def volume(self):
"""The abstract volume associated with this VolumeSource
This object does the heavy lifting to access data in an efficient manner
using a KDTree
"""
if self._volume is None:
mylog.info("Creating volume")
volume = AMRKDTree(self.data_source.ds, data_source=self.data_source)
self._volume = volume
return self._volume
def test_amr_kdtree_coverage():
return #TESTDISABLED
domain_dims = (32, 32, 32)
data = np.zeros(domain_dims) + 0.25
fo = [
ic.CoredSphere(0.05, 0.3, [0.7, 0.4, 0.75], {"density": (0.25, 100.0)})
]
rc = [fm.flagging_method_registry["overdensity"](8.0)]
ug = load_uniform_grid({"density": data}, domain_dims, 1.0)
ds = refine_amr(ug, rc, fo, 5)
kd = AMRKDTree(ds)
volume = kd.count_volume()
yield assert_equal, volume, \
np.prod(ds.domain_right_edge - ds.domain_left_edge)
cells = kd.count_cells()
true_cells = ds.all_data().quantities['TotalQuantity']('Ones')[0]
yield assert_equal, cells, true_cells
# This largely reproduces the AMRKDTree.tree.check_tree() functionality
tree_ok = True
for node in kd.tree.trunk.depth_traverse():
if node.grid is None:
continue
grid = ds.index.grids[node.grid - kd._id_offset]
dds = grid.dds
gle = grid.LeftEdge
nle = node.get_left_edge()
nre = node.get_right_edge()
li = np.rint((nle - gle) / dds).astype('int32')
ri = np.rint((nre - gle) / dds).astype('int32')
dims = (ri - li).astype('int32')
tree_ok *= np.all(grid.LeftEdge <= nle)
tree_ok *= np.all(grid.RightEdge >= nre)
tree_ok *= np.all(dims > 0)
yield assert_equal, True, tree_ok
def test_amr_kdtree_coverage():
return #TESTDISABLED
domain_dims = (32, 32, 32)
data = np.zeros(domain_dims) + 0.25
fo = [ic.CoredSphere(0.05, 0.3, [0.7, 0.4, 0.75],
{"density": (0.25, 100.0)})]
rc = [fm.flagging_method_registry["overdensity"](8.0)]
ug = load_uniform_grid({"density": data}, domain_dims, 1.0)
ds = refine_amr(ug, rc, fo, 5)
kd = AMRKDTree(ds)
volume = kd.count_volume()
yield assert_equal, volume, \
np.prod(ds.domain_right_edge - ds.domain_left_edge)
cells = kd.count_cells()
true_cells = ds.all_data().quantities['TotalQuantity']('Ones')[0]
yield assert_equal, cells, true_cells
# This largely reproduces the AMRKDTree.tree.check_tree() functionality
tree_ok = True
for node in depth_traverse(kd.tree.trunk):
if node.grid is None:
continue
grid = ds.index.grids[node.grid - kd._id_offset]
dds = grid.dds
gle = grid.LeftEdge
nle = get_left_edge(node)
nre = get_right_edge(node)
li = np.rint((nle-gle)/dds).astype('int32')
ri = np.rint((nre-gle)/dds).astype('int32')
dims = (ri - li).astype('int32')
tree_ok *= np.all(grid.LeftEdge <= nle)
tree_ok *= np.all(grid.RightEdge >= nre)
tree_ok *= np.all(dims > 0)
yield assert_equal, True, tree_ok
def __init__(self, ds, positions, xfield='velocity_x', yfield='velocity_x',
zfield='velocity_x', volume=None,
dx=None, length=None, direction=1,
get_magnitude=False):
ParallelAnalysisInterface.__init__(self)
self.ds = ds
self.start_positions = sanitize_length(positions, ds)
self.N = self.start_positions.shape[0]
# I need a data object to resolve the field names to field tuples
# via _determine_fields()
ad = self.ds.all_data()
self.xfield = ad._determine_fields(xfield)[0]
self.yfield = ad._determine_fields(yfield)[0]
self.zfield = ad._determine_fields(zfield)[0]
self.get_magnitude=get_magnitude
self.direction = np.sign(direction)
if volume is None:
volume = AMRKDTree(self.ds)
volume.set_fields([self.xfield,self.yfield,self.zfield],
[False,False,False],
False)
volume.join_parallel_trees()
self.volume = volume
if dx is None:
dx = self.ds.index.get_smallest_dx()
self.dx = sanitize_length(dx, ds)
if length is None:
length = np.max(self.ds.domain_right_edge-self.ds.domain_left_edge)
self.length = sanitize_length(length, ds)
