[1846] | 1 | """This module defines the scantable class.""" |
---|
| 2 | |
---|
[1697] | 3 | import os |
---|
[2751] | 4 | import re |
---|
[2315] | 5 | import tempfile |
---|
[1948] | 6 | import numpy |
---|
[1691] | 7 | try: |
---|
| 8 | from functools import wraps as wraps_dec |
---|
| 9 | except ImportError: |
---|
| 10 | from asap.compatibility import wraps as wraps_dec |
---|
| 11 | |
---|
[1824] | 12 | from asap.env import is_casapy |
---|
[876] | 13 | from asap._asap import Scantable |
---|
[2004] | 14 | from asap._asap import filler, msfiller |
---|
[1824] | 15 | from asap.parameters import rcParams |
---|
[1862] | 16 | from asap.logging import asaplog, asaplog_post_dec |
---|
[1824] | 17 | from asap.selector import selector |
---|
| 18 | from asap.linecatalog import linecatalog |
---|
[1600] | 19 | from asap.coordinate import coordinate |
---|
[1859] | 20 | from asap.utils import _n_bools, mask_not, mask_and, mask_or, page |
---|
[1907] | 21 | from asap.asapfitter import fitter |
---|
[102] | 22 | |
---|
[1689] | 23 | def preserve_selection(func): |
---|
[1691] | 24 | @wraps_dec(func) |
---|
[1689] | 25 | def wrap(obj, *args, **kw): |
---|
| 26 | basesel = obj.get_selection() |
---|
[1857] | 27 | try: |
---|
| 28 | val = func(obj, *args, **kw) |
---|
| 29 | finally: |
---|
| 30 | obj.set_selection(basesel) |
---|
[1689] | 31 | return val |
---|
| 32 | return wrap |
---|
| 33 | |
---|
[1846] | 34 | def is_scantable(filename): |
---|
| 35 | """Is the given file a scantable? |
---|
[1689] | 36 | |
---|
[1846] | 37 | Parameters: |
---|
| 38 | |
---|
| 39 | filename: the name of the file/directory to test |
---|
| 40 | |
---|
| 41 | """ |
---|
[1883] | 42 | if ( os.path.isdir(filename) |
---|
| 43 | and os.path.exists(filename+'/table.info') |
---|
| 44 | and os.path.exists(filename+'/table.dat') ): |
---|
| 45 | f=open(filename+'/table.info') |
---|
| 46 | l=f.readline() |
---|
| 47 | f.close() |
---|
[2753] | 48 | match_pattern = '^Type = (Scantable)? *$' |
---|
[2751] | 49 | if re.match(match_pattern,l): |
---|
[1883] | 50 | return True |
---|
| 51 | else: |
---|
| 52 | return False |
---|
| 53 | else: |
---|
| 54 | return False |
---|
| 55 | ## return (os.path.isdir(filename) |
---|
| 56 | ## and not os.path.exists(filename+'/table.f1') |
---|
| 57 | ## and os.path.exists(filename+'/table.info')) |
---|
[1697] | 58 | |
---|
[1883] | 59 | def is_ms(filename): |
---|
| 60 | """Is the given file a MeasurementSet? |
---|
[1697] | 61 | |
---|
[1883] | 62 | Parameters: |
---|
| 63 | |
---|
| 64 | filename: the name of the file/directory to test |
---|
| 65 | |
---|
| 66 | """ |
---|
| 67 | if ( os.path.isdir(filename) |
---|
| 68 | and os.path.exists(filename+'/table.info') |
---|
| 69 | and os.path.exists(filename+'/table.dat') ): |
---|
| 70 | f=open(filename+'/table.info') |
---|
| 71 | l=f.readline() |
---|
| 72 | f.close() |
---|
| 73 | if ( l.find('Measurement Set') != -1 ): |
---|
| 74 | return True |
---|
| 75 | else: |
---|
| 76 | return False |
---|
| 77 | else: |
---|
| 78 | return False |
---|
[2186] | 79 | |
---|
| 80 | def normalise_edge_param(edge): |
---|
| 81 | """\ |
---|
| 82 | Convert a given edge value to a one-dimensional array that can be |
---|
| 83 | given to baseline-fitting/subtraction functions. |
---|
| 84 | The length of the output value will be an even because values for |
---|
| 85 | the both sides of spectra are to be contained for each IF. When |
---|
| 86 | the length is 2, the values will be applied to all IFs. If the length |
---|
| 87 | is larger than 2, it will be 2*ifnos(). |
---|
| 88 | Accepted format of edge include: |
---|
| 89 | * an integer - will be used for both sides of spectra of all IFs. |
---|
| 90 | e.g. 10 is converted to [10,10] |
---|
[2277] | 91 | * an empty list/tuple [] - converted to [0, 0] and used for all IFs. |
---|
[2186] | 92 | * a list/tuple containing an integer - same as the above case. |
---|
| 93 | e.g. [10] is converted to [10,10] |
---|
| 94 | * a list/tuple containing two integers - will be used for all IFs. |
---|
| 95 | e.g. [5,10] is output as it is. no need to convert. |
---|
| 96 | * a list/tuple of lists/tuples containing TWO integers - |
---|
| 97 | each element of edge will be used for each IF. |
---|
[2277] | 98 | e.g. [[5,10],[15,20]] - [5,10] for IF[0] and [15,20] for IF[1]. |
---|
| 99 | |
---|
| 100 | If an element contains the same integer values, the input 'edge' |
---|
| 101 | parameter can be given in a simpler shape in the following cases: |
---|
[2186] | 102 | ** when len(edge)!=2 |
---|
[2277] | 103 | any elements containing the same values can be replaced |
---|
| 104 | to single integers. |
---|
| 105 | e.g. [[15,15]] can be simplified to [15] (or [15,15] or 15 also). |
---|
| 106 | e.g. [[1,1],[2,2],[3,3]] can be simplified to [1,2,3]. |
---|
[2186] | 107 | ** when len(edge)=2 |
---|
| 108 | care is needed for this case: ONLY ONE of the |
---|
| 109 | elements can be a single integer, |
---|
| 110 | e.g. [[5,5],[10,10]] can be simplified to [5,[10,10]] |
---|
[2277] | 111 | or [[5,5],10], but can NOT be simplified to [5,10]. |
---|
[2186] | 112 | when [5,10] given, it is interpreted as |
---|
[2277] | 113 | [[5,10],[5,10],[5,10],...] instead, as shown before. |
---|
[2186] | 114 | """ |
---|
| 115 | from asap import _is_sequence_or_number as _is_valid |
---|
| 116 | if isinstance(edge, list) or isinstance(edge, tuple): |
---|
| 117 | for edgepar in edge: |
---|
| 118 | if not _is_valid(edgepar, int): |
---|
| 119 | raise ValueError, "Each element of the 'edge' tuple has \ |
---|
| 120 | to be a pair of integers or an integer." |
---|
| 121 | if isinstance(edgepar, list) or isinstance(edgepar, tuple): |
---|
| 122 | if len(edgepar) != 2: |
---|
| 123 | raise ValueError, "Each element of the 'edge' tuple has \ |
---|
| 124 | to be a pair of integers or an integer." |
---|
| 125 | else: |
---|
| 126 | if not _is_valid(edge, int): |
---|
| 127 | raise ValueError, "Parameter 'edge' has to be an integer or a \ |
---|
| 128 | pair of integers specified as a tuple. \ |
---|
| 129 | Nested tuples are allowed \ |
---|
| 130 | to make individual selection for different IFs." |
---|
| 131 | |
---|
| 132 | |
---|
| 133 | if isinstance(edge, int): |
---|
| 134 | edge = [ edge, edge ] # e.g. 3 => [3,3] |
---|
| 135 | elif isinstance(edge, list) or isinstance(edge, tuple): |
---|
| 136 | if len(edge) == 0: |
---|
| 137 | edge = [0, 0] # e.g. [] => [0,0] |
---|
| 138 | elif len(edge) == 1: |
---|
| 139 | if isinstance(edge[0], int): |
---|
| 140 | edge = [ edge[0], edge[0] ] # e.g. [1] => [1,1] |
---|
| 141 | |
---|
| 142 | commonedge = True |
---|
| 143 | if len(edge) > 2: commonedge = False |
---|
| 144 | else: |
---|
| 145 | for edgepar in edge: |
---|
| 146 | if isinstance(edgepar, list) or isinstance(edgepar, tuple): |
---|
| 147 | commonedge = False |
---|
| 148 | break |
---|
| 149 | |
---|
| 150 | if commonedge: |
---|
| 151 | if len(edge) > 1: |
---|
| 152 | norm_edge = edge |
---|
| 153 | else: |
---|
| 154 | norm_edge = edge + edge |
---|
| 155 | else: |
---|
| 156 | norm_edge = [] |
---|
| 157 | for edgepar in edge: |
---|
| 158 | if isinstance(edgepar, int): |
---|
| 159 | norm_edge += [edgepar, edgepar] |
---|
| 160 | else: |
---|
| 161 | norm_edge += edgepar |
---|
| 162 | |
---|
| 163 | return norm_edge |
---|
| 164 | |
---|
| 165 | def raise_fitting_failure_exception(e): |
---|
| 166 | msg = "The fit failed, possibly because it didn't converge." |
---|
| 167 | if rcParams["verbose"]: |
---|
| 168 | asaplog.push(str(e)) |
---|
| 169 | asaplog.push(str(msg)) |
---|
| 170 | else: |
---|
| 171 | raise RuntimeError(str(e)+'\n'+msg) |
---|
| 172 | |
---|
[2189] | 173 | def pack_progress_params(showprogress, minnrow): |
---|
| 174 | return str(showprogress).lower() + ',' + str(minnrow) |
---|
| 175 | |
---|
[2767] | 176 | def pack_blinfo(blinfo, maxirow): |
---|
| 177 | """\ |
---|
| 178 | convert a dictionary or a list of dictionaries of baseline info |
---|
| 179 | into a list of comma-separated strings. |
---|
| 180 | """ |
---|
| 181 | if isinstance(blinfo, dict): |
---|
| 182 | res = do_pack_blinfo(blinfo, maxirow) |
---|
| 183 | return [res] if res != '' else [] |
---|
| 184 | elif isinstance(blinfo, list) or isinstance(blinfo, tuple): |
---|
| 185 | res = [] |
---|
| 186 | for i in xrange(len(blinfo)): |
---|
| 187 | resi = do_pack_blinfo(blinfo[i], maxirow) |
---|
| 188 | if resi != '': |
---|
| 189 | res.append(resi) |
---|
| 190 | return res |
---|
| 191 | |
---|
| 192 | def do_pack_blinfo(blinfo, maxirow): |
---|
| 193 | """\ |
---|
| 194 | convert a dictionary of baseline info for a spectrum into |
---|
| 195 | a comma-separated string. |
---|
| 196 | """ |
---|
| 197 | dinfo = {} |
---|
| 198 | for key in ['row', 'blfunc', 'masklist']: |
---|
| 199 | if blinfo.has_key(key): |
---|
| 200 | val = blinfo[key] |
---|
| 201 | if key == 'row': |
---|
| 202 | irow = val |
---|
| 203 | if isinstance(val, list) or isinstance(val, tuple): |
---|
| 204 | slval = [] |
---|
| 205 | for i in xrange(len(val)): |
---|
| 206 | if isinstance(val[i], list) or isinstance(val[i], tuple): |
---|
| 207 | for j in xrange(len(val[i])): |
---|
| 208 | slval.append(str(val[i][j])) |
---|
| 209 | else: |
---|
| 210 | slval.append(str(val[i])) |
---|
| 211 | sval = ",".join(slval) |
---|
| 212 | else: |
---|
| 213 | sval = str(val) |
---|
| 214 | |
---|
| 215 | dinfo[key] = sval |
---|
| 216 | else: |
---|
| 217 | raise ValueError("'"+key+"' is missing in blinfo.") |
---|
| 218 | |
---|
| 219 | if irow >= maxirow: return '' |
---|
| 220 | |
---|
| 221 | for key in ['order', 'npiece', 'nwave']: |
---|
| 222 | if blinfo.has_key(key): |
---|
| 223 | val = blinfo[key] |
---|
| 224 | if isinstance(val, list) or isinstance(val, tuple): |
---|
| 225 | slval = [] |
---|
| 226 | for i in xrange(len(val)): |
---|
| 227 | slval.append(str(val[i])) |
---|
| 228 | sval = ",".join(slval) |
---|
| 229 | else: |
---|
| 230 | sval = str(val) |
---|
| 231 | dinfo[key] = sval |
---|
| 232 | |
---|
| 233 | blfunc = dinfo['blfunc'] |
---|
| 234 | fspec_keys = {'poly': 'order', 'chebyshev': 'order', 'cspline': 'npiece', 'sinusoid': 'nwave'} |
---|
| 235 | |
---|
| 236 | fspec_key = fspec_keys[blfunc] |
---|
| 237 | if not blinfo.has_key(fspec_key): |
---|
| 238 | raise ValueError("'"+fspec_key+"' is missing in blinfo.") |
---|
| 239 | |
---|
| 240 | clip_params_n = 0 |
---|
| 241 | for key in ['clipthresh', 'clipniter']: |
---|
| 242 | if blinfo.has_key(key): |
---|
| 243 | clip_params_n += 1 |
---|
| 244 | dinfo[key] = str(blinfo[key]) |
---|
| 245 | |
---|
| 246 | if clip_params_n == 0: |
---|
| 247 | dinfo['clipthresh'] = '0.0' |
---|
| 248 | dinfo['clipniter'] = '0' |
---|
| 249 | elif clip_params_n != 2: |
---|
| 250 | raise ValueError("both 'clipthresh' and 'clipniter' must be given for n-sigma clipping.") |
---|
| 251 | |
---|
| 252 | lf_params_n = 0 |
---|
| 253 | for key in ['thresh', 'edge', 'chan_avg_limit']: |
---|
| 254 | if blinfo.has_key(key): |
---|
| 255 | lf_params_n += 1 |
---|
| 256 | val = blinfo[key] |
---|
| 257 | if isinstance(val, list) or isinstance(val, tuple): |
---|
| 258 | slval = [] |
---|
| 259 | for i in xrange(len(val)): |
---|
| 260 | slval.append(str(val[i])) |
---|
| 261 | sval = ",".join(slval) |
---|
| 262 | else: |
---|
| 263 | sval = str(val) |
---|
| 264 | dinfo[key] = sval |
---|
| 265 | |
---|
| 266 | if lf_params_n == 3: |
---|
| 267 | dinfo['use_linefinder'] = 'true' |
---|
[2810] | 268 | elif lf_params_n == 0: |
---|
[2767] | 269 | dinfo['use_linefinder'] = 'false' |
---|
| 270 | dinfo['thresh'] = '' |
---|
| 271 | dinfo['edge'] = '' |
---|
| 272 | dinfo['chan_avg_limit'] = '' |
---|
| 273 | else: |
---|
| 274 | raise ValueError("all of 'thresh', 'edge' and 'chan_avg_limit' must be given to use linefinder.") |
---|
| 275 | |
---|
| 276 | slblinfo = [dinfo['row'], blfunc, dinfo[fspec_key], dinfo['masklist'], \ |
---|
| 277 | dinfo['clipthresh'], dinfo['clipniter'], \ |
---|
| 278 | dinfo['use_linefinder'], dinfo['thresh'], dinfo['edge'], dinfo['chan_avg_limit']] |
---|
| 279 | |
---|
| 280 | return ":".join(slblinfo) |
---|
| 281 | |
---|
| 282 | def parse_fitresult(sres): |
---|
| 283 | """\ |
---|
| 284 | Parse the returned value of apply_bltable() or sub_baseline() and |
---|
| 285 | extract row number, the best-fit coefficients and rms, then pack |
---|
| 286 | them into a dictionary. |
---|
| 287 | The input value is generated by Scantable::packFittingResults() and |
