Skip to content

GitLab

N
NIFTy
Commit 3454501e authored by Jait Dixit's avatar Jait Dixit
Browse files
Options

Merge branch 'feature/field_multiple_space' into smooth_operator

parents aaa02478 f21c8952
Expand all Hide whitespace changes
Inline Side-by-side
Showing with 1037 additions and 1279 deletions
nifty/__init__.py
View file @ 3454501e
......@@ -55,6 +55,8 @@ from spaces import *
from operators import *
from probing import *
from demos import get_demo_dir
#import pyximport; pyximport.install(pyimport = True)
nifty/field.py
View file @ 3454501e
......@@ -4,7 +4,8 @@ import numpy as np
from d2o import distributed_data_object,\
STRATEGIES as DISTRIBUTION_STRATEGIES
from nifty.config import about, nifty_configuration as gc
from nifty.config import about,\
nifty_configuration as gc
from nifty.field_types import FieldType
......@@ -489,15 +490,23 @@ class Field(object):
else:
dtype = np.dtype(dtype)
casted_x = self._actual_cast(x, dtype=dtype)
for ind, sp in enumerate(self.domain):
casted_x = sp.pre_cast(x,
axes=self.domain_axes[ind])
for ind, ft in enumerate(self.field_type):
casted_x = ft.pre_cast(casted_x,
axes=self.field_type_axes[ind])
casted_x = self._actual_cast(casted_x, dtype=dtype)
for ind, sp in enumerate(self.domain):
casted_x = sp.complement_cast(casted_x,
axes=self.domain_axes[ind])
casted_x = sp.post_cast(casted_x,
axes=self.domain_axes[ind])
for ind, ft in enumerate(self.field_type):
casted_x = ft.complement_cast(casted_x,
axes=self.field_type_axes[ind])
casted_x = ft.post_cast(casted_x,
axes=self.field_type_axes[ind])
return casted_x
......@@ -576,7 +585,7 @@ class Field(object):
new_field.__class__ = self.__class__
# copy domain, codomain and val
for key, value in self.__dict__.items():
if key != 'val':
if key != '_val':
new_field.__dict__[key] = value
else:
new_field.__dict__[key] = self.val.copy_empty()
......
nifty/field_types/field_type.py
View file @ 3454501e
......@@ -51,5 +51,8 @@ class FieldType(object):
return result_array
def complement_cast(self, x, axes=None):
def pre_cast(self, x, axes=None):
return x
def post_cast(self, x, axes=None):
return x
nifty/operators/diagonal_operator/diagonal_operator.py
View file @ 3454501e
......@@ -15,12 +15,11 @@ class DiagonalOperator(EndomorphicOperator):
# ---Overwritten properties and methods---
def __init__(self, domain=(), field_type=(), implemented=False,
def __init__(self, domain=(), field_type=(), implemented=True,
diagonal=None, bare=False, copy=True,
distribution_strategy=None):
super(DiagonalOperator, self).__init__(domain=domain,
field_type=field_type,
implemented=implemented)
field_type=field_type)
self._implemented = bool(implemented)
......@@ -30,61 +29,27 @@ class DiagonalOperator(EndomorphicOperator):
elif isinstance(diagonal, Field):
distribution_strategy = diagonal.distribution_strategy
self.distribution_strategy = self._parse_distribution_strategy(
self._distribution_strategy = self._parse_distribution_strategy(
distribution_strategy=distribution_strategy,
val=diagonal)
self.set_diagonal(diagonal=diagonal, bare=bare, copy=copy)
def _times(self, x, spaces, types):
# if the distribution_strategy of self is sub-slice compatible to
# the one of x, reshape the local data of self and apply it directly
active_axes = []
if spaces is None:
for axes in x.domain_axes:
active_axes += axes
else:
for space_index in spaces:
active_axes += x.domain_axes[space_index]
if types is None:
for axes in x.field_type_axes:
active_axes += axes
else:
for type_index in types:
active_axes += x.field_type_axes[type_index]
if x.val.get_axes_local_distribution_strategy(active_axes) == \
self.distribution_strategy:
local_data = self._diagonal.val.get_local_data(copy=False)