self.steps = int(length/dx)+1
# Fix up the dx.
self.dx = 1.0*self.length/self.steps
self.streamlines = np.zeros((self.N,self.steps,3), dtype='float64')
self.magnitudes = None
if self.get_magnitude:
self.magnitudes = np.zeros((self.N,self.steps), dtype='float64')
def __init__(
self,
ds,
positions,
xfield="velocity_x",
yfield="velocity_x",
zfield="velocity_x",
volume=None,
dx=None,
length=None,
direction=1,
get_magnitude=False,
):
ParallelAnalysisInterface.__init__(self)
self.ds = ds
self.start_positions = sanitize_length(positions, ds)
self.N = self.start_positions.shape[0]
# I need a data object to resolve the field names to field tuples
# via _determine_fields()
ad = self.ds.all_data()
self.xfield = ad._determine_fields(xfield)[0]
self.yfield = ad._determine_fields(yfield)[0]
self.zfield = ad._determine_fields(zfield)[0]
self.get_magnitude = get_magnitude
self.direction = np.sign(direction)
if volume is None:
volume = AMRKDTree(self.ds)
volume.set_fields([self.xfield, self.yfield, self.zfield],
[False, False, False], False)
volume.join_parallel_trees()
self.volume = volume
if dx is None:
dx = self.ds.index.get_smallest_dx()
self.dx = sanitize_length(dx, ds)
if length is None:
length = np.max(self.ds.domain_right_edge -
self.ds.domain_left_edge)
self.length = sanitize_length(length, ds)
self.steps = int(self.length / self.dx) + 1
# Fix up the dx.
self.dx = 1.0 * self.length / self.steps
self.streamlines = np.zeros((self.N, self.steps, 3), dtype="float64")
self.magnitudes = None
if self.get_magnitude:
self.magnitudes = np.zeros((self.N, self.steps), dtype="float64")
def __init__(self,
ds,
positions,
xfield='velocity_x',
yfield='velocity_x',
zfield='velocity_x',
volume=None,
dx=None,
length=None,
direction=1,
get_magnitude=False):
ParallelAnalysisInterface.__init__(self)
self.ds = ds
self.start_positions = np.array(positions)
self.N = self.start_positions.shape[0]
self.xfield = xfield
self.yfield = yfield
self.zfield = zfield
self.get_magnitude = get_magnitude
self.direction = np.sign(direction)
if volume is None:
volume = AMRKDTree(self.ds)
volume.set_fields([self.xfield, self.yfield, self.zfield],
[False, False, False], False)
volume.join_parallel_trees()
self.volume = volume
if dx is None:
dx = self.ds.index.get_smallest_dx()
self.dx = dx
if length is None:
length = np.max(self.ds.domain_right_edge -
self.ds.domain_left_edge)
self.length = length
self.steps = int(length / dx) + 1
# Fix up the dx.
self.dx = 1.0 * self.length / self.steps
self.streamlines = np.zeros((self.N, self.steps, 3), dtype='float64')
self.magnitudes = None
if self.get_magnitude:
self.magnitudes = np.zeros((self.N, self.steps), dtype='float64')
# Using AMRKDTree Homogenized Volumes to examine large datasets
# at lower resolution.
# In this example we will show how to use the AMRKDTree to take a simulation
# with 8 levels of refinement and only use levels 0-3 to render the dataset.
# We begin by loading up yt, and importing the AMRKDTree
import numpy as np
import yt
from yt.utilities.amr_kdtree.api import AMRKDTree
# Load up a dataset and define the kdtree
ds = yt.load('IsolatedGalaxy/galaxy0030/galaxy0030')
kd = AMRKDTree(ds)
# Print out specifics of KD Tree
print("Total volume of all bricks = %i" % kd.count_volume())
print("Total number of cells = %i" % kd.count_cells())
# Define a camera and take an volume rendering.
tf = yt.ColorTransferFunction((-30, -22))
cam = ds.camera([0.5, 0.5, 0.5], [0.2, 0.3, 0.4], 0.10, 256,
tf, volume=kd)
tf.add_layers(4, 0.01, col_bounds=[-27.5, -25.5], colormap='RdBu_r')
cam.snapshot("v1.png", clip_ratio=6.0)