---|
| 288 | formatted as 'row:coeff[0],coeff[1],..,coeff[n-1]:rms'. |
---|
| 289 | """ |
---|
| 290 | res = [] |
---|
| 291 | for i in xrange(len(sres)): |
---|
| 292 | (srow, scoeff, srms) = sres[i].split(":") |
---|
| 293 | row = int(srow) |
---|
| 294 | rms = float(srms) |
---|
| 295 | lscoeff = scoeff.split(",") |
---|
| 296 | coeff = [] |
---|
| 297 | for j in xrange(len(lscoeff)): |
---|
| 298 | coeff.append(float(lscoeff[j])) |
---|
| 299 | res.append({'row': row, 'coeff': coeff, 'rms': rms}) |
---|
| 300 | |
---|
| 301 | return res |
---|
| 302 | |
---|
[2882] | 303 | def is_number(s): |
---|
| 304 | s = s.strip() |
---|
| 305 | res = True |
---|
| 306 | try: |
---|
| 307 | a = float(s) |
---|
| 308 | res = True |
---|
| 309 | except: |
---|
| 310 | res = False |
---|
| 311 | finally: |
---|
| 312 | return res |
---|
| 313 | |
---|
| 314 | def is_frequency(s): |
---|
| 315 | s = s.strip() |
---|
| 316 | return (s[-2:].lower() == "hz") |
---|
| 317 | |
---|
[2884] | 318 | def get_freq_by_string(s1, s2): |
---|
| 319 | if not (is_number(s1) and is_frequency(s2)): |
---|
[2882] | 320 | raise RuntimeError("Invalid input string.") |
---|
| 321 | |
---|
| 322 | prefix_list = ["a", "f", "p", "n", "u", "m", ".", "k", "M", "G", "T", "P", "E"] |
---|
| 323 | factor_list = [1e-18, 1e-15, 1e-12, 1e-9, 1e-6, 1e-3, 1.0, 1e+3, 1e+6, 1e+9, 1e+12, 1e+15, 1e+18] |
---|
| 324 | |
---|
[2884] | 325 | s1 = s1.strip() |
---|
| 326 | s2 = s2.strip() |
---|
[2882] | 327 | |
---|
[2884] | 328 | prefix = s2[-3:-2] |
---|
[2882] | 329 | if is_number(prefix): |
---|
[2884] | 330 | res1 = float(s1) |
---|
| 331 | res2 = float(s2[:-2]) |
---|
[2882] | 332 | else: |
---|
[2884] | 333 | factor = factor_list[prefix_list.index(prefix)] |
---|
| 334 | res1 = float(s1) * factor |
---|
| 335 | res2 = float(s2[:-3]) * factor |
---|
[2882] | 336 | |
---|
[2884] | 337 | return (res1, res2) |
---|
[2882] | 338 | |
---|
| 339 | def is_velocity(s): |
---|
| 340 | s = s.strip() |
---|
| 341 | return (s[-3:].lower() == "m/s") |
---|
| 342 | |
---|
[2884] | 343 | def get_velocity_by_string(s1, s2): |
---|
| 344 | if not (is_number(s1) and is_velocity(s2)): |
---|
[2882] | 345 | raise RuntimeError("Invalid input string.") |
---|
| 346 | |
---|
[2884] | 347 | # note that the default velocity unit is km/s |
---|
[2882] | 348 | prefix_list = [".", "k"] |
---|
| 349 | factor_list = [1e-3, 1.0] |
---|
| 350 | |
---|
[2884] | 351 | s1 = s1.strip() |
---|
| 352 | s2 = s2.strip() |
---|
[2882] | 353 | |
---|
[2884] | 354 | prefix = s2[-4:-3] |
---|
| 355 | if is_number(prefix): # in case velocity unit m/s |
---|
| 356 | res1 = float(s1) * 1e-3 |
---|
| 357 | res2 = float(s2[:-3]) * 1e-3 |
---|
[2882] | 358 | else: |
---|
[2884] | 359 | factor = factor_list[prefix_list.index(prefix)] |
---|
| 360 | res1 = float(s1) * factor |
---|
| 361 | res2 = float(s2[:-4]) * factor |
---|
[2882] | 362 | |
---|
[2884] | 363 | return (res1, res2) |
---|
[2882] | 364 | |
---|
[2889] | 365 | def get_frequency_by_velocity(restfreq, vel, doppler): |
---|
[2882] | 366 | # vel is in unit of km/s |
---|
| 367 | |
---|
| 368 | # speed of light |
---|
| 369 | vel_c = 299792.458 |
---|
| 370 | |
---|
| 371 | import math |
---|
| 372 | r = vel / vel_c |
---|
| 373 | |
---|
[2889] | 374 | if doppler.lower() == 'radio': |
---|
| 375 | return restfreq * (1.0 - r) |
---|
| 376 | if doppler.lower() == 'optical': |
---|
| 377 | return restfreq / (1.0 + r) |
---|
| 378 | else: |
---|
| 379 | return restfreq * math.sqrt((1.0 - r) / (1.0 + r)) |
---|
[2882] | 380 | |
---|
[2891] | 381 | def get_restfreq_in_Hz(s_restfreq): |
---|
| 382 | value = 0.0 |
---|
| 383 | unit = "" |
---|
| 384 | s = s_restfreq.replace(" ","") |
---|
[2889] | 385 | |
---|
[2891] | 386 | for i in range(len(s))[::-1]: |
---|
| 387 | if s[i].isalpha(): |
---|
| 388 | unit = s[i] + unit |
---|
| 389 | else: |
---|
| 390 | value = float(s[0:i+1]) |
---|
| 391 | break |
---|
| 392 | |
---|
| 393 | if (unit == "") or (unit.lower() == "hz"): |
---|
| 394 | return value |
---|
| 395 | elif (len(unit) == 3) and (unit[1:3].lower() == "hz"): |
---|
| 396 | unitprefix = unit[0] |
---|
| 397 | factor = 1.0 |
---|
| 398 | |
---|
| 399 | prefix_list = ["a", "f", "p", "n", "u", "m", ".", "k", "M", "G", "T", "P", "E"] |
---|
| 400 | factor_list = [1e-18, 1e-15, 1e-12, 1e-9, 1e-6, 1e-3, 1.0, 1e+3, 1e+6, 1e+9, 1e+12, 1e+15, 1e+18] |
---|
| 401 | factor = factor_list[prefix_list.index(unitprefix)] |
---|
| 402 | """ |
---|
| 403 | if (unitprefix == 'a'): |
---|
| 404 | factor = 1.0e-18 |
---|
| 405 | elif (unitprefix == 'f'): |
---|
| 406 | factor = 1.0e-15 |
---|
| 407 | elif (unitprefix == 'p'): |
---|
| 408 | factor = 1.0e-12 |
---|
| 409 | elif (unitprefix == 'n'): |
---|
| 410 | factor = 1.0e-9 |
---|
| 411 | elif (unitprefix == 'u'): |
---|
| 412 | factor = 1.0e-6 |
---|
| 413 | elif (unitprefix == 'm'): |
---|
| 414 | factor = 1.0e-3 |
---|
| 415 | elif (unitprefix == 'k'): |
---|
| 416 | factor = 1.0e+3 |
---|
| 417 | elif (unitprefix == 'M'): |
---|
| 418 | factor = 1.0e+6 |
---|
| 419 | elif (unitprefix == 'G'): |
---|
| 420 | factor = 1.0e+9 |
---|
| 421 | elif (unitprefix == 'T'): |
---|
| 422 | factor = 1.0e+12 |
---|
| 423 | elif (unitprefix == 'P'): |
---|
| 424 | factor = 1.0e+15 |
---|
| 425 | elif (unitprefix == 'E'): |
---|
| 426 | factor = 1.0e+18 |
---|
| 427 | """ |
---|
| 428 | return value*factor |
---|
| 429 | else: |
---|
| 430 | mesg = "wrong unit of restfreq." |
---|
| 431 | raise Exception, mesg |
---|
| 432 | |
---|
| 433 | def normalise_restfreq(in_restfreq): |
---|
| 434 | if isinstance(in_restfreq, float): |
---|
| 435 | return in_restfreq |
---|
| 436 | elif isinstance(in_restfreq, int) or isinstance(in_restfreq, long): |
---|
| 437 | return float(in_restfreq) |
---|
| 438 | elif isinstance(in_restfreq, str): |
---|
| 439 | return get_restfreq_in_Hz(in_restfreq) |
---|
| 440 | elif isinstance(in_restfreq, list) or isinstance(in_restfreq, numpy.ndarray): |
---|
| 441 | if isinstance(in_restfreq, numpy.ndarray): |
---|
| 442 | if len(in_restfreq.shape) > 1: |
---|
| 443 | mesg = "given in numpy.ndarray, in_restfreq must be 1-D." |
---|
| 444 | raise Exception, mesg |
---|
| 445 | |
---|
| 446 | res = [] |
---|
| 447 | for i in xrange(len(in_restfreq)): |
---|
| 448 | elem = in_restfreq[i] |
---|
| 449 | if isinstance(elem, float): |
---|
| 450 | res.append(elem) |
---|
| 451 | elif isinstance(elem, int) or isinstance(elem, long): |
---|
| 452 | res.append(float(elem)) |
---|
| 453 | elif isinstance(elem, str): |
---|
| 454 | res.append(get_restfreq_in_Hz(elem)) |
---|
| 455 | elif isinstance(elem, dict): |
---|
| 456 | if isinstance(elem["value"], float): |
---|
| 457 | res.append(elem) |
---|
| 458 | elif isinstance(elem["value"], int): |
---|
| 459 | dictelem = {} |
---|
| 460 | dictelem["name"] = elem["name"] |
---|
| 461 | dictelem["value"] = float(elem["value"]) |
---|
| 462 | res.append(dictelem) |
---|
| 463 | elif isinstance(elem["value"], str): |
---|
| 464 | dictelem = {} |
---|
| 465 | dictelem["name"] = elem["name"] |
---|
| 466 | dictelem["value"] = get_restfreq_in_Hz(elem["value"]) |
---|
| 467 | res.append(dictelem) |
---|
| 468 | else: |
---|
| 469 | mesg = "restfreq elements must be float, int, or string." |
---|
| 470 | raise Exception, mesg |
---|
| 471 | return res |
---|
| 472 | else: |
---|
| 473 | mesg = "wrong type of restfreq given." |
---|
| 474 | raise Exception, mesg |
---|
| 475 | |
---|
| 476 | def set_restfreq(s, restfreq): |
---|
| 477 | rfset = (restfreq != '') and (restfreq != []) |
---|
| 478 | if rfset: |
---|
| 479 | s.set_restfreqs(normalise_restfreq(restfreq)) |
---|
| 480 | |
---|
[876] | 481 | class scantable(Scantable): |
---|
[1846] | 482 | """\ |
---|
| 483 | The ASAP container for scans (single-dish data). |
---|
[102] | 484 | """ |
---|
[1819] | 485 | |
---|
[1862] | 486 | @asaplog_post_dec |
---|
[2315] | 487 | def __init__(self, filename, average=None, unit=None, parallactify=None, |
---|
| 488 | **args): |
---|
[1846] | 489 | """\ |
---|
[102] | 490 | Create a scantable from a saved one or make a reference |
---|
[1846] | 491 | |
---|
[102] | 492 | Parameters: |
---|
[1846] | 493 | |
---|
| 494 | filename: the name of an asap table on disk |
---|
| 495 | or |
---|
| 496 | the name of a rpfits/sdfits/ms file |
---|
| 497 | (integrations within scans are auto averaged |
---|
| 498 | and the whole file is read) or |
---|
| 499 | [advanced] a reference to an existing scantable |
---|
| 500 | |
---|
| 501 | average: average all integrations withinb a scan on read. |
---|
| 502 | The default (True) is taken from .asaprc. |
---|
| 503 | |
---|
[484] | 504 | unit: brightness unit; must be consistent with K or Jy. |
---|
[1846] | 505 | Over-rides the default selected by the filler |
---|
| 506 | (input rpfits/sdfits/ms) or replaces the value |
---|
| 507 | in existing scantables |
---|
| 508 | |
---|
[1920] | 509 | antenna: for MeasurementSet input data only: |
---|
[2349] | 510 | Antenna selection. integer (id) or string |
---|
| 511 | (name or id). |
---|
[1846] | 512 | |
---|
[2349] | 513 | parallactify: Indicate that the data had been parallactified. |
---|
| 514 | Default (false) is taken from rc file. |
---|
[1846] | 515 | |
---|
[2754] | 516 | getpt: Whether to import direction from MS/POINTING |
---|
| 517 | table properly or not. |
---|
| 518 | This is effective only when filename is MS. |
---|
| 519 | The default (True) is to import direction |
---|
| 520 | from MS/POINTING. |
---|
[710] | 521 | """ |
---|
[976] | 522 | if average is None: |
---|
[710] | 523 | average = rcParams['scantable.autoaverage'] |
---|
[1593] | 524 | parallactify = parallactify or rcParams['scantable.parallactify'] |
---|
[1259] | 525 | varlist = vars() |
---|
[876] | 526 | from asap._asap import stmath |
---|
[1819] | 527 | self._math = stmath( rcParams['insitu'] ) |
---|
[876] | 528 | if isinstance(filename, Scantable): |
---|
| 529 | Scantable.__init__(self, filename) |
---|
[181] | 530 | else: |
---|
[1697] | 531 | if isinstance(filename, str): |
---|
[976] | 532 | filename = os.path.expandvars(filename) |
---|
| 533 | filename = os.path.expanduser(filename) |
---|
| 534 | if not os.path.exists(filename): |
---|
| 535 | s = "File '%s' not found." % (filename) |
---|
| 536 | raise IOError(s) |
---|
[1697] | 537 | if is_scantable(filename): |
---|
| 538 | ondisk = rcParams['scantable.storage'] == 'disk' |
---|
| 539 | Scantable.__init__(self, filename, ondisk) |
---|
| 540 | if unit is not None: |
---|
| 541 | self.set_fluxunit(unit) |
---|
[2008] | 542 | if average: |
---|
| 543 | self._assign( self.average_time( scanav=True ) ) |
---|
[1819] | 544 | # do not reset to the default freqframe |
---|
| 545 | #self.set_freqframe(rcParams['scantable.freqframe']) |
---|
[1883] | 546 | elif is_ms(filename): |
---|
[1916] | 547 | # Measurement Set |
---|
| 548 | opts={'ms': {}} |
---|
[2844] | 549 | mskeys=['getpt','antenna'] |
---|
[1916] | 550 | for key in mskeys: |
---|
| 551 | if key in args.keys(): |
---|
| 552 | opts['ms'][key] = args[key] |
---|
| 553 | self._fill([filename], unit, average, opts) |
---|
[1893] | 554 | elif os.path.isfile(filename): |
---|
[2761] | 555 | opts={'nro': {}} |
---|
| 556 | nrokeys=['freqref'] |
---|
| 557 | for key in nrokeys: |
---|
| 558 | if key in args.keys(): |
---|
| 559 | opts['nro'][key] = args[key] |
---|
| 560 | self._fill([filename], unit, average, opts) |
---|
[2350] | 561 | # only apply to new data not "copy constructor" |
---|
| 562 | self.parallactify(parallactify) |
---|
[1883] | 563 | else: |
---|
[1819] | 564 | msg = "The given file '%s'is not a valid " \ |
---|
| 565 | "asap table." % (filename) |
---|
[1859] | 566 | raise IOError(msg) |
---|
[1118] | 567 | elif (isinstance(filename, list) or isinstance(filename, tuple)) \ |
---|
[976] | 568 | and isinstance(filename[-1], str): |
---|
[1916] | 569 | self._fill(filename, unit, average) |
---|
[1586] | 570 | self.parallactify(parallactify) |
---|
[1259] | 571 | self._add_history("scantable", varlist) |
---|
[102] | 572 | |
---|
[1862] | 573 | @asaplog_post_dec |
---|
[876] | 574 | def save(self, name=None, format=None, overwrite=False): |
---|
[1846] | 575 | """\ |
---|
[1280] | 576 | Store the scantable on disk. This can be an asap (aips++) Table, |
---|
| 577 | SDFITS or MS2 format. |
---|
[1846] | 578 | |
---|
[116] | 579 | Parameters: |
---|
[1846] | 580 | |
---|
