def _test4():
from numpy import sin, zeros, exp
# check vectorization evaluation:
g = UniformBoxGrid(min=(0, 0), max=(1, 1), division=(3, 3))
try:
g.vectorized_eval(lambda x, y: 2)
except TypeError as msg:
# fine, expect to arrive here
print '*** Expected error in this test:', msg
try:
g.vectorized_eval(lambda x, y: zeros((2, 2)) + 2)
except IndexError as msg:
# fine, expect to arrive here
print '*** Expected error in this test:', msg
a = g.vectorized_eval(lambda x, y: sin(x) * exp(y - x))
print a
a = g.vectorized_eval(lambda x, y: zeros(g.shape) + 2)
print a
def _test4():
from numpy import sin, zeros, exp
# check vectorization evaluation:
g = UniformBoxGrid(min=(0,0), max=(1,1), division=(3,3))
try:
g.vectorized_eval(lambda x,y: 2)
except TypeError as msg:
# fine, expect to arrive here
print '*** Expected error in this test:', msg
try:
g.vectorized_eval(lambda x,y: zeros((2,2))+2)
except IndexError as msg:
# fine, expect to arrive here
print '*** Expected error in this test:', msg
a = g.vectorized_eval(lambda x,y: sin(x)*exp(y-x))
print a
a = g.vectorized_eval(lambda x,y: zeros(g.shape)+2)
print a
def __init__(
self,
min=(0, 0), # minimum coordinates
max=(1, 1), # maximum coordinates
division=(4, 4), # cell divisions
dirnames=('x', 'y', 'z')): # names of the directions
"""
Initialize a BoxGrid by giving domain range (minimum and
maximum coordinates: min and max tuples/lists/arrays)
and number of cells in each space direction (division tuple/list/array).
The dirnames tuple/list holds the names of the coordinates in
the various spatial directions.
>>> g = UniformBoxGrid(min=0, max=1, division=10)
>>> g = UniformBoxGrid(min=(0,-1), max=(1,1), division=(10,4))
>>> g = UniformBoxGrid(min=(0,0,-1), max=(2,1,1), division=(2,3,5))
"""
# Allow int/float specifications in one-dimensional grids
# (just turn to lists for later multi-dimensional processing)
if isinstance(min, (int, float)):
min = [min]
if isinstance(max, (int, float)):
max = [max]
if isinstance(division, (int, float)):
division = [division]
if isinstance(dirnames, str):
dirnames = [dirnames]
self.nsd = len(min)
# strip dirnames down to right space dim (in case the default
# with three components were unchanged by the user):
dirnames = dirnames[:self.nsd]
# check consistent lengths:
for a in max, division:
if len(a) != self.nsd:
raise ValueError(
'Incompatible lengths of arguments to constructor'\
' (%d != %d)' % (len(a), self.nsd))
self.min_coor = array(min, float)
self.max_coor = array(max, float)
self.dirnames = dirnames
self.division = division
self.coor = [None] * self.nsd
self.shape = [0] * self.nsd
self.delta = zeros(self.nsd)
for i in range(self.nsd):
self.delta[i] = \
(self.max_coor[i] - self.min_coor[i])/float(self.division[i])
self.shape[i] = self.division[i] + 1 # no of grid points
self.coor[i] = \
linspace(self.min_coor[i], self.max_coor[i], self.shape[i])
self._more_init()
def __init__(self,
min=(0,0), # minimum coordinates
max=(1,1), # maximum coordinates
division=(4,4), # cell divisions
dirnames=('x', 'y', 'z')): # names of the directions
"""
Initialize a BoxGrid by giving domain range (minimum and
maximum coordinates: min and max tuples/lists/arrays)
and number of cells in each space direction (division tuple/list/array).
The dirnames tuple/list holds the names of the coordinates in
the various spatial directions.
