : array_like
Numpy array of shape (Npts1, 3) containing 3-D positions of points.
See the :ref:`mock_obs_pos_formatting` documentation page, or the
Examples section below, for instructions on how to transform
your coordinate position arrays into the
format accepted by the ``sample1`` and ``sample2`` arguments.
Length units are comoving and assumed to be in Mpc/h, here and throughout Halotools.
sample2 : array_like, optional
Numpy array of shape (Npts2, 3) containing 3-D positions of points.
Should be identical to sample1 for cases of auto-sample pair counts.
Length units are comoving and assumed to be in Mpc/h, here and throughout Halotools.
rbins : array_like
Boundaries defining the bins in which pairs are counted.
Length units are comoving and assumed to be in Mpc/h, here and throughout Halotools.
jtags1 : array_like
Numpy array of shape (Npts1, ) containing integer tags used to define jackknife sample
membership. Tags are in the range [1, N_samples].
The tag '0' is a reserved tag and should not be used.
jtags2 : array_like
Numpy array of shape (Npts2, ) containing integer tags used to define jackknife sample
membership. Tags are in the range [1, N_samples].
The tag '0' is a reserved tag and should not be used.
N_samples : int
Total number of jackknife samples. All values of ``jtags1`` and ``jtags2``
should be in the range [1, N_samples].
period : array_like, optional
Length-3 sequence defining the periodic boundary conditions
in each dimension. If you instead provide a single scalar, Lbox,
period is assumed to be the same in all Cartesian directions.
weights1 : array_like, optional
Numpy array of shape (Npts1, ) containing weights used for weighted pair counts.
weights2 : array_like, optional
Numpy array of shape (Npts2, ) containing weights used for weighted pair counts.
num_threads : int, optional
Number of threads to use in calculation, where parallelization is performed
using the python ``multiprocessing`` module. Default is 1 for a purely serial
calculation, in which case a multiprocessing Pool object will
never be instantiated. A string 'max' may be used to indicate that
the pair counters should use all available cores on the machine.
approx_cell1_size : array_like, optional
Length-3 array serving as a guess for the optimal manner by how points
will be apportioned into subvolumes of the simulation box.
The optimum choice unavoidably depends on the specs of your machine.
Default choice is to use Lbox/10 in each dimension,
which will return reasonable result performance for most use-cases.
Performance can vary sensitively with this parameter, so it is highly
recommended that you experiment with this parameter when carrying out
performance-critical calculations.
approx_cell2_size : array_like, optional
Analogous to ``approx_cell1_size``, but for sample2. See comments for
``approx_cell1_size`` for details.
Returns
-------
N_pairs : array_like
Numpy array of shape (N_samples+1,len(rbins)).
The sub-array N_pairs[0, :] stores numbers of pairs
in the input bins for the entire sample.
The sub-array N_pairs[i, :] stores numbers of pairs
in the input bins for the :math:`i^{\rm th}` jackknife sub-sample.
Notes
-----
Jackknife weights are calculated using a weighting function.
If both points are outside the sample, the weighting function returns 0.
If both points are inside the sample, the weighting function returns (w1 * w2)
If one point is inside, and the other is outside, the weighting function returns (w1 * w2)/2
Examples
--------
For demonstration purposes we create randomly distributed sets of points within a
periodic unit cube.
>>> from halotools.mock_observables.pair_counters import npairs_jackknife_3d
>>> Npts1, Npts2, Lbox = 1000, 1000, 250.