[2431] | 581 | name: the name of the outputfile. For format 'ASCII' |
---|
[1093] | 582 | this is the root file name (data in 'name'.txt |
---|
[497] | 583 | and header in 'name'_header.txt) |
---|
[1855] | 584 | |
---|
[116] | 585 | format: an optional file format. Default is ASAP. |
---|
[1855] | 586 | Allowed are: |
---|
| 587 | |
---|
| 588 | * 'ASAP' (save as ASAP [aips++] Table), |
---|
| 589 | * 'SDFITS' (save as SDFITS file) |
---|
| 590 | * 'ASCII' (saves as ascii text file) |
---|
| 591 | * 'MS2' (saves as an casacore MeasurementSet V2) |
---|
[2315] | 592 | * 'FITS' (save as image FITS - not readable by |
---|
| 593 | class) |
---|
[1855] | 594 | * 'CLASS' (save as FITS readable by CLASS) |
---|
| 595 | |
---|
[411] | 596 | overwrite: If the file should be overwritten if it exists. |
---|
[256] | 597 | The default False is to return with warning |
---|
[411] | 598 | without writing the output. USE WITH CARE. |
---|
[1855] | 599 | |
---|
[1846] | 600 | Example:: |
---|
| 601 | |
---|
[116] | 602 | scan.save('myscan.asap') |
---|
[1118] | 603 | scan.save('myscan.sdfits', 'SDFITS') |
---|
[1846] | 604 | |
---|
[116] | 605 | """ |
---|
[411] | 606 | from os import path |
---|
[1593] | 607 | format = format or rcParams['scantable.save'] |
---|
[256] | 608 | suffix = '.'+format.lower() |
---|
[1118] | 609 | if name is None or name == "": |
---|
[256] | 610 | name = 'scantable'+suffix |
---|
[718] | 611 | msg = "No filename given. Using default name %s..." % name |
---|
| 612 | asaplog.push(msg) |
---|
[411] | 613 | name = path.expandvars(name) |
---|
[256] | 614 | if path.isfile(name) or path.isdir(name): |
---|
| 615 | if not overwrite: |
---|
[718] | 616 | msg = "File %s exists." % name |
---|
[1859] | 617 | raise IOError(msg) |
---|
[451] | 618 | format2 = format.upper() |
---|
| 619 | if format2 == 'ASAP': |
---|
[116] | 620 | self._save(name) |
---|
[2029] | 621 | elif format2 == 'MS2': |
---|
| 622 | msopt = {'ms': {'overwrite': overwrite } } |
---|
| 623 | from asap._asap import mswriter |
---|
| 624 | writer = mswriter( self ) |
---|
| 625 | writer.write( name, msopt ) |
---|
[116] | 626 | else: |
---|
[989] | 627 | from asap._asap import stwriter as stw |
---|
[1118] | 628 | writer = stw(format2) |
---|
| 629 | writer.write(self, name) |
---|
[116] | 630 | return |
---|
| 631 | |
---|
[102] | 632 | def copy(self): |
---|
[1846] | 633 | """Return a copy of this scantable. |
---|
| 634 | |
---|
| 635 | *Note*: |
---|
| 636 | |
---|
[1348] | 637 | This makes a full (deep) copy. scan2 = scan1 makes a reference. |
---|
[1846] | 638 | |
---|
| 639 | Example:: |
---|
| 640 | |
---|
[102] | 641 | copiedscan = scan.copy() |
---|
[1846] | 642 | |
---|
[102] | 643 | """ |
---|
[876] | 644 | sd = scantable(Scantable._copy(self)) |
---|
[113] | 645 | return sd |
---|
| 646 | |
---|
[1093] | 647 | def drop_scan(self, scanid=None): |
---|
[1846] | 648 | """\ |
---|
[1093] | 649 | Return a new scantable where the specified scan number(s) has(have) |
---|
| 650 | been dropped. |
---|
[1846] | 651 | |
---|
[1093] | 652 | Parameters: |
---|
[1846] | 653 | |
---|
[1093] | 654 | scanid: a (list of) scan number(s) |
---|
[1846] | 655 | |
---|
[1093] | 656 | """ |
---|
| 657 | from asap import _is_sequence_or_number as _is_valid |
---|
| 658 | from asap import _to_list |
---|
| 659 | from asap import unique |
---|
| 660 | if not _is_valid(scanid): |
---|
[2315] | 661 | raise RuntimeError( 'Please specify a scanno to drop from the' |
---|
| 662 | ' scantable' ) |
---|
[1859] | 663 | scanid = _to_list(scanid) |
---|
| 664 | allscans = unique([ self.getscan(i) for i in range(self.nrow())]) |
---|
| 665 | for sid in scanid: allscans.remove(sid) |
---|
| 666 | if len(allscans) == 0: |
---|
| 667 | raise ValueError("Can't remove all scans") |
---|
| 668 | sel = selector(scans=allscans) |
---|
| 669 | return self._select_copy(sel) |
---|
[1093] | 670 | |
---|
[1594] | 671 | def _select_copy(self, selection): |
---|
| 672 | orig = self.get_selection() |
---|
| 673 | self.set_selection(orig+selection) |
---|
| 674 | cp = self.copy() |
---|
| 675 | self.set_selection(orig) |
---|
| 676 | return cp |
---|
| 677 | |
---|
[102] | 678 | def get_scan(self, scanid=None): |
---|
[1855] | 679 | """\ |
---|
[102] | 680 | Return a specific scan (by scanno) or collection of scans (by |
---|
| 681 | source name) in a new scantable. |
---|
[1846] | 682 | |
---|
| 683 | *Note*: |
---|
| 684 | |
---|
[1348] | 685 | See scantable.drop_scan() for the inverse operation. |
---|
[1846] | 686 | |
---|
[102] | 687 | Parameters: |
---|
[1846] | 688 | |
---|
[513] | 689 | scanid: a (list of) scanno or a source name, unix-style |
---|
| 690 | patterns are accepted for source name matching, e.g. |
---|
| 691 | '*_R' gets all 'ref scans |
---|
[1846] | 692 | |
---|
| 693 | Example:: |
---|
| 694 | |
---|
[513] | 695 | # get all scans containing the source '323p459' |
---|
| 696 | newscan = scan.get_scan('323p459') |
---|
| 697 | # get all 'off' scans |
---|
| 698 | refscans = scan.get_scan('*_R') |
---|
| 699 | # get a susbset of scans by scanno (as listed in scan.summary()) |
---|
[1118] | 700 | newscan = scan.get_scan([0, 2, 7, 10]) |
---|
[1846] | 701 | |
---|
[102] | 702 | """ |
---|
| 703 | if scanid is None: |
---|
[1859] | 704 | raise RuntimeError( 'Please specify a scan no or name to ' |
---|
| 705 | 'retrieve from the scantable' ) |
---|
[102] | 706 | try: |
---|
[946] | 707 | bsel = self.get_selection() |
---|
| 708 | sel = selector() |
---|
[102] | 709 | if type(scanid) is str: |
---|
[946] | 710 | sel.set_name(scanid) |
---|
[1594] | 711 | return self._select_copy(sel) |
---|
[102] | 712 | elif type(scanid) is int: |
---|
[946] | 713 | sel.set_scans([scanid]) |
---|
[1594] | 714 | return self._select_copy(sel) |
---|
[381] | 715 | elif type(scanid) is list: |
---|
[946] | 716 | sel.set_scans(scanid) |
---|
[1594] | 717 | return self._select_copy(sel) |
---|
[381] | 718 | else: |
---|
[718] | 719 | msg = "Illegal scanid type, use 'int' or 'list' if ints." |
---|
[1859] | 720 | raise TypeError(msg) |
---|
[102] | 721 | except RuntimeError: |
---|
[1859] | 722 | raise |
---|
[102] | 723 | |
---|
| 724 | def __str__(self): |
---|
[2315] | 725 | tempFile = tempfile.NamedTemporaryFile() |
---|
| 726 | Scantable._summary(self, tempFile.name) |
---|
| 727 | tempFile.seek(0) |
---|
| 728 | asaplog.clear() |
---|
| 729 | return tempFile.file.read() |
---|
[102] | 730 | |
---|
[2315] | 731 | @asaplog_post_dec |
---|
[976] | 732 | def summary(self, filename=None): |
---|
[1846] | 733 | """\ |
---|
[102] | 734 | Print a summary of the contents of this scantable. |
---|
[1846] | 735 | |
---|
[102] | 736 | Parameters: |
---|
[1846] | 737 | |
---|
[1931] | 738 | filename: the name of a file to write the putput to |
---|
[102] | 739 | Default - no file output |
---|
[1846] | 740 | |
---|
[102] | 741 | """ |
---|
| 742 | if filename is not None: |
---|
[256] | 743 | if filename is "": |
---|
| 744 | filename = 'scantable_summary.txt' |
---|
[415] | 745 | from os.path import expandvars, isdir |
---|
[411] | 746 | filename = expandvars(filename) |
---|
[2286] | 747 | if isdir(filename): |
---|
[718] | 748 | msg = "Illegal file name '%s'." % (filename) |
---|
[1859] | 749 | raise IOError(msg) |
---|
[2286] | 750 | else: |
---|
| 751 | filename = "" |
---|
| 752 | Scantable._summary(self, filename) |
---|
[710] | 753 | |
---|
[1512] | 754 | def get_spectrum(self, rowno): |
---|
[1471] | 755 | """Return the spectrum for the current row in the scantable as a list. |
---|
[1846] | 756 | |
---|
[1471] | 757 | Parameters: |
---|
[1846] | 758 | |
---|
[1573] | 759 | rowno: the row number to retrieve the spectrum from |
---|
[1846] | 760 | |
---|
[1471] | 761 | """ |
---|
| 762 | return self._getspectrum(rowno) |
---|
[946] | 763 | |
---|
[1471] | 764 | def get_mask(self, rowno): |
---|
| 765 | """Return the mask for the current row in the scantable as a list. |
---|
[1846] | 766 | |
---|
[1471] | 767 | Parameters: |
---|
[1846] | 768 | |
---|
[1573] | 769 | rowno: the row number to retrieve the mask from |
---|
[1846] | 770 | |
---|
[1471] | 771 | """ |
---|
| 772 | return self._getmask(rowno) |
---|
| 773 | |
---|
| 774 | def set_spectrum(self, spec, rowno): |
---|
[1938] | 775 | """Set the spectrum for the current row in the scantable. |
---|
[1846] | 776 | |
---|
[1471] | 777 | Parameters: |
---|
[1846] | 778 | |
---|
[1855] | 779 | spec: the new spectrum |
---|
[1846] | 780 | |
---|
[1855] | 781 | rowno: the row number to set the spectrum for |
---|
| 782 | |
---|
[1471] | 783 | """ |
---|
[2348] | 784 | assert(len(spec) == self.nchan(self.getif(rowno))) |
---|
[1471] | 785 | return self._setspectrum(spec, rowno) |
---|
| 786 | |
---|
[1600] | 787 | def get_coordinate(self, rowno): |
---|
| 788 | """Return the (spectral) coordinate for a a given 'rowno'. |
---|
[1846] | 789 | |
---|
| 790 | *Note*: |
---|
| 791 | |
---|
[1600] | 792 | * This coordinate is only valid until a scantable method modifies |
---|
| 793 | the frequency axis. |
---|
| 794 | * This coordinate does contain the original frequency set-up |
---|
| 795 | NOT the new frame. The conversions however are done using the user |
---|
| 796 | specified frame (e.g. LSRK/TOPO). To get the 'real' coordinate, |
---|
| 797 | use scantable.freq_align first. Without it there is no closure, |
---|
[1846] | 798 | i.e.:: |
---|
[1600] | 799 | |
---|
[1846] | 800 | c = myscan.get_coordinate(0) |
---|
| 801 | c.to_frequency(c.get_reference_pixel()) != c.get_reference_value() |
---|
| 802 | |
---|
[1600] | 803 | Parameters: |
---|
[1846] | 804 | |
---|
[1600] | 805 | rowno: the row number for the spectral coordinate |
---|
| 806 | |
---|
| 807 | """ |
---|
| 808 | return coordinate(Scantable.get_coordinate(self, rowno)) |
---|
| 809 | |
---|
[946] | 810 | def get_selection(self): |
---|
[1846] | 811 | """\ |
---|
[1005] | 812 | Get the selection object currently set on this scantable. |
---|
[1846] | 813 | |
---|
| 814 | Example:: |
---|
| 815 | |
---|
[1005] | 816 | sel = scan.get_selection() |
---|
| 817 | sel.set_ifs(0) # select IF 0 |
---|
| 818 | scan.set_selection(sel) # apply modified selection |
---|
[1846] | 819 | |
---|
[946] | 820 | """ |
---|
| 821 | return selector(self._getselection()) |
---|
| 822 | |
---|
[1576] | 823 | def set_selection(self, selection=None, **kw): |
---|
[1846] | 824 | """\ |
---|
[1005] | 825 | Select a subset of the data. All following operations on this scantable |
---|
| 826 | are only applied to thi selection. |
---|
[1846] | 827 | |
---|
[1005] | 828 | Parameters: |
---|
[1697] | 829 | |
---|
[1846] | 830 | selection: a selector object (default unset the selection), or |
---|
[2431] | 831 | any combination of 'pols', 'ifs', 'beams', 'scans', |
---|
| 832 | 'cycles', 'name', 'query' |
---|
[1697] | 833 | |
---|
[1846] | 834 | Examples:: |
---|
[1697] | 835 | |
---|
[1005] | 836 | sel = selector() # create a selection object |
---|
[1118] | 837 | self.set_scans([0, 3]) # select SCANNO 0 and 3 |
---|
[1005] | 838 | scan.set_selection(sel) # set the selection |
---|
| 839 | scan.summary() # will only print summary of scanno 0 an 3 |
---|
| 840 | scan.set_selection() # unset the selection |
---|
[1697] | 841 | # or the equivalent |
---|
| 842 | scan.set_selection(scans=[0,3]) |
---|
| 843 | scan.summary() # will only print summary of scanno 0 an 3 |
---|
| 844 | scan.set_selection() # unset the selection |
---|
[1846] | 845 | |
---|
[946] | 846 | """ |
---|
[1576] | 847 | if selection is None: |
---|
| 848 | # reset |
---|
| 849 | if len(kw) == 0: |
---|
| 850 | selection = selector() |
---|
| 851 | else: |
---|
| 852 | # try keywords |
---|
| 853 | for k in kw: |
---|
| 854 | if k not in selector.fields: |
---|
[2320] | 855 | raise KeyError("Invalid selection key '%s', " |
---|
| 856 | "valid keys are %s" % (k, |
---|
| 857 | selector.fields)) |
---|
[1576] | 858 | selection = selector(**kw) |
---|
[946] | 859 | self._setselection(selection) |
---|
| 860 | |
---|
[1819] | 861 | def get_row(self, row=0, insitu=None): |
---|
[1846] | 862 | """\ |
---|
[1819] | 863 | Select a row in the scantable. |
---|
| 864 | Return a scantable with single row. |
---|
[1846] | 865 | |
---|
[1819] | 866 | Parameters: |
---|
[1846] | 867 | |
---|
| 868 | row: row no of integration, default is 0. |
---|
| 869 | insitu: if False a new scantable is returned. Otherwise, the |
---|
| 870 | scaling is done in-situ. The default is taken from .asaprc |