# check if domains match completely
# -> multiply directly
# check if axes_local_distribution_strategy matches.
# If yes, extract local data of self.diagonal and x and use numpy
# reshape.
# assert that indices in spaces and types are striktly increasing
# otherwise a wild transpose would be necessary
# build new shape (1,1,x,1,y,1,1,z)
# copy self.diagonal into new shape
# apply reshaped array to x
return self._times_helper(x, spaces, types,
operation=lambda z: z.__mul__)
def _adjoint_times(self, x, spaces, types):
pass
return self._times_helper(x, spaces, types,
operation=lambda z: z.adjoint().__mul__)
def _inverse_times(self, x, spaces, types):
pass
return self._times_helper(x, spaces, types,
operation=lambda z: z.__rdiv__)
def _adjoint_inverse_times(self, x, spaces, types):
pass
def _inverse_adjoint_times(self, x, spaces, types):
pass
return self._times_helper(x, spaces, types,
operation=lambda z: z.adjoint().__rdiv__)
def diagonal(self, bare=False, copy=True):
if bare:
......@@ -178,3 +143,49 @@ class DiagonalOperator(EndomorphicOperator):
# store the diagonal-field
self._diagonal = f
def _times_helper(self, x, spaces, types, operation):
# if the domain and field_type match directly
# -> multiply the fields directly
if x.domain == self.domain and x.field_type == self.field_type:
# here the actual multiplication takes place
return operation(self.diagonal(copy=False))(x)
# if the distribution_strategy of self is sub-slice compatible to
# the one of x, reshape the local data of self and apply it directly
active_axes = []
if spaces is None:
for axes in x.domain_axes:
active_axes += axes
else:
for space_index in spaces:
active_axes += x.domain_axes[space_index]
if types is None:
for axes in x.field_type_axes:
active_axes += axes
else:
for type_index in types:
active_axes += x.field_type_axes[type_index]
axes_local_distribution_strategy = \
x.val.get_axes_local_distribution_strategy(active_axes)
if axes_local_distribution_strategy == self.distribution_strategy:
local_diagonal = self._diagonal.val.get_local_data(copy=False)
else:
# create an array that is sub-slice compatible
redistr_diagonal_val = self._diagonal.val.copy(
distribution_strategy=axes_local_distribution_strategy)
local_diagonal = redistr_diagonal_val.get_local_data(copy=False)
reshaper = [x.shape[i] if i in active_axes else 1
for i in xrange(len(x.shape))]
reshaped_local_diagonal = np.reshape(local_diagonal, reshaper)
# here the actual multiplication takes place
local_result = operation(reshaped_local_diagonal)(
x.val.get_local_data(copy=False))
result_field = x.copy_empty(dtype=local_result.dtype)
result_field.val.set_local_data(local_result, copy=False)
return result_field
nifty/operators/fft_operator/fft_operator.py
View file @ 3454501e
from nifty.config import about
import nifty.nifty_utilities as utilities
from nifty.spaces import RGSpace,\
GLSpace,\
HPSpace,\
LMSpace
from nifty.operators.linear_operator import LinearOperator
from transformations import TransformationFactory
from transformations import RGRGTransformation,\
LMGLTransformation,\
LMHPTransformation,\
GLLMTransformation,\
HPLMTransformation,\
TransformationCache
class FFTOperator(LinearOperator):
# ---Class attributes---
default_codomain_dictionary = {RGSpace: RGSpace,
HPSpace: LMSpace,
GLSpace: LMSpace,
LMSpace: HPSpace,
}
transformation_dictionary = {(RGSpace, RGSpace): RGRGTransformation,
(HPSpace, LMSpace): HPLMTransformation,
(GLSpace, LMSpace): GLLMTransformation,
(LMSpace, HPSpace): LMHPTransformation,
(LMSpace, GLSpace): LMGLTransformation
}
# ---Overwritten properties and methods---
def __init__(self, domain=(), field_type=(), target=None):
def __init__(self, domain=(), field_type=(), target=None, module=None):
super(FFTOperator, self).__init__(domain=domain,
field_type=field_type)
# Initialize domain and target
if len(self.domain) != 1:
raise ValueError(about._errors.cstring(
'ERROR: TransformationOperator accepts only exactly one '
......@@ -24,17 +51,30 @@ class FFTOperator(LinearOperator):
))
if target is None:
target = utilities.get_default_codomain(self.domain[0])
target = (self.get_default_codomain(self.domain[0]), )
self._target = self._parse_domain(target)
self._forward_transformation = TransformationFactory.create(
self.domain[0], self.target[0]
)
self._inverse_transformation = TransformationFactory.create(
self.target[0], self.domain[0]
)
# Create transformation instances
try:
forward_class = self.transformation_dictionary[
(self.domain[0].__class__, self.target[0].__class__)]
except KeyError:
raise TypeError(about._errors.cstring(
"ERROR: No forward transformation for domain-target pair "
"found."))
try:
backward_class = self.transformation_dictionary[
(self.target[0].__class__, self.domain[0].__class__)]
except KeyError:
raise TypeError(about._errors.cstring(
"ERROR: No backward transformation for domain-target pair "
"found."))