# This rendering is okay, but lets say I'd like to improve it, and I don't want
# to spend the time rendering the high resolution data. What we can do is
# generate a low resolution version of the AMRKDTree and pass that in to the
# camera. We do this by specifying a maximum refinement level of 6.
from yt.utilities.amr_kdtree.api import AMRKDTree
# Load up a dataset and define the kdtree
ds = yt.load('IsolatedGalaxy/galaxy0030/galaxy0030')
im, sc = yt.volume_render(ds, 'density', fname='v0.png')
sc.camera.set_width(ds.arr(100, 'kpc'))
render_source = sc.get_source(0)
kd = render_source.volume
# Print out specifics of KD Tree
print("Total volume of all bricks = %i" % kd.count_volume())
print("Total number of cells = %i" % kd.count_cells())
new_source = ds.all_data()
new_source.max_level = 3
kd_low_res = AMRKDTree(ds, data_source=new_source)
print(kd_low_res.count_volume())
print(kd_low_res.count_cells())
# Now we pass this in as the volume to our camera, and render the snapshot
# again.
render_source.set_volume(kd_low_res)
render_source.set_field('density')
sc.render()
sc.save("v1.png", sigma_clip=6.0)
# This operation was substantiall faster. Now lets modify the low resolution
# rendering until we find something we like.
tf = render_source.transfer_function
def surfacePressureEnergy(clump):
steps = np.array([[-1, -1, -1], [-1, -1, 0], [-1, -1, 1], [-1, 0, -1],
[-1, 0, 0], [-1, 0, 1], [-1, 1, -1], [-1, 1, 0],
[-1, 1, 1], [0, -1, -1], [0, -1, 0], [0, -1, 1],
[0, 0, -1], [0, 0, 1], [0, 1, -1], [0, 1, 0], [0, 1, 1],
[1, -1, -1], [1, -1, 0], [1, -1, 1], [1, 0, -1],
[1, 0, 0], [1, 0, 1], [1, 1, -1], [1, 1, 0], [1, 1, 1]])
mag = np.sqrt((steps**2).sum(axis=1))
mag = np.array([mag, mag, mag]).T.astype(float)
normals = -steps / mag
neighborsTrue = np.ones([3, 3, 3], bool)
neighborsTrue[1, 1, 1] = False
ds = clump.data.ds
amrTree = clump.data.tiles
cm = clump.quantities.center_of_mass()
LE = clump.data.tiles.data_source.base_object.left_edge
RE = clump.data.tiles.data_source.base_object.right_edge
newLE = LE - (RE - LE) / 10
newRE = RE + (RE - LE) / 10
biggerBox = ds.box(newLE, newRE)
amrTree = AMRKDTree(ds, data_source=biggerBox)
surfPressureTerm = 0
grid_mask_dict = dict((g, m) for g, m in clump.data.blocks)
pad = 2
# dictionary to keep track of surface cells that have been
# visited and the corresponding grid they were visited from.
allSurfCells = {'cell': [], 'grid': []}
for grid, clumpMask in clump.data.blocks:
print(grid)
# Get surface cells using masks
surfMask = np.zeros(np.array(clumpMask.shape) + 2 * pad, dtype=bool)
L = surfMask.shape[0] - pad
W = surfMask.shape[1] - pad
H = surfMask.shape[2] - pad
surfMask[2:L, 2:W, 2:H] = clumpMask
padClumpMask = surfMask.copy()
for i in [-1, 0, 1]:
for j in [-1, 0, 1]:
for k in [-1, 0, 1]:
surfMask[pad + i:L + i, pad + j:W + j,
pad + k:H + k] |= clumpMask
surfMask = np.logical_xor(surfMask, padClumpMask)
surfCells_fromPad = np.argwhere(surfMask)
surfCells = surfCells_fromPad - np.array([pad, pad, pad])
surfInGrid = np.all((surfCells >= 0) * (surfCells < clumpMask.shape),
axis=1)
surfOutGrid = np.logical_not(surfInGrid)
# *** Think about looping over surfInGrid surfOutGrid separately *** #
for cell in surfCells:
surfNormals = []
cell = np.array(cell)
center_dds = grid.dds
# get physical position of surface cell
position = grid.LeftEdge + (np.array(cell) + 0.5) * grid.dds
new_position = periodic_position(position, ds)