---|
| 871 | (False) |
---|
| 872 | |
---|
[1819] | 873 | """ |
---|
[2349] | 874 | if insitu is None: |
---|
| 875 | insitu = rcParams['insitu'] |
---|
[1819] | 876 | if not insitu: |
---|
| 877 | workscan = self.copy() |
---|
| 878 | else: |
---|
| 879 | workscan = self |
---|
| 880 | # Select a row |
---|
[2349] | 881 | sel = selector() |
---|
[1992] | 882 | sel.set_rows([row]) |
---|
[1819] | 883 | workscan.set_selection(sel) |
---|
| 884 | if not workscan.nrow() == 1: |
---|
[2349] | 885 | msg = "Could not identify single row. %d rows selected." \ |
---|
| 886 | % (workscan.nrow()) |
---|
[1819] | 887 | raise RuntimeError(msg) |
---|
| 888 | if insitu: |
---|
| 889 | self._assign(workscan) |
---|
| 890 | else: |
---|
| 891 | return workscan |
---|
| 892 | |
---|
[1862] | 893 | @asaplog_post_dec |
---|
[2957] | 894 | def stats(self, stat='stddev', mask=None, form='3.3f', row=None, skip_flaggedrow=False): |
---|
[1846] | 895 | """\ |
---|
[135] | 896 | Determine the specified statistic of the current beam/if/pol |
---|
[102] | 897 | Takes a 'mask' as an optional parameter to specify which |
---|
| 898 | channels should be excluded. |
---|
[1846] | 899 | |
---|
[102] | 900 | Parameters: |
---|
[1846] | 901 | |
---|
[1819] | 902 | stat: 'min', 'max', 'min_abc', 'max_abc', 'sumsq', 'sum', |
---|
| 903 | 'mean', 'var', 'stddev', 'avdev', 'rms', 'median' |
---|
[1855] | 904 | |
---|
[135] | 905 | mask: an optional mask specifying where the statistic |
---|
[102] | 906 | should be determined. |
---|
[1855] | 907 | |
---|
[1819] | 908 | form: format string to print statistic values |
---|
[1846] | 909 | |
---|
[1907] | 910 | row: row number of spectrum to process. |
---|
| 911 | (default is None: for all rows) |
---|
[1846] | 912 | |
---|
[2957] | 913 | skip_flaggedrow: if True, skip outputting text for flagged |
---|
| 914 | spectra. default is False. |
---|
| 915 | |
---|
[1907] | 916 | Example: |
---|
[113] | 917 | scan.set_unit('channel') |
---|
[1118] | 918 | msk = scan.create_mask([100, 200], [500, 600]) |
---|
[135] | 919 | scan.stats(stat='mean', mask=m) |
---|
[1846] | 920 | |
---|
[102] | 921 | """ |
---|
[1593] | 922 | mask = mask or [] |
---|
[876] | 923 | if not self._check_ifs(): |
---|
[1118] | 924 | raise ValueError("Cannot apply mask as the IFs have different " |
---|
| 925 | "number of channels. Please use setselection() " |
---|
| 926 | "to select individual IFs") |
---|
[1819] | 927 | getchan = False |
---|
| 928 | if stat.lower().startswith('min') or stat.lower().startswith('max'): |
---|
| 929 | chan = self._math._minmaxchan(self, mask, stat) |
---|
| 930 | getchan = True |
---|
| 931 | statvals = [] |
---|
[2957] | 932 | |
---|
| 933 | rtnabc = False |
---|
| 934 | if stat.lower().endswith('_abc'): |
---|
| 935 | rtnabc = True |
---|
| 936 | else: |
---|
[1907] | 937 | if row == None: |
---|
| 938 | statvals = self._math._stats(self, mask, stat) |
---|
| 939 | else: |
---|
| 940 | statvals = self._math._statsrow(self, mask, stat, int(row)) |
---|
[256] | 941 | |
---|
[1819] | 942 | #def cb(i): |
---|
| 943 | # return statvals[i] |
---|
[256] | 944 | |
---|
[1819] | 945 | #return self._row_callback(cb, stat) |
---|
[102] | 946 | |
---|
[1819] | 947 | label=stat |
---|
| 948 | #callback=cb |
---|
| 949 | out = "" |
---|
| 950 | #outvec = [] |
---|
| 951 | sep = '-'*50 |
---|
[1907] | 952 | |
---|
| 953 | if row == None: |
---|
| 954 | rows = xrange(self.nrow()) |
---|
| 955 | elif isinstance(row, int): |
---|
| 956 | rows = [ row ] |
---|
| 957 | |
---|
| 958 | for i in rows: |
---|
[1819] | 959 | refstr = '' |
---|
| 960 | statunit= '' |
---|
| 961 | if getchan: |
---|
| 962 | qx, qy = self.chan2data(rowno=i, chan=chan[i]) |
---|
| 963 | if rtnabc: |
---|
| 964 | statvals.append(qx['value']) |
---|
| 965 | refstr = ('(value: %'+form) % (qy['value'])+' ['+qy['unit']+'])' |
---|
| 966 | statunit= '['+qx['unit']+']' |
---|
| 967 | else: |
---|
| 968 | refstr = ('(@ %'+form) % (qx['value'])+' ['+qx['unit']+'])' |
---|
| 969 | |
---|
[2957] | 970 | if skip_flaggedrow and self._getflagrow(i): continue |
---|
| 971 | |
---|
[1819] | 972 | tm = self._gettime(i) |
---|
| 973 | src = self._getsourcename(i) |
---|
| 974 | out += 'Scan[%d] (%s) ' % (self.getscan(i), src) |
---|
| 975 | out += 'Time[%s]:\n' % (tm) |
---|
[1907] | 976 | if self.nbeam(-1) > 1: out += ' Beam[%d] ' % (self.getbeam(i)) |
---|
| 977 | if self.nif(-1) > 1: out += ' IF[%d] ' % (self.getif(i)) |
---|
| 978 | if self.npol(-1) > 1: out += ' Pol[%d] ' % (self.getpol(i)) |
---|
[1819] | 979 | #outvec.append(callback(i)) |
---|
[1907] | 980 | if len(rows) > 1: |
---|
| 981 | # out += ('= %'+form) % (outvec[i]) +' '+refstr+'\n' |
---|
| 982 | out += ('= %'+form) % (statvals[i]) +' '+refstr+'\n' |
---|
| 983 | else: |
---|
| 984 | # out += ('= %'+form) % (outvec[0]) +' '+refstr+'\n' |
---|
| 985 | out += ('= %'+form) % (statvals[0]) +' '+refstr+'\n' |
---|
[1819] | 986 | out += sep+"\n" |
---|
| 987 | |
---|
[1859] | 988 | import os |
---|
| 989 | if os.environ.has_key( 'USER' ): |
---|
| 990 | usr = os.environ['USER'] |
---|
| 991 | else: |
---|
| 992 | import commands |
---|
| 993 | usr = commands.getoutput( 'whoami' ) |
---|
| 994 | tmpfile = '/tmp/tmp_'+usr+'_casapy_asap_scantable_stats' |
---|
| 995 | f = open(tmpfile,'w') |
---|
| 996 | print >> f, sep |
---|
| 997 | print >> f, ' %s %s' % (label, statunit) |
---|
| 998 | print >> f, sep |
---|
| 999 | print >> f, out |
---|
| 1000 | f.close() |
---|
| 1001 | f = open(tmpfile,'r') |
---|
| 1002 | x = f.readlines() |
---|
| 1003 | f.close() |
---|
| 1004 | asaplog.push(''.join(x), False) |
---|
| 1005 | |
---|
[1819] | 1006 | return statvals |
---|
| 1007 | |
---|
| 1008 | def chan2data(self, rowno=0, chan=0): |
---|
[1846] | 1009 | """\ |
---|
[1819] | 1010 | Returns channel/frequency/velocity and spectral value |
---|
| 1011 | at an arbitrary row and channel in the scantable. |
---|
[1846] | 1012 | |
---|
[1819] | 1013 | Parameters: |
---|
[1846] | 1014 | |
---|
[1819] | 1015 | rowno: a row number in the scantable. Default is the |
---|
| 1016 | first row, i.e. rowno=0 |
---|
[1855] | 1017 | |
---|
[1819] | 1018 | chan: a channel in the scantable. Default is the first |
---|
| 1019 | channel, i.e. pos=0 |
---|
[1846] | 1020 | |
---|
[1819] | 1021 | """ |
---|
| 1022 | if isinstance(rowno, int) and isinstance(chan, int): |
---|
| 1023 | qx = {'unit': self.get_unit(), |
---|
| 1024 | 'value': self._getabcissa(rowno)[chan]} |
---|
| 1025 | qy = {'unit': self.get_fluxunit(), |
---|
| 1026 | 'value': self._getspectrum(rowno)[chan]} |
---|
| 1027 | return qx, qy |
---|
| 1028 | |
---|
[1118] | 1029 | def stddev(self, mask=None): |
---|
[1846] | 1030 | """\ |
---|
[135] | 1031 | Determine the standard deviation of the current beam/if/pol |
---|
| 1032 | Takes a 'mask' as an optional parameter to specify which |
---|
| 1033 | channels should be excluded. |
---|
[1846] | 1034 | |
---|
[135] | 1035 | Parameters: |
---|
[1846] | 1036 | |
---|
[135] | 1037 | mask: an optional mask specifying where the standard |
---|
| 1038 | deviation should be determined. |
---|
| 1039 | |
---|
[1846] | 1040 | Example:: |
---|
| 1041 | |
---|
[135] | 1042 | scan.set_unit('channel') |
---|
[1118] | 1043 | msk = scan.create_mask([100, 200], [500, 600]) |
---|
[135] | 1044 | scan.stddev(mask=m) |
---|
[1846] | 1045 | |
---|
[135] | 1046 | """ |
---|
[1118] | 1047 | return self.stats(stat='stddev', mask=mask); |
---|
[135] | 1048 | |
---|
[1003] | 1049 | |
---|
[1259] | 1050 | def get_column_names(self): |
---|
[1846] | 1051 | """\ |
---|
[1003] | 1052 | Return a list of column names, which can be used for selection. |
---|
| 1053 | """ |
---|
[1259] | 1054 | return list(Scantable.get_column_names(self)) |
---|
[1003] | 1055 | |
---|
[1730] | 1056 | def get_tsys(self, row=-1): |
---|
[1846] | 1057 | """\ |
---|
[113] | 1058 | Return the System temperatures. |
---|
[1846] | 1059 | |
---|
| 1060 | Parameters: |
---|
| 1061 | |
---|
| 1062 | row: the rowno to get the information for. (default all rows) |
---|
| 1063 | |
---|
[113] | 1064 | Returns: |
---|
[1846] | 1065 | |
---|
[876] | 1066 | a list of Tsys values for the current selection |
---|
[1846] | 1067 | |
---|
[113] | 1068 | """ |
---|
[1730] | 1069 | if row > -1: |
---|
| 1070 | return self._get_column(self._gettsys, row) |
---|
[876] | 1071 | return self._row_callback(self._gettsys, "Tsys") |
---|
[256] | 1072 | |
---|
[2406] | 1073 | def get_tsysspectrum(self, row=-1): |
---|
| 1074 | """\ |
---|
| 1075 | Return the channel dependent system temperatures. |
---|
[1730] | 1076 | |
---|
[2406] | 1077 | Parameters: |
---|
| 1078 | |
---|
| 1079 | row: the rowno to get the information for. (default all rows) |
---|
| 1080 | |
---|
| 1081 | Returns: |
---|
| 1082 | |
---|
| 1083 | a list of Tsys values for the current selection |
---|
| 1084 | |
---|
| 1085 | """ |
---|
| 1086 | return self._get_column( self._gettsysspectrum, row ) |
---|
| 1087 | |
---|
[2791] | 1088 | def set_tsys(self, values, row=-1): |
---|
| 1089 | """\ |
---|
| 1090 | Set the Tsys value(s) of the given 'row' or the whole scantable |
---|
| 1091 | (selection). |
---|
| 1092 | |
---|
| 1093 | Parameters: |
---|
| 1094 | |
---|
| 1095 | values: a scalar or list (if Tsys is a vector) of Tsys value(s) |
---|
| 1096 | row: the row number to apply Tsys values to. |
---|
| 1097 | (default all rows) |
---|
| 1098 | |
---|
| 1099 | """ |
---|
| 1100 | |
---|
| 1101 | if not hasattr(values, "__len__"): |
---|
| 1102 | values = [values] |
---|
| 1103 | self._settsys(values, row) |
---|
| 1104 | |
---|
[1730] | 1105 | def get_weather(self, row=-1): |
---|
[1846] | 1106 | """\ |
---|
[2930] | 1107 | Return the weather information. |
---|
[1846] | 1108 | |
---|
| 1109 | Parameters: |
---|
| 1110 | |
---|
| 1111 | row: the rowno to get the information for. (default all rows) |
---|
| 1112 | |
---|
| 1113 | Returns: |
---|
| 1114 | |
---|
| 1115 | a dict or list of of dicts of values for the current selection |
---|
| 1116 | |
---|
| 1117 | """ |
---|
[2930] | 1118 | if row >= len(self): |
---|
| 1119 | raise IndexError("row out of range") |
---|
[1730] | 1120 | values = self._get_column(self._get_weather, row) |
---|
| 1121 | if row > -1: |
---|
| 1122 | return {'temperature': values[0], |
---|
| 1123 | 'pressure': values[1], 'humidity' : values[2], |
---|
| 1124 | 'windspeed' : values[3], 'windaz' : values[4] |
---|
| 1125 | } |
---|
| 1126 | else: |
---|
| 1127 | out = [] |
---|
| 1128 | for r in values: |
---|
| 1129 | out.append({'temperature': r[0], |
---|
| 1130 | 'pressure': r[1], 'humidity' : r[2], |
---|
| 1131 | 'windspeed' : r[3], 'windaz' : r[4] |
---|
| 1132 | }) |
---|
| 1133 | return out |
---|
| 1134 | |
---|
[876] | 1135 | def _row_callback(self, callback, label): |
---|
| 1136 | out = "" |
---|
[1118] | 1137 | outvec = [] |
---|
[1590] | 1138 | sep = '-'*50 |
---|
[876] | 1139 | for i in range(self.nrow()): |
---|
| 1140 | tm = self._gettime(i) |
---|
| 1141 | src = self._getsourcename(i) |
---|
[1590] | 1142 | out += 'Scan[%d] (%s) ' % (self.getscan(i), src) |
---|
[876] | 1143 | out += 'Time[%s]:\n' % (tm) |
---|
[1590] | 1144 | if self.nbeam(-1) > 1: |
---|
| 1145 | out += ' Beam[%d] ' % (self.getbeam(i)) |
---|
| 1146 | if self.nif(-1) > 1: out += ' IF[%d] ' % (self.getif(i)) |
---|
| 1147 | if self.npol(-1) > 1: out += ' Pol[%d] ' % (self.getpol(i)) |
---|
[876] | 1148 | outvec.append(callback(i)) |
---|
| 1149 | out += '= %3.3f\n' % (outvec[i]) |
---|
[1590] | 1150 | out += sep+'\n' |
---|
[1859] | 1151 | |
---|
| 1152 | asaplog.push(sep) |
---|
| 1153 | asaplog.push(" %s" % (label)) |
---|
| 1154 | asaplog.push(sep) |
---|
| 1155 | asaplog.push(out) |
---|
[1861] | 1156 | asaplog.post() |
---|
[1175] | 1157 | return outvec |
---|
[256] | 1158 | |
---|
[1947] | 1159 | def _get_column(self, callback, row=-1, *args): |
---|
[1070] | 1160 | """ |
---|
| 1161 | """ |
---|
| 1162 | if row == -1: |
---|
[1947] | 1163 | return [callback(i, *args) for i in range(self.nrow())] |
---|
[1070] | 1164 | else: |
---|
[1819] | 1165 | if 0 <= row < self.nrow(): |
---|
[1947] | 1166 | return callback(row, *args) |
---|
[256] | 1167 | |
---|
[1070] | 1168 | |
---|
[1948] | 1169 | def get_time(self, row=-1, asdatetime=False, prec=-1): |
---|
[1846] | 1170 | """\ |
---|
[113] | 1171 | Get a list of time stamps for the observations. |
---|
[1938] | 1172 | Return a datetime object or a string (default) for each |
---|