self._forward_transformation = TransformationCache.create(
forward_class, self.domain[0], self.target[0], module=module)
self._backward_transformation = TransformationCache.create(
backward_class, self.target[0], self.domain[0], module=module)
def _times(self, x, spaces, types):
spaces = utilities.cast_axis_to_tuple(spaces, len(x.domain))
......@@ -69,7 +109,7 @@ class FFTOperator(LinearOperator):
else:
axes = x.domain_axes[spaces[0]]
new_val = self._inverse_transformation.transform(x.val, axes=axes)
new_val = self._backward_transformation.transform(x.val, axes=axes)
if spaces is None:
result_domain = self.domain
......@@ -99,3 +139,22 @@ class FFTOperator(LinearOperator):
@property
def unitary(self):
return True
# ---Added properties and methods---
@classmethod
def get_default_codomain(cls, domain):
domain_class = domain.__class__
try:
codomain_class = cls.default_codomain_dictionary[domain_class]
except KeyError:
raise TypeError(about._errors.cstring("ERROR: unknown domain"))
try:
transform_class = cls.transformation_dictionary[(domain_class,
codomain_class)]
except KeyError:
raise TypeError(about._errors.cstring(
"ERROR: No transformation for domain-codomain pair found."))
return transform_class.get_codomain(domain)
nifty/operators/fft_operator/transformations/__init__.py
View file @ 3454501e
......@@ -4,4 +4,4 @@ from hplmtransformation import HPLMTransformation
from lmgltransformation import LMGLTransformation
from lmhptransformation import LMHPTransformation
from transformation_factory import TransformationFactory
\ No newline at end of file
from transformation_cache import TransformationCache
\ No newline at end of file
nifty/operators/fft_operator/transformations/gllmtransformation.py
View file @ 3454501e
import numpy as np
from transformation import Transformation
from d2o import distributed_data_object
from nifty.config import dependency_injector as gdi
import nifty.nifty_utilities as utilities
from nifty.config import dependency_injector as gdi,\
about
from nifty import GLSpace, LMSpace
from slicing_transformation import SlicingTransformation
import lm_transformation_factory as ltf
gl = gdi.get('libsharp_wrapper_gl')
libsharp = gdi.get('libsharp_wrapper_gl')
class GLLMTransformation(Transformation):
class GLLMTransformation(SlicingTransformation):
# ---Overwritten properties and methods---
def __init__(self, domain, codomain=None, module=None):
if 'libsharp_wrapper_gl' not in gdi:
raise ImportError("The module libsharp is needed but not available")
if codomain is None:
self.domain = domain
self.codomain = self.get_codomain(domain)
elif self.check_codomain(domain, codomain):
self.domain = domain
self.codomain = codomain
else:
raise ValueError("ERROR: Incompatible codomain!")
raise ImportError(about._errors.cstring(
"The module libsharp is needed but not available."))
@staticmethod
def get_codomain(domain):
super(GLLMTransformation, self).__init__(domain, codomain, module)
# ---Mandatory properties and methods---
@classmethod
def get_codomain(cls, domain):
"""
Generates a compatible codomain to which transformations are
reasonable, i.e.\ an instance of the :py:class:`lm_space` class.
......@@ -38,96 +37,89 @@ class GLLMTransformation(Transformation):
codomain : LMSpace
A compatible codomain.
"""
if domain is None:
raise ValueError('ERROR: cannot generate codomain for None')
if not isinstance(domain, GLSpace):
raise TypeError('ERROR: domain needs to be a GLSpace')
raise TypeError(about._errors.cstring(
"ERROR: domain needs to be a GLSpace"))
nlat = domain.nlat
lmax = nlat - 1
mmax = nlat - 1
if domain.dtype == np.dtype('float32'):
return LMSpace(lmax=lmax, mmax=mmax, dtype=np.complex64)
return_dtype = np.float32
else:
return LMSpace(lmax=lmax, mmax=mmax, dtype=np.complex128)
return_dtype = np.float64
result = LMSpace(lmax=lmax, mmax=mmax, dtype=return_dtype)
cls.check_codomain(domain, result)
return result
@staticmethod
def check_codomain(domain, codomain):
if not isinstance(domain, GLSpace):
raise TypeError('ERROR: domain is not a GLSpace')
if codomain is None:
return False
raise TypeError(about._errors.cstring(
"ERROR: domain is not a GLSpace"))
if not isinstance(codomain, LMSpace):
raise TypeError('ERROR: codomain must be a LMSpace.')
raise TypeError(about._errors.cstring(
"ERROR: codomain must be a LMSpace."))
nlat = domain.nlat
nlon = domain.nlon
lmax = codomain.lmax
mmax = codomain.mmax
if (nlon != 2 * nlat - 1) or (lmax != nlat - 1) or (lmax != mmax):
return False
if lmax != mmax:
raise ValueError(about._errors.cstring(
'ERROR: codomain has lmax != mmax.'))
return True
def transform(self, val, axes=None, **kwargs):
"""
GL -> LM transform method.
if lmax != nlat - 1:
raise ValueError(about._errors.cstring(
'ERROR: codomain has lmax != nlat - 1.'))
