1 | """ |
---|
2 | Handle Q smearing |
---|
3 | """ |
---|
4 | ##################################################################### |
---|
5 | #This software was developed by the University of Tennessee as part of the |
---|
6 | #Distributed Data Analysis of Neutron Scattering Experiments (DANSE) |
---|
7 | #project funded by the US National Science Foundation. |
---|
8 | #See the license text in license.txt |
---|
9 | #copyright 2008, University of Tennessee |
---|
10 | ###################################################################### |
---|
11 | import numpy |
---|
12 | import math |
---|
13 | import logging |
---|
14 | import sys |
---|
15 | import numpy as np # type: ignore |
---|
16 | from numpy import pi, exp # type:ignore |
---|
17 | from sasmodels.resolution import Slit1D, Pinhole1D |
---|
18 | from sasmodels.sesans import SesansTransform |
---|
19 | from sasmodels.resolution2d import Pinhole2D |
---|
20 | from src.sas.sascalc.data_util.nxsunit import Converter |
---|
21 | |
---|
22 | def smear_selection(data, model = None): |
---|
23 | """ |
---|
24 | Creates the right type of smearer according |
---|
25 | to the data. |
---|
26 | The canSAS format has a rule that either |
---|
27 | slit smearing data OR resolution smearing data |
---|
28 | is available. |
---|
29 | |
---|
30 | For the present purpose, we choose the one that |
---|
31 | has none-zero data. If both slit and resolution |
---|
32 | smearing arrays are filled with good data |
---|
33 | (which should not happen), then we choose the |
---|
34 | resolution smearing data. |
---|
35 | |
---|
36 | :param data: Data1D object |
---|
37 | :param model: sas.model instance |
---|
38 | """ |
---|
39 | # Sanity check. If we are not dealing with a SAS Data1D |
---|
40 | # object, just return None |
---|
41 | # This checks for 2D data (does not throw exception because fail is common) |
---|
42 | if data.__class__.__name__ not in ['Data1D', 'Theory1D']: |
---|
43 | if data == None: |
---|
44 | return None |
---|
45 | elif data.dqx_data == None or data.dqy_data == None: |
---|
46 | return None |
---|
47 | return PySmear2D(data) |
---|
48 | # This checks for 1D data with smearing info in the data itself (again, fail is likely; no exceptions) |
---|
49 | if not hasattr(data, "dx") and not hasattr(data, "dxl")\ |
---|
50 | and not hasattr(data, "dxw"): |
---|
51 | return None |
---|
52 | |
---|
53 | # Look for resolution smearing data |
---|
54 | # This is the code that checks for SESANS data; it looks for the file loader |
---|
55 | # TODO: change other sanity checks to check for file loader instead of data structure? |
---|
56 | _found_sesans = False |
---|
57 | #if data.dx is not None and data.meta_data['loader']=='SESANS': |
---|
58 | if data.dx is not None and data.isSesans: |
---|
59 | #if data.dx[0] > 0.0: |
---|
60 | if numpy.size(data.dx[data.dx <= 0]) == 0: |
---|
61 | _found_sesans = True |
---|
62 | # if data.dx[0] <= 0.0: |
---|
63 | if numpy.size(data.dx[data.dx <= 0]) > 0: |
---|
64 | raise ValueError('one or more of your dx values are negative, please check the data file!') |
---|
65 | |
---|
66 | if _found_sesans == True: |
---|
67 | #Pre-compute the Hankel matrix (H) |
---|
68 | qmax, qunits = data.sample.zacceptance |
---|
69 | SElength = Converter(data._xunit)(data.x, "A") |
---|
70 | zaccept = Converter(qunits)(qmax, "1/A"), |
---|
71 | Rmax = 10000000 |
---|