| 1173 | integration time stamp in the scantable. |
---|
[1846] | 1174 | |
---|
[113] | 1175 | Parameters: |
---|
[1846] | 1176 | |
---|
[1348] | 1177 | row: row no of integration. Default -1 return all rows |
---|
[1855] | 1178 | |
---|
[1348] | 1179 | asdatetime: return values as datetime objects rather than strings |
---|
[1846] | 1180 | |
---|
[2349] | 1181 | prec: number of digits shown. Default -1 to automatic |
---|
| 1182 | calculation. |
---|
[1948] | 1183 | Note this number is equals to the digits of MVTime, |
---|
| 1184 | i.e., 0<prec<3: dates with hh:: only, |
---|
| 1185 | <5: with hh:mm:, <7 or 0: with hh:mm:ss, |
---|
[1947] | 1186 | and 6> : with hh:mm:ss.tt... (prec-6 t's added) |
---|
| 1187 | |
---|
[113] | 1188 | """ |
---|
[1175] | 1189 | from datetime import datetime |
---|
[1948] | 1190 | if prec < 0: |
---|
| 1191 | # automagically set necessary precision +1 |
---|
[2349] | 1192 | prec = 7 - \ |
---|
| 1193 | numpy.floor(numpy.log10(numpy.min(self.get_inttime(row)))) |
---|
[1948] | 1194 | prec = max(6, int(prec)) |
---|
| 1195 | else: |
---|
| 1196 | prec = max(0, prec) |
---|
| 1197 | if asdatetime: |
---|
| 1198 | #precision can be 1 millisecond at max |
---|
| 1199 | prec = min(12, prec) |
---|
| 1200 | |
---|
[1947] | 1201 | times = self._get_column(self._gettime, row, prec) |
---|
[1348] | 1202 | if not asdatetime: |
---|
[1392] | 1203 | return times |
---|
[1947] | 1204 | format = "%Y/%m/%d/%H:%M:%S.%f" |
---|
| 1205 | if prec < 7: |
---|
| 1206 | nsub = 1 + (((6-prec)/2) % 3) |
---|
| 1207 | substr = [".%f","%S","%M"] |
---|
| 1208 | for i in range(nsub): |
---|
| 1209 | format = format.replace(substr[i],"") |
---|
[1175] | 1210 | if isinstance(times, list): |
---|
[1947] | 1211 | return [datetime.strptime(i, format) for i in times] |
---|
[1175] | 1212 | else: |
---|
[1947] | 1213 | return datetime.strptime(times, format) |
---|
[102] | 1214 | |
---|
[1348] | 1215 | |
---|
| 1216 | def get_inttime(self, row=-1): |
---|
[1846] | 1217 | """\ |
---|
[1348] | 1218 | Get a list of integration times for the observations. |
---|
| 1219 | Return a time in seconds for each integration in the scantable. |
---|
[1846] | 1220 | |
---|
[1348] | 1221 | Parameters: |
---|
[1846] | 1222 | |
---|
[1348] | 1223 | row: row no of integration. Default -1 return all rows. |
---|
[1846] | 1224 | |
---|
[1348] | 1225 | """ |
---|
[1573] | 1226 | return self._get_column(self._getinttime, row) |
---|
[1348] | 1227 | |
---|
[1573] | 1228 | |
---|
[714] | 1229 | def get_sourcename(self, row=-1): |
---|
[1846] | 1230 | """\ |
---|
[794] | 1231 | Get a list source names for the observations. |
---|
[714] | 1232 | Return a string for each integration in the scantable. |
---|
| 1233 | Parameters: |
---|
[1846] | 1234 | |
---|
[1348] | 1235 | row: row no of integration. Default -1 return all rows. |
---|
[1846] | 1236 | |
---|
[714] | 1237 | """ |
---|
[1070] | 1238 | return self._get_column(self._getsourcename, row) |
---|
[714] | 1239 | |
---|
[794] | 1240 | def get_elevation(self, row=-1): |
---|
[1846] | 1241 | """\ |
---|
[794] | 1242 | Get a list of elevations for the observations. |
---|
| 1243 | Return a float for each integration in the scantable. |
---|
[1846] | 1244 | |
---|
[794] | 1245 | Parameters: |
---|
[1846] | 1246 | |
---|
[1348] | 1247 | row: row no of integration. Default -1 return all rows. |
---|
[1846] | 1248 | |
---|
[794] | 1249 | """ |
---|
[1070] | 1250 | return self._get_column(self._getelevation, row) |
---|
[794] | 1251 | |
---|
| 1252 | def get_azimuth(self, row=-1): |
---|
[1846] | 1253 | """\ |
---|
[794] | 1254 | Get a list of azimuths for the observations. |
---|
| 1255 | Return a float for each integration in the scantable. |
---|
[1846] | 1256 | |
---|
[794] | 1257 | Parameters: |
---|
[1348] | 1258 | row: row no of integration. Default -1 return all rows. |
---|
[1846] | 1259 | |
---|
[794] | 1260 | """ |
---|
[1070] | 1261 | return self._get_column(self._getazimuth, row) |
---|
[794] | 1262 | |
---|
| 1263 | def get_parangle(self, row=-1): |
---|
[1846] | 1264 | """\ |
---|
[794] | 1265 | Get a list of parallactic angles for the observations. |
---|
| 1266 | Return a float for each integration in the scantable. |
---|
[1846] | 1267 | |
---|
[794] | 1268 | Parameters: |
---|
[1846] | 1269 | |
---|
[1348] | 1270 | row: row no of integration. Default -1 return all rows. |
---|
[1846] | 1271 | |
---|
[794] | 1272 | """ |
---|
[1070] | 1273 | return self._get_column(self._getparangle, row) |
---|
[794] | 1274 | |
---|
[1070] | 1275 | def get_direction(self, row=-1): |
---|
| 1276 | """ |
---|
| 1277 | Get a list of Positions on the sky (direction) for the observations. |
---|
[1594] | 1278 | Return a string for each integration in the scantable. |
---|
[1855] | 1279 | |
---|
[1070] | 1280 | Parameters: |
---|
[1855] | 1281 | |
---|
[1070] | 1282 | row: row no of integration. Default -1 return all rows |
---|
[1855] | 1283 | |
---|
[1070] | 1284 | """ |
---|
| 1285 | return self._get_column(self._getdirection, row) |
---|
| 1286 | |
---|
[1391] | 1287 | def get_directionval(self, row=-1): |
---|
[1846] | 1288 | """\ |
---|
[1391] | 1289 | Get a list of Positions on the sky (direction) for the observations. |
---|
| 1290 | Return a float for each integration in the scantable. |
---|
[1846] | 1291 | |
---|
[1391] | 1292 | Parameters: |
---|
[1846] | 1293 | |
---|
[1391] | 1294 | row: row no of integration. Default -1 return all rows |
---|
[1846] | 1295 | |
---|
[1391] | 1296 | """ |
---|
| 1297 | return self._get_column(self._getdirectionvec, row) |
---|
| 1298 | |
---|
[1862] | 1299 | @asaplog_post_dec |
---|
[102] | 1300 | def set_unit(self, unit='channel'): |
---|
[1846] | 1301 | """\ |
---|
[102] | 1302 | Set the unit for all following operations on this scantable |
---|
[1846] | 1303 | |
---|
[102] | 1304 | Parameters: |
---|
[1846] | 1305 | |
---|
| 1306 | unit: optional unit, default is 'channel'. Use one of '*Hz', |
---|
| 1307 | 'km/s', 'channel' or equivalent '' |
---|
| 1308 | |
---|
[102] | 1309 | """ |
---|
[484] | 1310 | varlist = vars() |
---|
[1118] | 1311 | if unit in ['', 'pixel', 'channel']: |
---|
[113] | 1312 | unit = '' |
---|
| 1313 | inf = list(self._getcoordinfo()) |
---|
| 1314 | inf[0] = unit |
---|
| 1315 | self._setcoordinfo(inf) |
---|
[1118] | 1316 | self._add_history("set_unit", varlist) |
---|
[113] | 1317 | |
---|
[1862] | 1318 | @asaplog_post_dec |
---|
[484] | 1319 | def set_instrument(self, instr): |
---|
[1846] | 1320 | """\ |
---|
[1348] | 1321 | Set the instrument for subsequent processing. |
---|
[1846] | 1322 | |
---|
[358] | 1323 | Parameters: |
---|
[1846] | 1324 | |
---|
[710] | 1325 | instr: Select from 'ATPKSMB', 'ATPKSHOH', 'ATMOPRA', |
---|
[407] | 1326 | 'DSS-43' (Tid), 'CEDUNA', and 'HOBART' |
---|
[1846] | 1327 | |
---|
[358] | 1328 | """ |
---|
| 1329 | self._setInstrument(instr) |
---|
[1118] | 1330 | self._add_history("set_instument", vars()) |
---|
[358] | 1331 | |
---|
[1862] | 1332 | @asaplog_post_dec |
---|
[1190] | 1333 | def set_feedtype(self, feedtype): |
---|
[1846] | 1334 | """\ |
---|
[1190] | 1335 | Overwrite the feed type, which might not be set correctly. |
---|
[1846] | 1336 | |
---|
[1190] | 1337 | Parameters: |
---|
[1846] | 1338 | |
---|
[1190] | 1339 | feedtype: 'linear' or 'circular' |
---|
[1846] | 1340 | |
---|
[1190] | 1341 | """ |
---|
| 1342 | self._setfeedtype(feedtype) |
---|
| 1343 | self._add_history("set_feedtype", vars()) |
---|
| 1344 | |
---|
[1862] | 1345 | @asaplog_post_dec |
---|
[2897] | 1346 | def get_doppler(self): |
---|
| 1347 | """\ |
---|
| 1348 | Get the doppler. |
---|
| 1349 | """ |
---|
| 1350 | return self._getcoordinfo()[2] |
---|
| 1351 | |
---|
| 1352 | @asaplog_post_dec |
---|
[276] | 1353 | def set_doppler(self, doppler='RADIO'): |
---|
[1846] | 1354 | """\ |
---|
[276] | 1355 | Set the doppler for all following operations on this scantable. |
---|
[1846] | 1356 | |
---|
[276] | 1357 | Parameters: |
---|
[1846] | 1358 | |
---|
[276] | 1359 | doppler: One of 'RADIO', 'OPTICAL', 'Z', 'BETA', 'GAMMA' |
---|
[1846] | 1360 | |
---|
[276] | 1361 | """ |
---|
[484] | 1362 | varlist = vars() |
---|
[276] | 1363 | inf = list(self._getcoordinfo()) |
---|
| 1364 | inf[2] = doppler |
---|
| 1365 | self._setcoordinfo(inf) |
---|
[1118] | 1366 | self._add_history("set_doppler", vars()) |
---|
[710] | 1367 | |
---|
[1862] | 1368 | @asaplog_post_dec |
---|
[226] | 1369 | def set_freqframe(self, frame=None): |
---|
[1846] | 1370 | """\ |
---|
[113] | 1371 | Set the frame type of the Spectral Axis. |
---|
[1846] | 1372 | |
---|
[113] | 1373 | Parameters: |
---|
[1846] | 1374 | |
---|
[591] | 1375 | frame: an optional frame type, default 'LSRK'. Valid frames are: |
---|
[1819] | 1376 | 'TOPO', 'LSRD', 'LSRK', 'BARY', |
---|
[1118] | 1377 | 'GEO', 'GALACTO', 'LGROUP', 'CMB' |
---|
[1846] | 1378 | |
---|
| 1379 | Example:: |
---|
| 1380 | |
---|
[113] | 1381 | scan.set_freqframe('BARY') |
---|
[1846] | 1382 | |
---|
[113] | 1383 | """ |
---|
[1593] | 1384 | frame = frame or rcParams['scantable.freqframe'] |
---|
[484] | 1385 | varlist = vars() |
---|
[1819] | 1386 | # "REST" is not implemented in casacore |
---|
| 1387 | #valid = ['REST', 'TOPO', 'LSRD', 'LSRK', 'BARY', \ |
---|
| 1388 | # 'GEO', 'GALACTO', 'LGROUP', 'CMB'] |
---|
| 1389 | valid = ['TOPO', 'LSRD', 'LSRK', 'BARY', \ |
---|
[1118] | 1390 | 'GEO', 'GALACTO', 'LGROUP', 'CMB'] |
---|
[591] | 1391 | |
---|
[989] | 1392 | if frame in valid: |
---|
[113] | 1393 | inf = list(self._getcoordinfo()) |
---|
| 1394 | inf[1] = frame |
---|
| 1395 | self._setcoordinfo(inf) |
---|
[1118] | 1396 | self._add_history("set_freqframe", varlist) |
---|
[102] | 1397 | else: |
---|
[1118] | 1398 | msg = "Please specify a valid freq type. Valid types are:\n", valid |
---|
[1859] | 1399 | raise TypeError(msg) |
---|
[710] | 1400 | |
---|
[1862] | 1401 | @asaplog_post_dec |
---|
[989] | 1402 | def set_dirframe(self, frame=""): |
---|
[1846] | 1403 | """\ |
---|
[989] | 1404 | Set the frame type of the Direction on the sky. |
---|
[1846] | 1405 | |
---|
[989] | 1406 | Parameters: |
---|
[1846] | 1407 | |
---|
[989] | 1408 | frame: an optional frame type, default ''. Valid frames are: |
---|
| 1409 | 'J2000', 'B1950', 'GALACTIC' |
---|
[1846] | 1410 | |
---|
| 1411 | Example: |
---|
| 1412 | |
---|
[989] | 1413 | scan.set_dirframe('GALACTIC') |
---|
[1846] | 1414 | |
---|
[989] | 1415 | """ |
---|
| 1416 | varlist = vars() |
---|
[1859] | 1417 | Scantable.set_dirframe(self, frame) |
---|
[1118] | 1418 | self._add_history("set_dirframe", varlist) |
---|
[989] | 1419 | |
---|
[113] | 1420 | def get_unit(self): |
---|
[1846] | 1421 | """\ |
---|
[113] | 1422 | Get the default unit set in this scantable |
---|
[1846] | 1423 | |
---|
[113] | 1424 | Returns: |
---|
[1846] | 1425 | |
---|
[113] | 1426 | A unit string |
---|
[1846] | 1427 | |
---|
[113] | 1428 | """ |
---|
| 1429 | inf = self._getcoordinfo() |
---|
| 1430 | unit = inf[0] |
---|
| 1431 | if unit == '': unit = 'channel' |
---|
| 1432 | return unit |
---|
[102] | 1433 | |
---|
[1862] | 1434 | @asaplog_post_dec |
---|
[158] | 1435 | def get_abcissa(self, rowno=0): |
---|
[1846] | 1436 | """\ |
---|
[158] | 1437 | Get the abcissa in the current coordinate setup for the currently |
---|
[113] | 1438 | selected Beam/IF/Pol |
---|
[1846] | 1439 | |
---|
[113] | 1440 | Parameters: |
---|
[1846] | 1441 | |
---|
[226] | 1442 | rowno: an optional row number in the scantable. Default is the |
---|
| 1443 | first row, i.e. rowno=0 |
---|
[1846] | 1444 | |
---|
[113] | 1445 | Returns: |
---|
[1846] | 1446 | |
---|
[1348] | 1447 | The abcissa values and the format string (as a dictionary) |
---|
[1846] | 1448 | |
---|
[113] | 1449 | """ |
---|
[256] | 1450 | abc = self._getabcissa(rowno) |
---|
[710] | 1451 | lbl = self._getabcissalabel(rowno) |
---|
[158] | 1452 | return abc, lbl |
---|
[113] | 1453 | |
---|
[1862] | 1454 | @asaplog_post_dec |
---|
[2322] | 1455 | def flag(self, mask=None, unflag=False, row=-1): |
---|
[1846] | 1456 | """\ |
---|
[1001] | 1457 | Flag the selected data using an optional channel mask. |
---|
[1846] | 1458 | |
---|
[1001] | 1459 | Parameters: |
---|
[1846] | 1460 | |
---|
[1001] | 1461 | mask: an optional channel mask, created with create_mask. Default |
---|
| 1462 | (no mask) is all channels. |
---|
[1855] | 1463 | |
---|
[1819] | 1464 | unflag: if True, unflag the data |
---|
[1846] | 1465 | |
---|