Parameters
----------
val : np.ndarray or distributed_data_object
The value array which is to be transformed
if nlon != 2 * nlat - 1:
raise ValueError(about._errors.cstring(
'ERROR: domain has nlon != 2 * nlat - 1.'))
axes : None or tuple
The axes along which the transformation should take place
return None
"""
if self.domain.discrete:
val = self.domain.weight(val, power=-0.5, axes=axes)
# shorthands for transform parameters
def _transformation_of_slice(self, inp, **kwargs):
nlat = self.domain.nlat
nlon = self.domain.nlon
lmax = self.codomain.lmax
mmax = self.codomain.mmax
if isinstance(val, distributed_data_object):
temp_val = val.get_full_data()
else:
temp_val = val
return_val = None
for slice_list in utilities.get_slice_list(temp_val.shape, axes):
if slice_list == [slice(None, None)]:
inp = temp_val
else:
if return_val is None:
return_val = np.empty_like(temp_val)
inp = temp_val[slice_list]
if self.domain.dtype == np.dtype('float32'):
inp = gl.map2alm_f(inp,
nlat=nlat, nlon=nlon,
lmax=lmax, mmax=mmax)
else:
inp = gl.map2alm(inp,
nlat=nlat, nlon=nlon,
lmax=lmax, mmax=mmax)
if slice_list == [slice(None, None)]:
return_val = inp
else:
return_val[slice_list] = inp
if isinstance(val, distributed_data_object):
new_val = val.copy_empty(dtype=self.codomain.dtype)
new_val.set_full_data(return_val, copy=False)
if issubclass(inp.dtype.type, np.complexfloating):
[resultReal, resultImag] = [self.libsharpMap2Alm(x,
nlat=nlat,
nlon=nlon,
lmax=lmax,
mmax=mmax,
**kwargs)
for x in (inp.real, inp.imag)]
[resultReal, resultImag] = [ltf.buildIdx(x, lmax=lmax)
for x in [resultReal, resultImag]]
result = self._combine_complex_result(resultReal, resultImag)
else:
return_val = return_val.astype(self.codomain.dtype, copy=False)
result = self.libsharpMap2Alm(inp, nlat=nlat, nlon=nlon, lmax=lmax,
mmax=mmax)
result = ltf.buildIdx(result, lmax=lmax)
return result
return return_val
# ---Added properties and methods---
def libsharpMap2Alm(self, inp, **kwargs):
if inp.dtype == np.dtype('float32'):
return libsharp.map2alm_f(inp, **kwargs)
elif inp.dtype == np.dtype('float64'):
return libsharp.map2alm(inp, **kwargs)
else:
about.warnings.cprint("WARNING: performing dtype conversion for "
"libsharp compatibility.")
casted_inp = inp.astype(np.dtype('float64'), copy=False)
result = libsharp.map2alm(casted_inp, **kwargs)
return result
\ No newline at end of file
nifty/operators/fft_operator/transformations/hplmtransformation.py
View file @ 3454501e
import numpy as np
from transformation import Transformation
from d2o import distributed_data_object
from nifty.config import dependency_injector as gdi
import nifty.nifty_utilities as utilities
from nifty.config import dependency_injector as gdi,\
about
from nifty import HPSpace, LMSpace
from slicing_transformation import SlicingTransformation
import lm_transformation_factory as ltf
hp = gdi.get('healpy')
class HPLMTransformation(Transformation):
class HPLMTransformation(SlicingTransformation):
# ---Overwritten properties and methods---
def __init__(self, domain, codomain=None, module=None):
if 'healpy' not in gdi:
raise ImportError("The module healpy is needed but not available")
if codomain is None:
self.domain = domain
self.codomain = self.get_codomain(domain)
elif self.check_codomain(domain, codomain):
self.domain = domain
self.codomain = codomain
else:
raise ValueError("ERROR: Incompatible codomain!")
raise ImportError(about._errors.cstring(
"The module healpy is needed but not available"))
@staticmethod
def get_codomain(domain):
super(HPLMTransformation, self).__init__(domain, codomain, module)
# ---Mandatory properties and methods---
@classmethod
def get_codomain(cls, domain):
"""
Generates a compatible codomain to which transformations are
reasonable, i.e.\ an instance of the :py:class:`lm_space` class.
......@@ -38,87 +38,65 @@ class HPLMTransformation(Transformation):
codomain : LMSpace
A compatible codomain.
"""
if domain is None:
raise ValueError('ERROR: cannot generate codomain for None')
if not isinstance(domain, HPSpace):
raise TypeError('ERROR: domain needs to be a HPSpace')
raise TypeError(about._errors.cstring(
"ERROR: domain needs to be a HPSpace"))