>>> g = UniformBoxGrid(min=0, max=1, division=10)
>>> g = UniformBoxGrid(min=(0,-1), max=(1,1), division=(10,4))
>>> g = UniformBoxGrid(min=(0,0,-1), max=(2,1,1), division=(2,3,5))
"""
# Allow int/float specifications in one-dimensional grids
# (just turn to lists for later multi-dimensional processing)
if isinstance(min, (int,float)):
min = [min]
if isinstance(max, (int,float)):
max = [max]
if isinstance(division, (int,float)):
division = [division]
if isinstance(dirnames, str):
dirnames = [dirnames]
self.nsd = len(min)
# strip dirnames down to right space dim (in case the default
# with three components were unchanged by the user):
dirnames = dirnames[:self.nsd]
# check consistent lengths:
for a in max, division:
if len(a) != self.nsd:
raise ValueError(
'Incompatible lengths of arguments to constructor'\
' (%d != %d)' % (len(a), self.nsd))
self.min_coor = array(min, float)
self.max_coor = array(max, float)
self.dirnames = dirnames
self.division = division
self.coor = [None]*self.nsd
self.shape = [0]*self.nsd
self.delta = zeros(self.nsd)
for i in range(self.nsd):
self.delta[i] = \
(self.max_coor[i] - self.min_coor[i])/float(self.division[i])
self.shape[i] = self.division[i] + 1 # no of grid points
self.coor[i] = \
linspace(self.min_coor[i], self.max_coor[i], self.shape[i])
self._more_init()
def _test(g, points=None):
print 'g=%s' % str(g)
# dump all the contents of a grid object:
import scitools.misc
scitools.misc.dump(g, hide_nonpublic=False)
from numpy import zeros
def fv(*args):
# vectorized evaluation function
return zeros(g.shape) + 2
def fs(*args):
# scalar version
return 2
fv_arr = g.vectorized_eval(fv)
fs_arr = zeros(g.shape)
coor = [0.0] * g.nsd
itparts = [
'all', 'interior', 'all_boundary', 'interior_boundary', 'corners'
]
if g.nsd == 3:
itparts += ['all_edges', 'interior_edges']
for domain_part in itparts:
print '\niterator over "%s"' % domain_part
for i in g.iter(domain_part, vectorized_version=False):
if isinstance(i, int): i = [i] # wrap index as list (if 1D)
for k in range(g.nsd):
coor[k] = g.coor[k][i[k]]
print i, coor
if domain_part == 'all': # fs_arr shape corresponds to all points
fs_arr[i] = fs(*coor)
print 'vectorized iterator over "%s":' % domain_part
for slices in g.iter(domain_part, vectorized_version=True):
if domain_part == 'all':
fs_arr[slices] = fv(*g.coor)
# else: more complicated
print slices
# boundary slices...
if points is not None:
print '\n\nInterpolation in', g, '\n', g.coor
for p in points:
index, distance = g.locate_cell(p)
print 'point %s is in cell %s, distance=%s' % (p, index, distance)
def _test(g, points=None):
print 'g=%s' % str(g)
# dump all the contents of a grid object:
import scitools.misc; scitools.misc.dump(g, hide_nonpublic=False)
from numpy import zeros
def fv(*args):
# vectorized evaluation function
return zeros(g.shape)+2
def fs(*args):
# scalar version
return 2
fv_arr = g.vectorized_eval(fv)
fs_arr = zeros(g.shape)
coor = [0.0]*g.nsd
itparts = ['all', 'interior', 'all_boundary', 'interior_boundary',
'corners']
if g.nsd == 3:
itparts += ['all_edges', 'interior_edges']
for domain_part in itparts:
print '\niterator over "%s"' % domain_part
for i in g.iter(domain_part, vectorized_version=False):
if isinstance(i, int): i = [i] # wrap index as list (if 1D)
for k in range(g.nsd):
coor[k] = g.coor[k][i[k]]
print i, coor
if domain_part == 'all': # fs_arr shape corresponds to all points
fs_arr[i] = fs(*coor)
print 'vectorized iterator over "%s":' % domain_part
for slices in g.iter(domain_part, vectorized_version=True):
if domain_part == 'all':
fs_arr[slices] = fv(*g.coor)
# else: more complicated
print slices
# boundary slices...
if points is not None:
print '\n\nInterpolation in', g, '\n', g.coor
for p in points:
index, distance = g.locate_cell(p)
print 'point %s is in cell %s, distance=%s' % (p, index, distance)
def flow(*args):
# xx,yy,zz,vv = flow()
# xx,yy,zz,vv = flow(n)
# xx,yy,zz,vv = flow(xx,yy,zz)
if len(args) == 0:
xx, yy, zz = ndgrid(linspace(0.1, 10, 50),
linspace(-3, 3, 25),
linspace(-3, 3, 25),
sparse=False)
elif len(args) == 1:
n = int(args[0])
xx, yy, zz = ndgrid(linspace(0.1, 10, 2 * n),
linspace(-3, 3, n),
linspace(-3, 3, n),
sparse=False)
elif len(args) == 3:
xx, yy, zz = args
else:
raise SyntaxError("Invalid number of arguments.")