[2322] | 1466 | row: an optional row number in the scantable. |
---|
| 1467 | Default -1 flags all rows |
---|
| 1468 | |
---|
[1001] | 1469 | """ |
---|
| 1470 | varlist = vars() |
---|
[1593] | 1471 | mask = mask or [] |
---|
[1994] | 1472 | self._flag(row, mask, unflag) |
---|
[1001] | 1473 | self._add_history("flag", varlist) |
---|
| 1474 | |
---|
[1862] | 1475 | @asaplog_post_dec |
---|
[2322] | 1476 | def flag_row(self, rows=None, unflag=False): |
---|
[1846] | 1477 | """\ |
---|
[1819] | 1478 | Flag the selected data in row-based manner. |
---|
[1846] | 1479 | |
---|
[1819] | 1480 | Parameters: |
---|
[1846] | 1481 | |
---|
[1843] | 1482 | rows: list of row numbers to be flagged. Default is no row |
---|
[2322] | 1483 | (must be explicitly specified to execute row-based |
---|
| 1484 | flagging). |
---|
[1855] | 1485 | |
---|
[1819] | 1486 | unflag: if True, unflag the data. |
---|
[1846] | 1487 | |
---|
[1819] | 1488 | """ |
---|
| 1489 | varlist = vars() |
---|
[2322] | 1490 | if rows is None: |
---|
| 1491 | rows = [] |
---|
[1859] | 1492 | self._flag_row(rows, unflag) |
---|
[1819] | 1493 | self._add_history("flag_row", varlist) |
---|
| 1494 | |
---|
[1862] | 1495 | @asaplog_post_dec |
---|
[1819] | 1496 | def clip(self, uthres=None, dthres=None, clipoutside=True, unflag=False): |
---|
[1846] | 1497 | """\ |
---|
[1819] | 1498 | Flag the selected data outside a specified range (in channel-base) |
---|
[1846] | 1499 | |
---|
[1819] | 1500 | Parameters: |
---|
[1846] | 1501 | |
---|
[1819] | 1502 | uthres: upper threshold. |
---|
[1855] | 1503 | |
---|
[1819] | 1504 | dthres: lower threshold |
---|
[1846] | 1505 | |
---|
[2322] | 1506 | clipoutside: True for flagging data outside the range |
---|
| 1507 | [dthres:uthres]. |
---|
[1928] | 1508 | False for flagging data inside the range. |
---|
[1855] | 1509 | |
---|
[1846] | 1510 | unflag: if True, unflag the data. |
---|
| 1511 | |
---|
[1819] | 1512 | """ |
---|
| 1513 | varlist = vars() |
---|
[1859] | 1514 | self._clip(uthres, dthres, clipoutside, unflag) |
---|
[1819] | 1515 | self._add_history("clip", varlist) |
---|
| 1516 | |
---|
[1862] | 1517 | @asaplog_post_dec |
---|
[1584] | 1518 | def lag_flag(self, start, end, unit="MHz", insitu=None): |
---|
[1846] | 1519 | """\ |
---|
[1192] | 1520 | Flag the data in 'lag' space by providing a frequency to remove. |
---|
[2177] | 1521 | Flagged data in the scantable get interpolated over the region. |
---|
[1192] | 1522 | No taper is applied. |
---|
[1846] | 1523 | |
---|
[1192] | 1524 | Parameters: |
---|
[1846] | 1525 | |
---|
[1579] | 1526 | start: the start frequency (really a period within the |
---|
| 1527 | bandwidth) or period to remove |
---|
[1855] | 1528 | |
---|
[1579] | 1529 | end: the end frequency or period to remove |
---|
[1855] | 1530 | |
---|
[2431] | 1531 | unit: the frequency unit (default 'MHz') or '' for |
---|
[1579] | 1532 | explicit lag channels |
---|
[1846] | 1533 | |
---|
| 1534 | *Notes*: |
---|
| 1535 | |
---|
[1579] | 1536 | It is recommended to flag edges of the band or strong |
---|
[1348] | 1537 | signals beforehand. |
---|
[1846] | 1538 | |
---|
[1192] | 1539 | """ |
---|
| 1540 | if insitu is None: insitu = rcParams['insitu'] |
---|
| 1541 | self._math._setinsitu(insitu) |
---|
| 1542 | varlist = vars() |
---|
[1579] | 1543 | base = { "GHz": 1000000000., "MHz": 1000000., "kHz": 1000., "Hz": 1.} |
---|
| 1544 | if not (unit == "" or base.has_key(unit)): |
---|
[1192] | 1545 | raise ValueError("%s is not a valid unit." % unit) |
---|
[1859] | 1546 | if unit == "": |
---|
| 1547 | s = scantable(self._math._lag_flag(self, start, end, "lags")) |
---|
| 1548 | else: |
---|
| 1549 | s = scantable(self._math._lag_flag(self, start*base[unit], |
---|
| 1550 | end*base[unit], "frequency")) |
---|
[1192] | 1551 | s._add_history("lag_flag", varlist) |
---|
| 1552 | if insitu: |
---|
| 1553 | self._assign(s) |
---|
| 1554 | else: |
---|
| 1555 | return s |
---|
[1001] | 1556 | |
---|
[1862] | 1557 | @asaplog_post_dec |
---|
[2349] | 1558 | def fft(self, rowno=None, mask=None, getrealimag=False): |
---|
[2177] | 1559 | """\ |
---|
| 1560 | Apply FFT to the spectra. |
---|
| 1561 | Flagged data in the scantable get interpolated over the region. |
---|
| 1562 | |
---|
| 1563 | Parameters: |
---|
[2186] | 1564 | |
---|
| 1565 | rowno: The row number(s) to be processed. int, list |
---|
[2349] | 1566 | and tuple are accepted. By default (None), FFT |
---|
[2186] | 1567 | is applied to the whole data. |
---|
| 1568 | |
---|
| 1569 | mask: Auxiliary channel mask(s). Given as a boolean |
---|
| 1570 | list, it is applied to all specified rows. |
---|
| 1571 | A list of boolean lists can also be used to |
---|
| 1572 | apply different masks. In the latter case, the |
---|
| 1573 | length of 'mask' must be the same as that of |
---|
[2349] | 1574 | 'rowno'. The default is None. |
---|
[2177] | 1575 | |
---|
| 1576 | getrealimag: If True, returns the real and imaginary part |
---|
| 1577 | values of the complex results. |
---|
| 1578 | If False (the default), returns the amplitude |
---|
| 1579 | (absolute value) normalised with Ndata/2 and |
---|
| 1580 | phase (argument, in unit of radian). |
---|
| 1581 | |
---|
| 1582 | Returns: |
---|
| 1583 | |
---|
[2186] | 1584 | A list of dictionaries containing the results for each spectrum. |
---|
| 1585 | Each dictionary contains two values, the real and the imaginary |
---|
| 1586 | parts when getrealimag = True, or the amplitude(absolute value) |
---|
| 1587 | and the phase(argument) when getrealimag = False. The key for |
---|
| 1588 | these values are 'real' and 'imag', or 'ampl' and 'phase', |
---|
[2177] | 1589 | respectively. |
---|
| 1590 | """ |
---|
[2349] | 1591 | if rowno is None: |
---|
| 1592 | rowno = [] |
---|
[2177] | 1593 | if isinstance(rowno, int): |
---|
| 1594 | rowno = [rowno] |
---|
| 1595 | elif not (isinstance(rowno, list) or isinstance(rowno, tuple)): |
---|
[2186] | 1596 | raise TypeError("The row number(s) must be int, list or tuple.") |
---|
| 1597 | if len(rowno) == 0: rowno = [i for i in xrange(self.nrow())] |
---|
| 1598 | |
---|
[2411] | 1599 | usecommonmask = True |
---|
| 1600 | |
---|
| 1601 | if mask is None: |
---|
| 1602 | mask = [] |
---|
| 1603 | if isinstance(mask, list) or isinstance(mask, tuple): |
---|
| 1604 | if len(mask) == 0: |
---|
| 1605 | mask = [[]] |
---|
| 1606 | else: |
---|
| 1607 | if isinstance(mask[0], bool): |
---|
| 1608 | if len(mask) != self.nchan(self.getif(rowno[0])): |
---|
| 1609 | raise ValueError("The spectra and the mask have " |
---|
| 1610 | "different length.") |
---|
| 1611 | mask = [mask] |
---|
| 1612 | elif isinstance(mask[0], list) or isinstance(mask[0], tuple): |
---|
| 1613 | usecommonmask = False |
---|
| 1614 | if len(mask) != len(rowno): |
---|
| 1615 | raise ValueError("When specifying masks for each " |
---|
| 1616 | "spectrum, the numbers of them " |
---|
| 1617 | "must be identical.") |
---|
| 1618 | for i in xrange(mask): |
---|
| 1619 | if len(mask[i]) != self.nchan(self.getif(rowno[i])): |
---|
| 1620 | raise ValueError("The spectra and the mask have " |
---|
| 1621 | "different length.") |
---|
| 1622 | else: |
---|
| 1623 | raise TypeError("The mask must be a boolean list or " |
---|
| 1624 | "a list of boolean list.") |
---|
| 1625 | else: |
---|
[2349] | 1626 | raise TypeError("The mask must be a boolean list or a list of " |
---|
| 1627 | "boolean list.") |
---|
[2186] | 1628 | |
---|
| 1629 | res = [] |
---|
| 1630 | |
---|
| 1631 | imask = 0 |
---|
| 1632 | for whichrow in rowno: |
---|
| 1633 | fspec = self._fft(whichrow, mask[imask], getrealimag) |
---|
| 1634 | nspec = len(fspec) |
---|
[2177] | 1635 | |
---|
[2349] | 1636 | i = 0 |
---|
| 1637 | v1 = [] |
---|
| 1638 | v2 = [] |
---|
| 1639 | reselem = {"real":[],"imag":[]} if getrealimag \ |
---|
| 1640 | else {"ampl":[],"phase":[]} |
---|
[2177] | 1641 | |
---|
[2186] | 1642 | while (i < nspec): |
---|
| 1643 | v1.append(fspec[i]) |
---|
| 1644 | v2.append(fspec[i+1]) |
---|
[2349] | 1645 | i += 2 |
---|
[2186] | 1646 | |
---|
[2177] | 1647 | if getrealimag: |
---|
[2186] | 1648 | reselem["real"] += v1 |
---|
| 1649 | reselem["imag"] += v2 |
---|
[2177] | 1650 | else: |
---|
[2186] | 1651 | reselem["ampl"] += v1 |
---|
| 1652 | reselem["phase"] += v2 |
---|
[2177] | 1653 | |
---|
[2186] | 1654 | res.append(reselem) |
---|
| 1655 | |
---|
[2349] | 1656 | if not usecommonmask: |
---|
| 1657 | imask += 1 |
---|
[2186] | 1658 | |
---|
[2177] | 1659 | return res |
---|
| 1660 | |
---|
| 1661 | @asaplog_post_dec |
---|
[113] | 1662 | def create_mask(self, *args, **kwargs): |
---|
[1846] | 1663 | """\ |
---|
[1118] | 1664 | Compute and return a mask based on [min, max] windows. |
---|
[189] | 1665 | The specified windows are to be INCLUDED, when the mask is |
---|
[113] | 1666 | applied. |
---|
[1846] | 1667 | |
---|
[102] | 1668 | Parameters: |
---|
[1846] | 1669 | |
---|
[1118] | 1670 | [min, max], [min2, max2], ... |
---|
[1024] | 1671 | Pairs of start/end points (inclusive)specifying the regions |
---|
[102] | 1672 | to be masked |
---|
[1855] | 1673 | |
---|
[189] | 1674 | invert: optional argument. If specified as True, |
---|
| 1675 | return an inverted mask, i.e. the regions |
---|
| 1676 | specified are EXCLUDED |
---|
[1855] | 1677 | |
---|
[513] | 1678 | row: create the mask using the specified row for |
---|
| 1679 | unit conversions, default is row=0 |
---|
| 1680 | only necessary if frequency varies over rows. |
---|
[1846] | 1681 | |
---|
| 1682 | Examples:: |
---|
| 1683 | |
---|
[113] | 1684 | scan.set_unit('channel') |
---|
[1846] | 1685 | # a) |
---|
[1118] | 1686 | msk = scan.create_mask([400, 500], [800, 900]) |
---|
[189] | 1687 | # masks everything outside 400 and 500 |
---|
[113] | 1688 | # and 800 and 900 in the unit 'channel' |
---|
| 1689 | |
---|
[1846] | 1690 | # b) |
---|
[1118] | 1691 | msk = scan.create_mask([400, 500], [800, 900], invert=True) |
---|
[189] | 1692 | # masks the regions between 400 and 500 |
---|
[113] | 1693 | # and 800 and 900 in the unit 'channel' |
---|
[1846] | 1694 | |
---|
| 1695 | # c) |
---|
| 1696 | #mask only channel 400 |
---|
[1554] | 1697 | msk = scan.create_mask([400]) |
---|
[1846] | 1698 | |
---|
[102] | 1699 | """ |
---|
[1554] | 1700 | row = kwargs.get("row", 0) |
---|
[513] | 1701 | data = self._getabcissa(row) |
---|
[113] | 1702 | u = self._getcoordinfo()[0] |
---|
[1859] | 1703 | if u == "": |
---|
| 1704 | u = "channel" |
---|
| 1705 | msg = "The current mask window unit is %s" % u |
---|
| 1706 | i = self._check_ifs() |
---|
| 1707 | if not i: |
---|
| 1708 | msg += "\nThis mask is only valid for IF=%d" % (self.getif(i)) |
---|
| 1709 | asaplog.push(msg) |
---|
[2348] | 1710 | n = len(data) |
---|
[1295] | 1711 | msk = _n_bools(n, False) |
---|
[710] | 1712 | # test if args is a 'list' or a 'normal *args - UGLY!!! |
---|
| 1713 | |
---|
[2349] | 1714 | ws = (isinstance(args[-1][-1], int) |
---|
| 1715 | or isinstance(args[-1][-1], float)) and args or args[0] |
---|
[710] | 1716 | for window in ws: |
---|
[1554] | 1717 | if len(window) == 1: |
---|
| 1718 | window = [window[0], window[0]] |
---|
| 1719 | if len(window) == 0 or len(window) > 2: |
---|
[2349] | 1720 | raise ValueError("A window needs to be defined as " |
---|
| 1721 | "[start(, end)]") |
---|
[1545] | 1722 | if window[0] > window[1]: |
---|
| 1723 | tmp = window[0] |
---|
| 1724 | window[0] = window[1] |
---|
| 1725 | window[1] = tmp |
---|
[102] | 1726 | for i in range(n): |
---|
[1024] | 1727 | if data[i] >= window[0] and data[i] <= window[1]: |
---|
[1295] | 1728 | msk[i] = True |
---|
[113] | 1729 | if kwargs.has_key('invert'): |
---|
| 1730 | if kwargs.get('invert'): |
---|
[1295] | 1731 | msk = mask_not(msk) |
---|
[102] | 1732 | return msk |
---|
[710] | 1733 | |
---|
[1931] | 1734 | def get_masklist(self, mask=None, row=0, silent=False): |
---|
[1846] | 1735 | """\ |
---|
[1819] | 1736 | Compute and return a list of mask windows, [min, max]. |
---|
[1846] | 1737 | |
---|
[1819] | 1738 | Parameters: |
---|
[1846] | 1739 | |
---|
[1819] | 1740 | mask: channel mask, created with create_mask. |
---|
[1855] | 1741 | |
---|
[1819] | 1742 | row: calcutate the masklist using the specified row |
---|
| 1743 | for unit conversions, default is row=0 |
---|
| 1744 | only necessary if frequency varies over rows. |
---|
[1846] | 1745 | |
---|
[1819] | 1746 | Returns: |
---|
[1846] | 1747 | |
---|