72 | hankel = SesansTransform(data.x, SElength, zaccept, Rmax) |
---|
73 | # Then return the actual transform, as if it were a smearing function |
---|
74 | return PySmear(hankel, model, offset=0) |
---|
75 | |
---|
76 | _found_resolution = False |
---|
77 | if data.dx is not None and len(data.dx) == len(data.x): |
---|
78 | |
---|
79 | # Check that we have non-zero data |
---|
80 | if data.dx[0] > 0.0: |
---|
81 | _found_resolution = True |
---|
82 | #print "_found_resolution",_found_resolution |
---|
83 | #print "data1D.dx[0]",data1D.dx[0],data1D.dxl[0] |
---|
84 | # If we found resolution smearing data, return a QSmearer |
---|
85 | if _found_resolution == True: |
---|
86 | return pinhole_smear(data, model) |
---|
87 | |
---|
88 | # Look for slit smearing data |
---|
89 | _found_slit = False |
---|
90 | if data.dxl is not None and len(data.dxl) == len(data.x) \ |
---|
91 | and data.dxw is not None and len(data.dxw) == len(data.x): |
---|
92 | |
---|
93 | # Check that we have non-zero data |
---|
94 | if data.dxl[0] > 0.0 or data.dxw[0] > 0.0: |
---|
95 | _found_slit = True |
---|
96 | |
---|
97 | # Sanity check: all data should be the same as a function of Q |
---|
98 | for item in data.dxl: |
---|
99 | if data.dxl[0] != item: |
---|
100 | _found_resolution = False |
---|
101 | break |
---|
102 | |
---|
103 | for item in data.dxw: |
---|
104 | if data.dxw[0] != item: |
---|
105 | _found_resolution = False |
---|
106 | break |
---|
107 | # If we found slit smearing data, return a slit smearer |
---|
108 | if _found_slit == True: |
---|
109 | return slit_smear(data, model) |
---|
110 | return None |
---|
111 | |
---|
112 | |
---|
113 | class PySmear(object): |
---|
114 | """ |
---|
115 | Wrapper for pure python sasmodels resolution functions. |
---|
116 | """ |
---|
117 | def __init__(self, resolution, model, offset=None): |
---|
118 | self.model = model |
---|
119 | self.resolution = resolution |
---|
120 | if offset is None: |
---|
121 | offset = numpy.searchsorted(self.resolution.q_calc, self.resolution.q[0]) |
---|
122 | self.offset = offset |
---|
123 | |
---|
124 | def apply(self, iq_in, first_bin=0, last_bin=None): |
---|
125 | """ |
---|
126 | Apply the resolution function to the data. |
---|
127 | Note that this is called with iq_in matching data.x, but with |
---|
128 | iq_in[first_bin:last_bin] set to theory values for these bins, |
---|
129 | and the remainder left undefined. The first_bin, last_bin values |
---|
130 | should be those returned from get_bin_range. |
---|
131 | The returned value is of the same length as iq_in, with the range |
---|
132 | first_bin:last_bin set to the resolution smeared values. |
---|
133 | """ |
---|
134 | if last_bin is None: last_bin = len(iq_in) |
---|
135 | start, end = first_bin + self.offset, last_bin + self.offset |
---|
136 | q_calc = self.resolution.q_calc |
---|
137 | iq_calc = numpy.empty_like(q_calc) |
---|
138 | if start > 0: |
---|
139 | iq_calc[:start] = self.model.evalDistribution(q_calc[:start]) |
---|
140 | if end+1 < len(q_calc): |
---|
141 | iq_calc[end+1:] = self.model.evalDistribution(q_calc[end+1:]) |
---|
142 | iq_calc[start:end+1] = iq_in[first_bin:last_bin+1] |
---|
143 | smeared = self.resolution.apply(iq_calc) |
---|
144 | return smeared |
---|
145 | __call__ = apply |
---|
146 | |
---|
147 | def get_bin_range(self, q_min=None, q_max=None): |
---|
148 | """ |
---|