# convert to spherical coordinates:
theta = arctan2(zz, yy)
phi = arctan2(xx, sqrt(yy**2 + zz**2))
r = sqrt(xx**2 + yy**2 + zz**2)
rv = 2 / r * (3 / (2 - cos(phi))**2 - 1)
phiv = -2 * sin(phi) / (2 - cos(phi)) / r
thetav = zeros(shape(r))
# convert back to cartesian coordinates:
xv = rv * cos(phiv) * cos(thetav)
yv = rv * cos(phiv) * sin(thetav)
zv = rv * sin(phiv)
vv = log(sqrt(xv**2 + yv**2 + zv**2))
return xx, yy, zz, vv
def flow(*args):
# xx,yy,zz,vv = flow()
# xx,yy,zz,vv = flow(n)
# xx,yy,zz,vv = flow(xx,yy,zz)
if len(args) == 0:
xx, yy, zz = ndgrid(linspace(0.1, 10, 50),
linspace(-3, 3, 25),
linspace(-3, 3, 25),
sparse=False)
elif len(args) == 1:
n = int(args[0])
xx, yy, zz = ndgrid(linspace(0.1, 10, 2*n),
linspace(-3, 3, n),
linspace(-3, 3, n),
sparse=False)
elif len(args) == 3:
xx, yy, zz = args
else:
raise SyntaxError("Invalid number of arguments.")
# convert to spherical coordinates:
theta = arctan2(zz, yy)
phi = arctan2(xx, sqrt(yy**2 + zz**2))
r = sqrt(xx**2 + yy**2 + zz**2)
rv = 2/r*(3/(2-cos(phi))**2 - 1)
phiv = -2*sin(phi)/(2-cos(phi))/r
thetav = zeros(shape(r))
# convert back to cartesian coordinates:
xv = rv*cos(phiv)*cos(thetav)
yv = rv*cos(phiv)*sin(thetav)
zv = rv*sin(phiv)
vv = log(sqrt(xv**2 + yv**2 + zv**2))
return xx, yy, zz, vv
def fv(*args):
# vectorized evaluation function
return zeros(g.shape) + 2
def locate_cell(self, point):
"""
Given a point (x, (x,y), (x,y,z)), locate the cell in which
the point is located, and return
1) the (i,j,k) vertex index
of the "lower-left" grid point in this cell,
2) the distances (dx, (dx,dy), or (dx,dy,dz)) from this point to
the given point,
3) a boolean list if point matches the
coordinates of any grid lines. If a point matches
the last grid point in a direction, the cell index is
set to the max index such that the (i,j,k) index can be used
directly for look up in an array of values. The corresponding
element in the distance array is then set 0.
4) the indices of the nearest grid point.
The method only works for uniform grid spacing.
Used for interpolation.
>>> g1 = UniformBoxGrid(min=0, max=1, division=4)
>>> cell_index, distance, match, nearest = g1.locate_cell(0.7)
>>> print cell_index
[2]
>>> print distance
[ 0.2]
>>> print match
[False]
>>> print nearest
[3]
>>>
>>> g1.locate_cell(0.5)
([2], array([ 0.]), [True], [2])
>>>
>>> g2 = UniformBoxGrid.init_fromstring('[-1,1]x[-1,2] [0:3]x[0:4]')
>>> print g2.coor
[array([-1. , -0.33333333, 0.33333333, 1. ]), array([-1. , -0.25, 0.5 , 1.25, 2. ])]
>>> g2.locate_cell((0.2,0.2))
([1, 1], array([ 0.53333333, 0.45 ]), [False, False], [2, 2])
>>> g2.locate_cell((1,2))
([3, 4], array([ 0., 0.]), [True, True], [3, 4])
>>>
>>>
>>>
"""
if isinstance(point, (int, float)):
point = [point]
nsd = len(point)
if nsd != self.nsd:
raise ValueError('point=%s has wrong dimension (this is a %dD grid!)' % \
(point, self.nsd))
#index = zeros(nsd, int)
index = [0] * nsd
distance = zeros(nsd)
grid_point = [False] * nsd
nearest_point = [0] * nsd
for i, coor in enumerate(point):