[1819] | 1748 | [min, max], [min2, max2], ... |
---|
| 1749 | Pairs of start/end points (inclusive)specifying |
---|
| 1750 | the masked regions |
---|
[1846] | 1751 | |
---|
[1819] | 1752 | """ |
---|
| 1753 | if not (isinstance(mask,list) or isinstance(mask, tuple)): |
---|
| 1754 | raise TypeError("The mask should be list or tuple.") |
---|
[2427] | 1755 | if len(mask) <= 0: |
---|
| 1756 | raise TypeError("The mask elements should be > 0") |
---|
[2348] | 1757 | data = self._getabcissa(row) |
---|
| 1758 | if len(data) != len(mask): |
---|
[1819] | 1759 | msg = "Number of channels in scantable != number of mask elements" |
---|
| 1760 | raise TypeError(msg) |
---|
| 1761 | u = self._getcoordinfo()[0] |
---|
[1859] | 1762 | if u == "": |
---|
| 1763 | u = "channel" |
---|
| 1764 | msg = "The current mask window unit is %s" % u |
---|
| 1765 | i = self._check_ifs() |
---|
| 1766 | if not i: |
---|
| 1767 | msg += "\nThis mask is only valid for IF=%d" % (self.getif(i)) |
---|
[1931] | 1768 | if not silent: |
---|
| 1769 | asaplog.push(msg) |
---|
[2349] | 1770 | masklist = [] |
---|
[1819] | 1771 | ist, ien = None, None |
---|
| 1772 | ist, ien=self.get_mask_indices(mask) |
---|
| 1773 | if ist is not None and ien is not None: |
---|
| 1774 | for i in xrange(len(ist)): |
---|
| 1775 | range=[data[ist[i]],data[ien[i]]] |
---|
| 1776 | range.sort() |
---|
| 1777 | masklist.append([range[0],range[1]]) |
---|
| 1778 | return masklist |
---|
| 1779 | |
---|
| 1780 | def get_mask_indices(self, mask=None): |
---|
[1846] | 1781 | """\ |
---|
[1819] | 1782 | Compute and Return lists of mask start indices and mask end indices. |
---|
[1855] | 1783 | |
---|
| 1784 | Parameters: |
---|
| 1785 | |
---|
[1819] | 1786 | mask: channel mask, created with create_mask. |
---|
[1846] | 1787 | |
---|
[1819] | 1788 | Returns: |
---|
[1846] | 1789 | |
---|
[1819] | 1790 | List of mask start indices and that of mask end indices, |
---|
| 1791 | i.e., [istart1,istart2,....], [iend1,iend2,....]. |
---|
[1846] | 1792 | |
---|
[1819] | 1793 | """ |
---|
| 1794 | if not (isinstance(mask,list) or isinstance(mask, tuple)): |
---|
| 1795 | raise TypeError("The mask should be list or tuple.") |
---|
[2427] | 1796 | if len(mask) <= 0: |
---|
| 1797 | raise TypeError("The mask elements should be > 0") |
---|
[2349] | 1798 | istart = [] |
---|
| 1799 | iend = [] |
---|
| 1800 | if mask[0]: |
---|
| 1801 | istart.append(0) |
---|
[1819] | 1802 | for i in range(len(mask)-1): |
---|
| 1803 | if not mask[i] and mask[i+1]: |
---|
| 1804 | istart.append(i+1) |
---|
| 1805 | elif mask[i] and not mask[i+1]: |
---|
| 1806 | iend.append(i) |
---|
[2349] | 1807 | if mask[len(mask)-1]: |
---|
| 1808 | iend.append(len(mask)-1) |
---|
[1819] | 1809 | if len(istart) != len(iend): |
---|
| 1810 | raise RuntimeError("Numbers of mask start != mask end.") |
---|
| 1811 | for i in range(len(istart)): |
---|
| 1812 | if istart[i] > iend[i]: |
---|
| 1813 | raise RuntimeError("Mask start index > mask end index") |
---|
| 1814 | break |
---|
| 1815 | return istart,iend |
---|
| 1816 | |
---|
[2013] | 1817 | @asaplog_post_dec |
---|
[2882] | 1818 | def parse_spw_selection(self, selectstring, restfreq=None, frame=None, doppler=None): |
---|
| 1819 | """ |
---|
| 1820 | Parse MS type spw/channel selection syntax. |
---|
| 1821 | |
---|
| 1822 | Parameters: |
---|
| 1823 | selectstring : A string expression of spw and channel selection. |
---|
| 1824 | Comma-separated expressions mean different spw - |
---|
| 1825 | channel combinations. Spws and channel selections |
---|
| 1826 | are partitioned by a colon ':'. In a single |
---|
| 1827 | selection expression, you can put multiple values |
---|
| 1828 | separated by semicolons ';'. Both for spw and |
---|
| 1829 | channel selection, allowed cases include single |
---|
| 1830 | value, blank('') or asterisk('*') to specify all |
---|
| 1831 | available values, two values connected with a |
---|
| 1832 | tilde ('~') to specify an inclusive range. Unit |
---|
| 1833 | strings for frequency or velocity can be added to |
---|
| 1834 | the tilde-connected values. For channel selection |
---|
| 1835 | expression, placing a '<' or a '>' is possible to |
---|
| 1836 | specify a semi-infinite interval as well. |
---|
| 1837 | |
---|
| 1838 | examples: |
---|
| 1839 | '' or '*' = all spws (all channels) |
---|
| 1840 | '<2,4~6,9' = Spws 0,1,4,5,6,9 (all channels) |
---|
| 1841 | '3:3~45;60' = channels 3 to 45 and 60 in spw 3 |
---|
| 1842 | '0~1:2~6,8' = channels 2 to 6 in spws 0,1, and |
---|
| 1843 | all channels in spw8 |
---|
[2936] | 1844 | '1.3~1.5GHz' = all spws whose central frequency |
---|
| 1845 | falls in frequency range between |
---|
| 1846 | 1.3GHz and 1.5GHz. |
---|
| 1847 | '1.3~1.5GHz:1.3~1.5GHz' = channels which fall |
---|
[2884] | 1848 | between the specified |
---|
| 1849 | frequency range in spws |
---|
[2936] | 1850 | whose central frequency |
---|
| 1851 | falls in the specified |
---|
| 1852 | frequency range. |
---|
[2884] | 1853 | '1:-200~250km/s' = channels that fall between the |
---|
| 1854 | specified velocity range in |
---|
| 1855 | spw 1. |
---|
[2897] | 1856 | restfreq: the rest frequency. |
---|
| 1857 | examples: '115.2712GHz', 115271201800.0 |
---|
| 1858 | frame: an optional frame type, default 'LSRK'. Valid frames are: |
---|
| 1859 | 'TOPO', 'LSRD', 'LSRK', 'BARY', |
---|
| 1860 | 'GEO', 'GALACTO', 'LGROUP', 'CMB' |
---|
| 1861 | doppler: one of 'RADIO', 'OPTICAL', 'Z', 'BETA', 'GAMMA' |
---|
[2882] | 1862 | Returns: |
---|
| 1863 | A dictionary of selected (valid) spw and masklist pairs, |
---|
| 1864 | e.g. {'0': [[50,250],[350,462]], '2': [[100,400],[550,974]]} |
---|
| 1865 | """ |
---|
| 1866 | if not isinstance(selectstring, str): |
---|
| 1867 | asaplog.post() |
---|
| 1868 | asaplog.push("Expression of spw/channel selection must be a string.") |
---|
| 1869 | asaplog.post("ERROR") |
---|
| 1870 | |
---|
| 1871 | orig_unit = self.get_unit() |
---|
| 1872 | self.set_unit('channel') |
---|
| 1873 | |
---|
[2891] | 1874 | if restfreq is not None: |
---|
[2892] | 1875 | orig_molids = self._getmolidcol_list() |
---|
| 1876 | set_restfreq(self, restfreq) |
---|
[2882] | 1877 | |
---|
[2897] | 1878 | orig_coord = self._getcoordinfo() |
---|
[2892] | 1879 | |
---|
| 1880 | if frame is not None: |
---|
| 1881 | orig_frame = orig_coord[1] |
---|
| 1882 | self.set_freqframe(frame) |
---|
| 1883 | |
---|
| 1884 | if doppler is not None: |
---|
| 1885 | orig_doppler = orig_coord[2] |
---|
| 1886 | self.set_doppler(doppler) |
---|
[2882] | 1887 | |
---|
| 1888 | valid_ifs = self.getifnos() |
---|
| 1889 | |
---|
| 1890 | comma_sep = selectstring.split(",") |
---|
| 1891 | res = {} |
---|
| 1892 | |
---|
| 1893 | for cms_elem in comma_sep: |
---|
| 1894 | colon_sep = cms_elem.split(":") |
---|
| 1895 | |
---|
| 1896 | if (len(colon_sep) > 2): |
---|
| 1897 | raise RuntimeError("Invalid selection expression: more than two colons!") |
---|
| 1898 | |
---|
| 1899 | # parse spw expression and store result in spw_list. |
---|
| 1900 | # allowed cases include '', '*', 'a', '<a', '>a', 'a~b', |
---|
[2884] | 1901 | # 'a~b*Hz' (where * can be '', 'k', 'M', 'G' etc.), |
---|
| 1902 | # 'a~b*m/s' (where * can be '' or 'k') and also |
---|
[2882] | 1903 | # several of the above expressions connected with ';'. |
---|
| 1904 | |
---|
| 1905 | spw_list = [] |
---|
| 1906 | |
---|
| 1907 | semicolon_sep = colon_sep[0].split(";") |
---|
| 1908 | |
---|
| 1909 | for scs_elem in semicolon_sep: |
---|
| 1910 | scs_elem = scs_elem.strip() |
---|
| 1911 | |
---|
| 1912 | lt_sep = scs_elem.split("<") |
---|
| 1913 | gt_sep = scs_elem.split(">") |
---|
| 1914 | ti_sep = scs_elem.split("~") |
---|
| 1915 | |
---|
| 1916 | lt_sep_length = len(lt_sep) |
---|
| 1917 | gt_sep_length = len(gt_sep) |
---|
| 1918 | ti_sep_length = len(ti_sep) |
---|
| 1919 | |
---|
| 1920 | len_product = lt_sep_length * gt_sep_length * ti_sep_length |
---|
| 1921 | |
---|
| 1922 | if (len_product > 2): |
---|
| 1923 | # '<', '>' and '~' must not coexist in a single spw expression |
---|
| 1924 | |
---|
| 1925 | raise RuntimeError("Invalid spw selection.") |
---|
| 1926 | |
---|
| 1927 | elif (len_product == 1): |
---|
| 1928 | # '', '*', or single spw number. |
---|
| 1929 | |
---|
| 1930 | if (scs_elem == "") or (scs_elem == "*"): |
---|
| 1931 | spw_list = valid_ifs[:] # deep copy |
---|
| 1932 | |
---|
| 1933 | else: # single number |
---|
[2887] | 1934 | expr = int(scs_elem) |
---|
| 1935 | spw_list.append(expr) |
---|
| 1936 | if expr not in valid_ifs: |
---|
| 1937 | asaplog.push("Invalid spw given. Ignored.") |
---|
| 1938 | |
---|
[2882] | 1939 | else: # (len_product == 2) |
---|
[2887] | 1940 | # namely, one of '<', '>' or '~' appears just once. |
---|
[2882] | 1941 | |
---|
| 1942 | if (lt_sep_length == 2): # '<a' |
---|
| 1943 | if is_number(lt_sep[1]): |
---|
[2886] | 1944 | no_valid_spw = True |
---|
[2882] | 1945 | for i in valid_ifs: |
---|
| 1946 | if (i < float(lt_sep[1])): |
---|
| 1947 | spw_list.append(i) |
---|
[2886] | 1948 | no_valid_spw = False |
---|
| 1949 | |
---|
| 1950 | if no_valid_spw: |
---|
| 1951 | raise ValueError("Invalid spw selection ('<" + str(lt_sep[1]) + "').") |
---|
[2882] | 1952 | |
---|
| 1953 | else: |
---|
[2886] | 1954 | raise RuntimeError("Invalid spw selection.") |
---|
[2882] | 1955 | |
---|
| 1956 | elif (gt_sep_length == 2): # '>a' |
---|
| 1957 | if is_number(gt_sep[1]): |
---|
[2886] | 1958 | no_valid_spw = True |
---|
[2882] | 1959 | for i in valid_ifs: |
---|
| 1960 | if (i > float(gt_sep[1])): |
---|
| 1961 | spw_list.append(i) |
---|
[2886] | 1962 | no_valid_spw = False |
---|
| 1963 | |
---|
| 1964 | if no_valid_spw: |
---|
| 1965 | raise ValueError("Invalid spw selection ('>" + str(gt_sep[1]) + "').") |
---|
[2882] | 1966 | |
---|
| 1967 | else: |
---|
[2886] | 1968 | raise RuntimeError("Invalid spw selection.") |
---|
[2882] | 1969 | |
---|
| 1970 | else: # (ti_sep_length == 2) where both boundaries inclusive |
---|
| 1971 | expr0 = ti_sep[0].strip() |
---|
| 1972 | expr1 = ti_sep[1].strip() |
---|
| 1973 | |
---|
| 1974 | if is_number(expr0) and is_number(expr1): |
---|
| 1975 | # 'a~b' |
---|
| 1976 | expr_pmin = min(float(expr0), float(expr1)) |
---|
| 1977 | expr_pmax = max(float(expr0), float(expr1)) |
---|
[2887] | 1978 | has_invalid_spw = False |
---|
[2886] | 1979 | no_valid_spw = True |
---|
| 1980 | |
---|
[2882] | 1981 | for i in valid_ifs: |
---|
| 1982 | if (expr_pmin <= i) and (i <= expr_pmax): |
---|
| 1983 | spw_list.append(i) |
---|
[2886] | 1984 | no_valid_spw = False |
---|
[2887] | 1985 | else: |
---|
| 1986 | has_invalid_spw = True |
---|
[2886] | 1987 | |
---|
[2887] | 1988 | if has_invalid_spw: |
---|
| 1989 | msg = "Invalid spw is given. Ignored." |
---|
| 1990 | asaplog.push(msg) |
---|
| 1991 | asaplog.post() |
---|
| 1992 | |
---|
[2886] | 1993 | if no_valid_spw: |
---|
| 1994 | raise ValueError("No valid spw in range ('" + str(expr_pmin) + "~" + str(expr_pmax) + "').") |
---|
[2887] | 1995 | |
---|
[2884] | 1996 | elif is_number(expr0) and is_frequency(expr1): |
---|
| 1997 | # 'a~b*Hz' |
---|
| 1998 | (expr_f0, expr_f1) = get_freq_by_string(expr0, expr1) |
---|
[2936] | 1999 | expr_fmin = min(expr_f0, expr_f1) |
---|
| 2000 | expr_fmax = max(expr_f0, expr_f1) |
---|
[2886] | 2001 | no_valid_spw = True |
---|
| 2002 | |
---|
[2882] | 2003 | for coord in self._get_coordinate_list(): |
---|
[2936] | 2004 | spw = coord['if'] |
---|
| 2005 | |
---|
| 2006 | """ |
---|
[2882] | 2007 | expr_p0 = coord['coord'].to_pixel(expr_f0) |
---|
| 2008 | expr_p1 = coord['coord'].to_pixel(expr_f1) |
---|
| 2009 | expr_pmin = min(expr_p0, expr_p1) |
---|
| 2010 | expr_pmax = max(expr_p0, expr_p1) |
---|
| 2011 | |
---|
| 2012 | pmin = 0.0 |
---|
| 2013 | pmax = float(self.nchan(spw) - 1) |
---|
[2936] | 2014 | |
---|
[2882] | 2015 | if ((expr_pmax - pmin)*(expr_pmin - pmax) <= 0.0): |
---|
| 2016 | spw_list.append(spw) |
---|
[2886] | 2017 | no_valid_spw = False |
---|
[2936] | 2018 | """ |
---|
[2886] | 2019 | |
---|
[2937] | 2020 | crd = coord['coord'] |
---|
| 2021 | fhead = crd.to_frequency(0) |
---|
| 2022 | ftail = crd.to_frequency(self.nchan(spw) - 1) |
---|
[2936] | 2023 | fcen = (fhead + ftail) / 2.0 |
---|
| 2024 | |
---|
| 2025 | if ((expr_fmin <= fcen) and (fcen <= expr_fmax)): |
---|
| 2026 | spw_list.append(spw) |
---|
| 2027 | no_valid_spw = False |
---|
| 2028 | |
---|
[2886] | 2029 | if no_valid_spw: |
