gl_space.py 6.48 KB
Newer Older
csongor's avatar
csongor committed
1
2
from __future__ import division

Jait Dixit's avatar
Jait Dixit committed
3
import itertools
csongor's avatar
csongor committed
4
5
import numpy as np

6
7
import d2o
from d2o import STRATEGIES as DISTRIBUTION_STRATEGIES
csongor's avatar
csongor committed
8

9
from nifty.spaces.space import Space
10
from nifty.config import nifty_configuration as gc,\
11
                         dependency_injector as gdi
csongor's avatar
csongor committed
12
13
14
15
16

gl = gdi.get('libsharp_wrapper_gl')

GL_DISTRIBUTION_STRATEGIES = DISTRIBUTION_STRATEGIES['global']

17

Theo Steininger's avatar
Theo Steininger committed
18
class GLSpace(Space):
csongor's avatar
csongor committed
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
    """
        ..                 __
        ..               /  /
        ..     ____ __  /  /
        ..   /   _   / /  /
        ..  /  /_/  / /  /_
        ..  \___   /  \___/  space class
        .. /______/

        NIFTY subclass for Gauss-Legendre pixelizations [#]_ of the two-sphere.

        Parameters
        ----------
        nlat : int
            Number of latitudinal bins, or rings.
        nlon : int, *optional*
            Number of longitudinal bins (default: ``2*nlat - 1``).
        dtype : numpy.dtype, *optional*
            Data type of the field values (default: numpy.float64).

        See Also
        --------
        hp_space : A class for the HEALPix discretization of the sphere [#]_.
        lm_space : A class for spherical harmonic components.

        Notes
        -----
        Only real-valued fields on the two-sphere are supported, i.e.
        `dtype` has to be either numpy.float64 or numpy.float32.

        References
        ----------
        .. [#] M. Reinecke and D. Sverre Seljebotn, 2013, "Libsharp - spherical
               harmonic transforms revisited";
               `arXiv:1303.4945 <http://www.arxiv.org/abs/1303.4945>`_
        .. [#] K.M. Gorski et al., 2005, "HEALPix: A Framework for
               High-Resolution Discretization and Fast Analysis of Data
               Distributed on the Sphere", *ApJ* 622..759G.

        Attributes
        ----------
        para : numpy.ndarray
            One-dimensional array containing the two numbers `nlat` and `nlon`.
        dtype : numpy.dtype
            Data type of the field values.
        discrete : bool
            Whether or not the underlying space is discrete, always ``False``
            for spherical spaces.
        vol : numpy.ndarray
            An array containing the pixel sizes.
    """

71
72
    # ---Overwritten properties and methods---

73
    def __init__(self, nlat=2, nlon=None, dtype=None):
csongor's avatar
csongor committed
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
        """
            Sets the attributes for a gl_space class instance.

            Parameters
            ----------
            nlat : int
                Number of latitudinal bins, or rings.
            nlon : int, *optional*
                Number of longitudinal bins (default: ``2*nlat - 1``).
            dtype : numpy.dtype, *optional*
                Data type of the field values (default: numpy.float64).

            Returns
            -------
            None

            Raises
            ------
            ImportError
                If the libsharp_wrapper_gl module is not available.
            ValueError
                If input `nlat` is invaild.

        """
        # check imports
        if not gc['use_libsharp']:
100
101
            raise ImportError(
                "libsharp_wrapper_gl not available or not loaded.")
102
103

        super(GLSpace, self).__init__(dtype)
csongor's avatar
csongor committed
104

105
106
        self._nlat = self._parse_nlat(nlat)
        self._nlon = self._parse_nlon(nlon)
csongor's avatar
csongor committed
107

108
    # ---Mandatory properties and methods---
csongor's avatar
csongor committed
109

110
111
112
    @property
    def harmonic(self):
        return False
csongor's avatar
csongor committed
113
114
115

    @property
    def shape(self):
116
        return (np.int((self.nlat * self.nlon)),)
csongor's avatar
csongor committed
117

118
    @property
119
    def dim(self):
120
        return np.int((self.nlat * self.nlon))
121
122
123
124

    @property
    def total_volume(self):
        return 4 * np.pi
125

126
127
128
129
130
    def copy(self):
        return self.__class__(nlat=self.nlat,
                              nlon=self.nlon,
                              dtype=self.dtype)