>>> period = [Lbox, Lbox, Lbox]
>>> rbins = np.logspace(-1, 1.5, 15)
>>> x1 = np.random.uniform(0, Lbox, Npts1)
>>> y1 = np.random.uniform(0, Lbox, Npts1)
>>> z1 = np.random.uniform(0, Lbox, Npts1)
>>> x2 = np.random.uniform(0, Lbox, Npts2)
>>> y2 = np.random.uniform(0, Lbox, Npts2)
>>> z2 = np.random.uniform(0, Lbox, Npts2)
We transform our *x, y, z* points into the array shape used by the pair-counter by
taking the transpose of the result of `numpy.vstack`. This boilerplate transformation
is used throughout the `~halotools.mock_observables` sub-package:
>>> sample1 = np.vstack([x1, y1, z1]).T
>>> sample2 = np.vstack([x2, y2, z2]).T
Ordinarily, you would create ``jtags`` for the points by properly subdivide
the points into spatial sub-volumes. For illustration purposes, we'll simply
use randomly assigned sub-volumes as this has no impact on the calling signature:
>>> N_samples = 10
>>> jtags1 = np.random.randint(1, N_samples+1, Npts1)
>>> jtags2 = np.random.randint(1, N_samples+1, Npts2)
>>> result = npairs_jackknife_3d(sample1, sample2, rbins, period = period, jtags1=jtags1, jtags2=jtags2, N_samples = N_samples)
"""
# Process the inputs with the helper function
result = _npairs_3d_process_args(sample1, sample2, rbins, period,
num_threads, approx_cell1_size, approx_cell2_size)
x1in, y1in, z1in, x2in, y2in, z2in = result[0:6]
rbins, period, num_threads, PBCs, approx_cell1_size, approx_cell2_size = result[6:]
xperiod, yperiod, zperiod = period
rmax = np.max(rbins)
search_xlength, search_ylength, search_zlength = rmax, rmax, rmax
# Process the input weights and jackknife-tags with the helper function
weights1, weights2, jtags1, jtags2 = (
_npairs_jackknife_3d_process_weights_jtags(sample1, sample2,
weights1, weights2, jtags1, jtags2, N_samples))
# Compute the estimates for the cell sizes
approx_cell1_size, approx_cell2_size = (
_set_approximate_cell_sizes(approx_cell1_size, approx_cell2_size, period)
)
approx_x1cell_size, approx_y1cell_size, approx_z1cell_size = approx_cell1_size
approx_x2cell_size, approx_y2cell_size, approx_z2cell_size = approx_cell2_size
# Build the rectangular mesh
double_mesh = RectangularDoubleMesh(x1in, y1in, z1in, x2in, y2in, z2in,
approx_x1cell_size, approx_y1cell_size, approx_z1cell_size,
approx_x2cell_size, approx_y2cell_size, approx_z2cell_size,
search_xlength, search_ylength, search_zlength, xperiod, yperiod, zperiod, PBCs)
# Create a function object that has a single argument, for parallelization purposes
engine = partial(npairs_jackknife_3d_engine,
double_mesh, x1in, y1in, z1in, x2in, y2in, z2in,
weights1, weights2, jtags1, jtags2, N_samples, rbins)
# Calculate the cell1 indices that will be looped over by the engine
num_threads, cell1_tuples = _cell1_parallelization_indices(
double_mesh.mesh1.ncells, num_threads)
if num_threads > 1:
pool = multiprocessing.Pool(num_threads)
result = pool.map(engine, cell1_tuples)
counts = np.sum(np.array(result), axis=0)
pool.close()
else:
counts = engine(cell1_tuples[0])
return np.array(counts)
def _npairs_jackknife_3d_process_weights_jtags(sample1, sample2,
weights1, weights2, jtags1, jtags2, N_samples):
"""
"""
# Process weights1 entry and check for consistency.
if weights1 is None:
weights1 = np.array([1.0]*np.shape(sample1)[0], dtype=np.float64)
else:
weights1 = np.asarray(weights1).astype("float64")
if np.shape(weights1)[0] != np.shape(sample1)[0]:
raise HalotoolsError("weights1 should have same len as sample1")
# Process weights2 entry and check for consistency.
if weights2 is None:
weights2 = np.array([1.0]*np.shape(sample2)[0], dtype=np.float64)
else:
weights2 = np.asarray(weights2).astype("float64")
if np.shape(weights2)[0] != np.shape(sample2)[0]:
raise HalotoolsError("weights2 should have same len as sample2")
# Process jtags_1 entry and check for consistency.
jtags1 = np.asarray(jtags1).astype("int")
if np.shape(jtags1)[0] != np.shape(sample1)[0]:
raise HalotoolsError("jtags1 should have same len as sample1")
# Process jtags_2 entry and check for consistency.
jtags2 = np.asarray(jtags2).astype("int")
if np.shape(jtags2)[0] != np.shape(sample2)[0]:
raise HalotoolsError("jtags2 should have same len as sample2")
# Check bounds of jackknife tags
if np.min(jtags1) < 1:
raise HalotoolsError("jtags1 must be >= 1")
if np.min(jtags2) < 1:
raise HalotoolsError("jtags2 must be >= 1")
if np.max(jtags1) > N_samples:
raise HalotoolsError("jtags1 must be <= N_samples")
if np.max(jtags2) > N_samples:
raise HalotoolsError("jtags2 must be <= N_samples")
# throw warning if some tags do not exist
if not np.array_equal(np.unique(jtags1), np.arange(1, N_samples+1)):
warn("Warning: sample1 does not contain points in every jackknife sample.")
if not np.array_equal(np.unique(jtags1), np.arange(1, N_samples+1)):
warn("Warning: sample2 does not contain points in every jackknife sample.")
return weights1, weights2, jtags1, jtags2