149 | For a given q_min, q_max, find the corresponding indices in the data. |
---|
150 | Returns first, last. |
---|
151 | Note that these are indexes into q from the data, not the q_calc |
---|
152 | needed by the resolution function. Note also that these are the |
---|
153 | indices, not the range limits. That is, the complete range will be |
---|
154 | q[first:last+1]. |
---|
155 | """ |
---|
156 | q = self.resolution.q |
---|
157 | first = numpy.searchsorted(q, q_min) |
---|
158 | last = numpy.searchsorted(q, q_max) |
---|
159 | return first, min(last,len(q)-1) |
---|
160 | |
---|
161 | def slit_smear(data, model=None): |
---|
162 | q = data.x |
---|
163 | width = data.dxw if data.dxw is not None else 0 |
---|
164 | height = data.dxl if data.dxl is not None else 0 |
---|
165 | # TODO: width and height seem to be reversed |
---|
166 | return PySmear(Slit1D(q, height, width), model) |
---|
167 | |
---|
168 | def pinhole_smear(data, model=None): |
---|
169 | q = data.x |
---|
170 | width = data.dx if data.dx is not None else 0 |
---|
171 | return PySmear(Pinhole1D(q, width), model) |
---|
172 | |
---|
173 | |
---|
174 | class PySmear2D(object): |
---|
175 | """ |
---|
176 | Q smearing class for SAS 2d pinhole data |
---|
177 | """ |
---|
178 | |
---|
179 | def __init__(self, data=None, model=None): |
---|
180 | self.data = data |
---|
181 | self.model = model |
---|
182 | self.accuracy = 'Low' |
---|
183 | self.limit = 3.0 |
---|
184 | self.index = None |
---|
185 | self.coords = 'polar' |
---|
186 | self.smearer = True |
---|
187 | |
---|
188 | def set_accuracy(self, accuracy='Low'): |
---|
189 | """ |
---|
190 | Set accuracy. |
---|
191 | |
---|
192 | :param accuracy: string |
---|
193 | """ |
---|
194 | self.accuracy = accuracy |
---|
195 | |
---|
196 | def set_smearer(self, smearer=True): |
---|
197 | """ |
---|
198 | Set whether or not smearer will be used |
---|
199 | |
---|
200 | :param smearer: smear object |
---|
201 | |
---|
202 | """ |
---|
203 | self.smearer = smearer |
---|
204 | |
---|
205 | def set_data(self, data=None): |
---|
206 | """ |
---|
207 | Set data. |
---|
208 | |
---|
209 | :param data: DataLoader.Data_info type |
---|
210 | """ |
---|
211 | self.data = data |
---|
212 | |
---|
213 | def set_model(self, model=None): |
---|
214 | """ |
---|
215 | Set model. |
---|
216 | |
---|
217 | :param model: sas.models instance |
---|
218 | """ |
---|
219 | self.model = model |
---|
220 | |
---|
221 | def set_index(self, index=None): |
---|
222 | """ |
---|
223 | Set index. |
---|
224 | |
---|
225 | :param index: 1d arrays |
---|
226 | """ |
---|
227 | self.index = index |
---|
228 | |
---|
229 | def get_value(self): |
---|
230 | """ |
---|
231 | Over sampling of r_nbins times phi_nbins, calculate Gaussian weights, |
---|
232 | then find smeared intensity |
---|
233 | """ |
---|
234 | if self.smearer: |
---|
235 | res = Pinhole2D(data=self.data, index=self.index, |
---|
236 | nsigma=3.0, accuracy=self.accuracy, |
---|
237 | coords=self.coords) |
---|
238 | val = self.model.evalDistribution(res.q_calc) |
---|
239 | return res.apply(val) |
---|
240 | else: |
---|
241 | index = self.index if self.index is not None else slice(None) |
---|
242 | qx_data = self.data.qx_data[index] |
---|
243 | qy_data = self.data.qy_data[index] |
---|
244 | q_calc = [qx_data, qy_data] |
---|
245 | val = self.model.evalDistribution(q_calc) |
---|
246 | return val |
---|
247 | |
---|