Jait Dixit's avatar
Jait Dixit committed
131
    def weight(self, x, power=1, axes=None, inplace=False):
132
133
        nlon = self.nlon
        nlat = self.nlat
Theo Steininger's avatar
Theo Steininger committed
134
        vol = gl.vol(nlat) ** power
135
        weight = np.array(list(itertools.chain.from_iterable(
Theo Steininger's avatar
Theo Steininger committed
136
                          itertools.repeat(x, nlon) for x in vol)))
Jait Dixit's avatar
Jait Dixit committed
137
138
139

        if axes is not None:
            # reshape the weight array to match the input shape
140
            new_shape = np.ones(len(x.shape), dtype=np.int)
141
142
            # we know len(axes) is always 1
            new_shape[axes[0]] = len(weight)
Jait Dixit's avatar
Jait Dixit committed
143
144
145
146
147
            weight = weight.reshape(new_shape)

        if inplace:
            x *= weight
            result_x = x
csongor's avatar
csongor committed
148
        else:
Jait Dixit's avatar
Jait Dixit committed
149
            result_x = x * weight
csongor's avatar
csongor committed
150

Jait Dixit's avatar
Jait Dixit committed
151
        return result_x
152

153
    def get_distance_array(self, distribution_strategy):
154
155
        dists = d2o.arange(start=0, stop=self.shape[0],
                           distribution_strategy=distribution_strategy)
156

157
        dists = dists.apply_scalar_function(
158
            lambda x: self._distance_array_helper(divmod(x, self.nlon)),
159
            dtype=np.float)
160
161
162

        return dists

theos's avatar
theos committed
163
    def _distance_array_helper(self, qr_tuple):
164
165
        lat = qr_tuple[0]*(np.pi/(self.nlat-1))
        lon = qr_tuple[1]*(2*np.pi/(self.nlon-1))
Jait Dixit's avatar
Jait Dixit committed
166
167
168
        numerator = np.sqrt(np.sin(lon)**2 +
                            (np.sin(lat) * np.cos(lon))**2)
        denominator = np.cos(lat) * np.cos(lon)
169

theos's avatar
theos committed
170
        return np.arctan(numerator / denominator)
171

172
    def get_fft_smoothing_kernel_function(self, sigma):
Jait Dixit's avatar
Jait Dixit committed
173
        if sigma is None:
174
            sigma = np.sqrt(2) * np.pi
Jait Dixit's avatar
Jait Dixit committed
175
176

        return lambda x: np.exp((-0.5 * x**2) / sigma**2)
177

178
179
180
181
182
183
184
185
186
187
188
189
190
    # ---Added properties and methods---

    @property
    def nlat(self):
        return self._nlat

    @property
    def nlon(self):
        return self._nlon

    def _parse_nlat(self, nlat):
        nlat = int(nlat)
        if nlat < 2:
191
            raise ValueError("nlat must be greater than 2.")
192
        elif nlat % 2 != 0:
193
            raise ValueError("nlat must be a multiple of 2.")
194
195
196
197
198
199
200
201
        return nlat

    def _parse_nlon(self, nlon):
        if nlon is None:
            nlon = 2 * self.nlat - 1
        else:
            nlon = int(nlon)
            if nlon != 2 * self.nlat - 1:
202
203
                self.logger.warn("nlon was set to an unrecommended value: "
                                 "nlon <> 2*nlat-1.")
204
        return nlon
205
206
207
208

    # ---Serialization---

    def _to_hdf5(self, hdf5_group):
Jait Dixit's avatar
Jait Dixit committed
209
210
        hdf5_group['nlat'] = self.nlat
        hdf5_group['nlon'] = self.nlon
211
        hdf5_group.attrs['dtype'] = self.dtype.name
Jait Dixit's avatar
Jait Dixit committed
212

213
214
215
        return None

    @classmethod
Theo Steininger's avatar
Theo Steininger committed
216
    def _from_hdf5(cls, hdf5_group, repository):
217
        result = cls(
Jait Dixit's avatar
Jait Dixit committed
218
219
            nlat=hdf5_group['nlat'][()],
            nlon=hdf5_group['nlon'][()],
220
            dtype=np.dtype(hdf5_group.attrs['dtype'])
Jait Dixit's avatar
Jait Dixit committed
221
222
            )

223
        return result