r = position - cm
# get cooresponding grid that surface cell lives in as well as its
# indicies relatvie to that grid.
trueGrid = ds.index.grids[ amrTree.locate_brick(new_position).grid \
- grid._id_offset]
trueCell = ((new_position - trueGrid.LeftEdge) / trueGrid.dds).v
trueCell = tuple(trueCell.astype(int))
# if surface cell has already been visited in a different grid loop,
# then we have already accounted for the surface pressure from that cell
# and we can skip it.
skip = False
for g, c in zip(allSurfCells['grid'], allSurfCells['cell']):
if (not (np.array(c) -
np.array(trueCell)).sum()) and grid.id != g:
skip = True
break
if skip:
continue
allSurfCells['cell'].append(trueCell)
allSurfCells['grid'].append(grid.id)
#print("\tgrid, trueGrid", grid, trueGrid)
#print("\tcell, trueCell", cell, trueCell)
# if trueCell is part of the clump in trueGrid it is not actually
# a surface cell and we can skip it. Also if trueGrid is not in
# grid_mask_dict, that means trueCell is not in this clump and is
# a true surface cells
possiblyInClump = False
try:
trueClumpMask = grid_mask_dict[trueGrid]
possiblyInClump = True
except:
pass
trueCellInClump = False
if possiblyInClump:
#print("\ttrue cell", trueCell, " in clump?")
trueCellMask = np.zeros_like(trueClumpMask)
trueCellMask[trueCell] = True
trueCellInClump = np.logical_and(trueCellMask,
trueClumpMask).any()
if trueCellInClump:
# not a true surface cell, continue to next one
#print("\t\tYES. Next surface cell")
continue
#print("\t\tNO. True surface cell")
# if we made it here we have a true surface cell. We know need
# the pressure and the surface normals of all the cells in the
# clump it is touching.
pressure = trueGrid['pressure'][trueCell]
dS = grid.dds[0] * grid.dds[0]
# find grid and cell indicies of neighboring cells
# which may be in separate grids
cell = np.array(cell)
center_dds = grid.dds
grids = np.empty(26, dtype='object')
neigborCells = np.empty([26, 3], dtype='int64')
offs = 0.5 * (center_dds + amrTree.sdx)
new_neigborCells = cell + steps
# index mask for cells this in grid
in_grid = np.all( (new_neigborCells >=0) * \
(new_neigborCells < grid.ActiveDimensions),
axis=1 )
# index mask for cells in a neighboring grid
not_in_grid = np.logical_not(in_grid)
# physical positions of neighboring cells (assumes a periodic box)
new_positions = position + steps * offs
new_positions = [periodic_position(p, ds) for p in new_positions]
grids[in_grid] = grid
neigborCells[in_grid] = new_neigborCells[in_grid]
# neigboring cell indicies of cells that are in other grids
get_them = np.argwhere(not_in_grid).ravel()
if not_in_grid.any():
# get grids containing cells outside current grids
grids[get_them] = \
[ds.index.grids[amrTree.locate_brick(new_positions[i]).grid - \
grid._id_offset] for i in get_them ]
# get cell location indicies in grids outside current grid
neigborCells[not_in_grid] = \
[ (new_positions[i] - grids[i].LeftEdge)/grids[i].dds \
for i in get_them ]
neigborCells = [tuple(_cs) for _cs in neigborCells]
# find which neigbor cells are in the clump. We only care about
# neighbor cells that are within the clump becaause we need to
# know the normal vectors for those ones.
#print("\t\tFind neigbor cells within clump")
for nGrid, nCell, norm in zip(grids, neigborCells, normals):
nCell = tuple(nCell)
#print("\t\tnCell, nGrid", nCell, nGrid)
try:
nClumpMask = grid_mask_dict[nGrid]
except:
# if nGrid is not in dict of grids for this clump,
# then nCell cannot be in this clump
continue
# is nCell in nClump?
nCellMask = np.zeros_like(nClumpMask)
nCellMask[nCell] = True
inClump = np.logical_and(nCellMask, nClumpMask).any()
if inClump:
#print("\t\t\tIN clump")
surfNormals.append(norm)
#else:
#print("\t\t\tNOT in clump")
surfNormals = np.array(surfNormals)
rs = ds.arr(np.empty_like(surfNormals), r.units)
rs[:] = r
# surface pressure term = Integrate{ p r dot dS }
r_dot_dS = (rs * surfNormals).sum(axis=1) * dS
surfPressureTerm += (pressure * r_dot_dS).sum()
#print(surfPressureTerm.to('erg'))
return surfPressureTerm