# is point inside the domain?
if coor < self.min_coor[i] or coor > self.max_coor[i]:
raise ValueError(
'locate_cell: point=%s is outside the domain [%s,%s]' % \
point, self.min_coor[i], self.max_coor[i])
index[i] = int((coor - self.min_coor[i]) //
self.delta[i]) # (need integer division)
distance[i] = coor - (self.min_coor[i] + index[i] * self.delta[i])
if distance[i] > self.delta[i] / 2:
nearest_point[i] = index[i] + 1
else:
nearest_point[i] = index[i]
if abs(distance[i]) < self.tolerance:
grid_point[i] = True
nearest_point[i] = index[i]
if (abs(distance[i] - self.delta[i])) < self.tolerance:
# last cell, update index such that it coincides with the point
grid_point[i] = True
index[i] += 1
nearest_point[i] = index[i]
distance[i] = 0.0
return index, distance, grid_point, nearest_point
#!/usr/bin/env python
import sys, math
import oscillator
# default values of input parameters:
m = 1.0; b = 0.7; c = 5.0; func = 'y'; A = 5.0; w = 2*math.pi
y0 = 0.2; tstop = 30.0; dt = 0.05
from scitools.numpyutils import seq, zeros
maxsteps = 10000
#maxsteps = 10
n = 2
y = zeros((n,maxsteps))
step = 0; time = 0.0
import Gnuplot
g1 = Gnuplot.Gnuplot(persist=1) # (y(t),dy/dt) plot
g2 = Gnuplot.Gnuplot(persist=1) # y(t) plot
def setprm():
oscillator.scan2(m, b, c, A, w, y0, tstop, dt, func)
def rewind(nsteps=0):
global step, time
if nsteps == 0: # start all over again?
step = 0
time = 0.0
else: # rewind nsteps
step -= nsteps
time -= nsteps*dt
def fv(*args):
# vectorized evaluation function
return zeros(g.shape)+2
def locate_cell(self, point):
"""
Given a point (x, (x,y), (x,y,z)), locate the cell in which
the point is located, and return
1) the (i,j,k) vertex index
of the "lower-left" grid point in this cell,
2) the distances (dx, (dx,dy), or (dx,dy,dz)) from this point to
the given point,
3) a boolean list if point matches the
coordinates of any grid lines. If a point matches
the last grid point in a direction, the cell index is
set to the max index such that the (i,j,k) index can be used
directly for look up in an array of values. The corresponding
element in the distance array is then set 0.
4) the indices of the nearest grid point.
The method only works for uniform grid spacing.
Used for interpolation.
>>> g1 = UniformBoxGrid(min=0, max=1, division=4)
>>> cell_index, distance, match, nearest = g1.locate_cell(0.7)
>>> print cell_index
[2]
>>> print distance
[ 0.2]
>>> print match
[False]
>>> print nearest
[3]
>>>
>>> g1.locate_cell(0.5)
([2], array([ 0.]), [True], [2])
>>>
>>> g2 = UniformBoxGrid.init_fromstring('[-1,1]x[-1,2] [0:3]x[0:4]')
>>> print g2.coor
[array([-1. , -0.33333333, 0.33333333, 1. ]), array([-1. , -0.25, 0.5 , 1.25, 2. ])]
>>> g2.locate_cell((0.2,0.2))
([1, 1], array([ 0.53333333, 0.45 ]), [False, False], [2, 2])
>>> g2.locate_cell((1,2))
([3, 4], array([ 0., 0.]), [True, True], [3, 4])
>>>
>>>
>>>
"""
if isinstance(point, (int,float)):
point = [point]
nsd = len(point)
if nsd != self.nsd:
raise ValueError('point=%s has wrong dimension (this is a %dD grid!)' % \
(point, self.nsd))
#index = zeros(nsd, int)
index = [0]*nsd
distance = zeros(nsd)
grid_point = [False]*nsd
nearest_point = [0]*nsd
for i, coor in enumerate(point):
# is point inside the domain?
if coor < self.min_coor[i] or coor > self.max_coor[i]:
raise ValueError(
'locate_cell: point=%s is outside the domain [%s,%s]' % \
point, self.min_coor[i], self.max_coor[i])
index[i] = int((coor - self.min_coor[i])//self.delta[i]) # (need integer division)
distance[i] = coor - (self.min_coor[i] + index[i]*self.delta[i])
if distance[i] > self.delta[i]/2:
nearest_point[i] = index[i] + 1
else:
nearest_point[i] = index[i]
if abs(distance[i]) < self.tolerance:
grid_point[i] = True
nearest_point[i] = index[i]
if (abs(distance[i] - self.delta[i])) < self.tolerance:
# last cell, update index such that it coincides with the point
grid_point[i] = True
index[i] += 1
nearest_point[i] = index[i]
distance[i] = 0.0
return index, distance, grid_point, nearest_point