---|
| 2030 | raise ValueError("No valid spw in range ('" + str(expr0) + "~" + str(expr1) + "').") |
---|
[2882] | 2031 | |
---|
[2884] | 2032 | elif is_number(expr0) and is_velocity(expr1): |
---|
| 2033 | # 'a~b*m/s' |
---|
| 2034 | (expr_v0, expr_v1) = get_velocity_by_string(expr0, expr1) |
---|
[2882] | 2035 | expr_vmin = min(expr_v0, expr_v1) |
---|
| 2036 | expr_vmax = max(expr_v0, expr_v1) |
---|
[2886] | 2037 | no_valid_spw = True |
---|
| 2038 | |
---|
[2882] | 2039 | for coord in self._get_coordinate_list(): |
---|
| 2040 | spw = coord['if'] |
---|
[2936] | 2041 | |
---|
| 2042 | """ |
---|
[2882] | 2043 | pmin = 0.0 |
---|
| 2044 | pmax = float(self.nchan(spw) - 1) |
---|
| 2045 | |
---|
| 2046 | vel0 = coord['coord'].to_velocity(pmin) |
---|
| 2047 | vel1 = coord['coord'].to_velocity(pmax) |
---|
| 2048 | |
---|
| 2049 | vmin = min(vel0, vel1) |
---|
| 2050 | vmax = max(vel0, vel1) |
---|
| 2051 | |
---|
| 2052 | if ((expr_vmax - vmin)*(expr_vmin - vmax) <= 0.0): |
---|
| 2053 | spw_list.append(spw) |
---|
[2886] | 2054 | no_valid_spw = False |
---|
[2936] | 2055 | """ |
---|
[2886] | 2056 | |
---|
[2937] | 2057 | crd = coord['coord'] |
---|
| 2058 | fhead = crd.to_frequency(0) |
---|
| 2059 | ftail = crd.to_frequency(self.nchan(spw) - 1) |
---|
| 2060 | fcen = (fhead + ftail) / 2.0 |
---|
| 2061 | vcen = crd.to_velocity(crd.to_pixel(fcen)) |
---|
[2936] | 2062 | |
---|
| 2063 | if ((expr_vmin <= vcen) and (vcen <= expr_vmax)): |
---|
| 2064 | spw_list.append(spw) |
---|
| 2065 | no_valid_spw = False |
---|
| 2066 | |
---|
[2886] | 2067 | if no_valid_spw: |
---|
| 2068 | raise ValueError("No valid spw in range ('" + str(expr0) + "~" + str(expr1) + "').") |
---|
[2882] | 2069 | |
---|
| 2070 | else: |
---|
| 2071 | # cases such as 'aGHz~bkm/s' are not allowed now |
---|
| 2072 | raise RuntimeError("Invalid spw selection.") |
---|
| 2073 | |
---|
[2887] | 2074 | # check spw list and remove invalid ones. |
---|
| 2075 | # if no valid spw left, emit ValueError. |
---|
| 2076 | if len(spw_list) == 0: |
---|
| 2077 | raise ValueError("No valid spw in given range.") |
---|
| 2078 | |
---|
[2882] | 2079 | # parse channel expression and store the result in crange_list. |
---|
| 2080 | # allowed cases include '', 'a~b', 'a*Hz~b*Hz' (where * can be |
---|
| 2081 | # '', 'k', 'M', 'G' etc.), 'a*m/s~b*m/s' (where * can be '' or 'k') |
---|
| 2082 | # and also several of the above expressions connected with ';'. |
---|
| 2083 | |
---|
| 2084 | for spw in spw_list: |
---|
| 2085 | pmin = 0.0 |
---|
| 2086 | pmax = float(self.nchan(spw) - 1) |
---|
[2909] | 2087 | |
---|
| 2088 | molid = self._getmolidcol_list()[self.get_first_rowno_by_if(spw)] |
---|
[2882] | 2089 | |
---|
| 2090 | if (len(colon_sep) == 1): |
---|
| 2091 | # no expression for channel selection, |
---|
| 2092 | # which means all channels are to be selected. |
---|
| 2093 | crange_list = [[pmin, pmax]] |
---|
| 2094 | |
---|
| 2095 | else: # (len(colon_sep) == 2) |
---|
| 2096 | crange_list = [] |
---|
| 2097 | |
---|
| 2098 | found = False |
---|
| 2099 | for i in self._get_coordinate_list(): |
---|
| 2100 | if (i['if'] == spw): |
---|
| 2101 | coord = i['coord'] |
---|
| 2102 | found = True |
---|
| 2103 | break |
---|
| 2104 | |
---|
[2887] | 2105 | if found: |
---|
| 2106 | semicolon_sep = colon_sep[1].split(";") |
---|
| 2107 | for scs_elem in semicolon_sep: |
---|
| 2108 | scs_elem = scs_elem.strip() |
---|
[2882] | 2109 | |
---|
[2887] | 2110 | ti_sep = scs_elem.split("~") |
---|
| 2111 | ti_sep_length = len(ti_sep) |
---|
[2882] | 2112 | |
---|
[2887] | 2113 | if (ti_sep_length > 2): |
---|
| 2114 | raise RuntimeError("Invalid channel selection.") |
---|
[2882] | 2115 | |
---|
[2887] | 2116 | elif (ti_sep_length == 1): |
---|
| 2117 | if (scs_elem == "") or (scs_elem == "*"): |
---|
| 2118 | # '' and '*' for all channels |
---|
| 2119 | crange_list = [[pmin, pmax]] |
---|
| 2120 | break |
---|
| 2121 | elif (is_number(scs_elem)): |
---|
| 2122 | # single channel given |
---|
| 2123 | crange_list.append([float(scs_elem), float(scs_elem)]) |
---|
| 2124 | else: |
---|
| 2125 | raise RuntimeError("Invalid channel selection.") |
---|
[2882] | 2126 | |
---|
[2887] | 2127 | else: #(ti_sep_length == 2) |
---|
| 2128 | expr0 = ti_sep[0].strip() |
---|
| 2129 | expr1 = ti_sep[1].strip() |
---|
[2882] | 2130 | |
---|
[2887] | 2131 | if is_number(expr0) and is_number(expr1): |
---|
| 2132 | # 'a~b' |
---|
| 2133 | expr_pmin = min(float(expr0), float(expr1)) |
---|
| 2134 | expr_pmax = max(float(expr0), float(expr1)) |
---|
[2882] | 2135 | |
---|
[2887] | 2136 | elif is_number(expr0) and is_frequency(expr1): |
---|
| 2137 | # 'a~b*Hz' |
---|
| 2138 | (expr_f0, expr_f1) = get_freq_by_string(expr0, expr1) |
---|
| 2139 | expr_p0 = coord.to_pixel(expr_f0) |
---|
| 2140 | expr_p1 = coord.to_pixel(expr_f1) |
---|
| 2141 | expr_pmin = min(expr_p0, expr_p1) |
---|
| 2142 | expr_pmax = max(expr_p0, expr_p1) |
---|
[2882] | 2143 | |
---|
[2887] | 2144 | elif is_number(expr0) and is_velocity(expr1): |
---|
| 2145 | # 'a~b*m/s' |
---|
[2909] | 2146 | restf = self.get_restfreqs()[molid][0] |
---|
[2887] | 2147 | (expr_v0, expr_v1) = get_velocity_by_string(expr0, expr1) |
---|
[2897] | 2148 | dppl = self.get_doppler() |
---|
| 2149 | expr_f0 = get_frequency_by_velocity(restf, expr_v0, dppl) |
---|
| 2150 | expr_f1 = get_frequency_by_velocity(restf, expr_v1, dppl) |
---|
[2887] | 2151 | expr_p0 = coord.to_pixel(expr_f0) |
---|
| 2152 | expr_p1 = coord.to_pixel(expr_f1) |
---|
| 2153 | expr_pmin = min(expr_p0, expr_p1) |
---|
| 2154 | expr_pmax = max(expr_p0, expr_p1) |
---|
[2882] | 2155 | |
---|
[2887] | 2156 | else: |
---|
| 2157 | # cases such as 'aGHz~bkm/s' are not allowed now |
---|
| 2158 | raise RuntimeError("Invalid channel selection.") |
---|
[2882] | 2159 | |
---|
[2887] | 2160 | cmin = max(pmin, expr_pmin) |
---|
| 2161 | cmax = min(pmax, expr_pmax) |
---|
| 2162 | # if the given range of channel selection has overwrap with |
---|
| 2163 | # that of current spw, output the overwrap area. |
---|
| 2164 | if (cmin <= cmax): |
---|
| 2165 | cmin = float(int(cmin + 0.5)) |
---|
| 2166 | cmax = float(int(cmax + 0.5)) |
---|
| 2167 | crange_list.append([cmin, cmax]) |
---|
[2882] | 2168 | |
---|
| 2169 | if (len(crange_list) == 0): |
---|
| 2170 | crange_list.append([]) |
---|
| 2171 | |
---|
[2910] | 2172 | if (len(crange_list[0]) > 0): |
---|
| 2173 | if res.has_key(spw): |
---|
| 2174 | res[spw].extend(crange_list) |
---|
| 2175 | else: |
---|
| 2176 | res[spw] = crange_list |
---|
[2882] | 2177 | |
---|
[2887] | 2178 | for spw in res.keys(): |
---|
[2932] | 2179 | if spw in valid_ifs: |
---|
[2933] | 2180 | # remove duplicated channel ranges |
---|
[2932] | 2181 | for i in reversed(xrange(len(res[spw]))): |
---|
| 2182 | for j in xrange(i): |
---|
[2933] | 2183 | if ((res[spw][i][0]-res[spw][j][1])*(res[spw][i][1]-res[spw][j][0]) <= 0) or \ |
---|
| 2184 | (min(abs(res[spw][i][0]-res[spw][j][1]),abs(res[spw][j][0]-res[spw][i][1])) == 1): |
---|
[2935] | 2185 | asaplog.post() |
---|
| 2186 | merge_warn_mesg = "Spw " + str(spw) + ": overwrapping channel ranges are merged." |
---|
| 2187 | asaplog.push(merge_warn_mesg) |
---|
| 2188 | asaplog.post('WARN') |
---|
[2932] | 2189 | res[spw][j][0] = min(res[spw][i][0], res[spw][j][0]) |
---|
| 2190 | res[spw][j][1] = max(res[spw][i][1], res[spw][j][1]) |
---|
| 2191 | res[spw].pop(i) |
---|
| 2192 | break |
---|
| 2193 | else: |
---|
[2887] | 2194 | del res[spw] |
---|
| 2195 | |
---|
| 2196 | if len(res) == 0: |
---|
| 2197 | raise RuntimeError("No valid spw.") |
---|
| 2198 | |
---|
[2882] | 2199 | # restore original values |
---|
[2892] | 2200 | self.set_unit(orig_unit) |
---|
[2891] | 2201 | if restfreq is not None: |
---|
[2892] | 2202 | self._setmolidcol_list(orig_molids) |
---|
| 2203 | if frame is not None: |
---|
| 2204 | self.set_freqframe(orig_frame) |
---|
| 2205 | if doppler is not None: |
---|
| 2206 | self.set_doppler(orig_doppler) |
---|
[2882] | 2207 | |
---|
| 2208 | return res |
---|
[2890] | 2209 | |
---|
[2882] | 2210 | @asaplog_post_dec |
---|
| 2211 | def get_first_rowno_by_if(self, ifno): |
---|
| 2212 | found = False |
---|
| 2213 | for irow in xrange(self.nrow()): |
---|
| 2214 | if (self.getif(irow) == ifno): |
---|
| 2215 | res = irow |
---|
| 2216 | found = True |
---|
| 2217 | break |
---|
| 2218 | |
---|
[2926] | 2219 | if not found: raise RuntimeError("No valid spw.") |
---|
[2882] | 2220 | |
---|
| 2221 | return res |
---|
| 2222 | |
---|
| 2223 | @asaplog_post_dec |
---|
| 2224 | def _get_coordinate_list(self): |
---|
| 2225 | res = [] |
---|
| 2226 | spws = self.getifnos() |
---|
| 2227 | for spw in spws: |
---|
| 2228 | elem = {} |
---|
| 2229 | elem['if'] = spw |
---|
| 2230 | elem['coord'] = self.get_coordinate(self.get_first_rowno_by_if(spw)) |
---|
| 2231 | res.append(elem) |
---|
| 2232 | |
---|
| 2233 | return res |
---|
| 2234 | |
---|
| 2235 | @asaplog_post_dec |
---|
[2349] | 2236 | def parse_maskexpr(self, maskstring): |
---|
[2013] | 2237 | """ |
---|
| 2238 | Parse CASA type mask selection syntax (IF dependent). |
---|
| 2239 | |
---|
| 2240 | Parameters: |
---|
| 2241 | maskstring : A string mask selection expression. |
---|
| 2242 | A comma separated selections mean different IF - |
---|
| 2243 | channel combinations. IFs and channel selections |
---|
| 2244 | are partitioned by a colon, ':'. |
---|
| 2245 | examples: |
---|
[2015] | 2246 | '' = all IFs (all channels) |
---|
[2013] | 2247 | '<2,4~6,9' = IFs 0,1,4,5,6,9 (all channels) |
---|
| 2248 | '3:3~45;60' = channels 3 to 45 and 60 in IF 3 |
---|
| 2249 | '0~1:2~6,8' = channels 2 to 6 in IFs 0,1, and |
---|
| 2250 | all channels in IF8 |
---|
| 2251 | Returns: |
---|
| 2252 | A dictionary of selected (valid) IF and masklist pairs, |
---|
| 2253 | e.g. {'0': [[50,250],[350,462]], '2': [[100,400],[550,974]]} |
---|
| 2254 | """ |
---|
| 2255 | if not isinstance(maskstring,str): |
---|
| 2256 | asaplog.post() |
---|
[2611] | 2257 | asaplog.push("Mask expression should be a string.") |
---|
[2013] | 2258 | asaplog.post("ERROR") |
---|
| 2259 | |
---|
| 2260 | valid_ifs = self.getifnos() |
---|
| 2261 | frequnit = self.get_unit() |
---|
| 2262 | seldict = {} |
---|
[2015] | 2263 | if maskstring == "": |
---|
| 2264 | maskstring = str(valid_ifs)[1:-1] |
---|
[2611] | 2265 | ## split each selection "IF range[:CHAN range]" |
---|
[2867] | 2266 | # split maskstring by "<spaces>,<spaces>" |
---|
| 2267 | comma_sep = re.compile('\s*,\s*') |
---|
| 2268 | sellist = comma_sep.split(maskstring) |
---|
| 2269 | # separator by "<spaces>:<spaces>" |
---|
| 2270 | collon_sep = re.compile('\s*:\s*') |
---|
[2013] | 2271 | for currselstr in sellist: |
---|
[2867] | 2272 | selset = collon_sep.split(currselstr) |
---|
[2013] | 2273 | # spw and mask string (may include ~, < or >) |
---|
[2349] | 2274 | spwmasklist = self._parse_selection(selset[0], typestr='integer', |
---|
[2611] | 2275 | minval=min(valid_ifs), |
---|
[2349] | 2276 | maxval=max(valid_ifs)) |
---|
[2013] | 2277 | for spwlist in spwmasklist: |
---|
| 2278 | selspws = [] |
---|
| 2279 | for ispw in range(spwlist[0],spwlist[1]+1): |
---|
| 2280 | # Put into the list only if ispw exists |
---|
| 2281 | if valid_ifs.count(ispw): |
---|
| 2282 | selspws.append(ispw) |
---|
| 2283 | del spwmasklist, spwlist |
---|
| 2284 | |
---|
| 2285 | # parse frequency mask list |
---|
| 2286 | if len(selset) > 1: |
---|
[2349] | 2287 | freqmasklist = self._parse_selection(selset[1], typestr='float', |
---|
| 2288 | offset=0.) |
---|
[2013] | 2289 | else: |
---|
| 2290 | # want to select the whole spectrum |
---|
| 2291 | freqmasklist = [None] |
---|
| 2292 | |
---|
| 2293 | ## define a dictionary of spw - masklist combination |
---|
| 2294 | for ispw in selspws: |
---|
| 2295 | #print "working on", ispw |
---|
| 2296 | spwstr = str(ispw) |
---|
| 2297 | if len(selspws) == 0: |
---|
| 2298 | # empty spw |
---|
| 2299 | continue |
---|
| 2300 | else: |
---|
| 2301 | ## want to get min and max of the spw and |
---|
| 2302 | ## offset to set for '<' and '>' |
---|
| 2303 | if frequnit == 'channel': |
---|
| 2304 | minfreq = 0 |
---|
| 2305 | maxfreq = self.nchan(ifno=ispw) |
---|
| 2306 | offset = 0.5 |
---|
| 2307 | else: |
---|
| 2308 | ## This is ugly part. need improvement |
---|
| 2309 | for ifrow in xrange(self.nrow()): |
---|
| 2310 | if self.getif(ifrow) == ispw: |
---|
| 2311 | #print "IF",ispw,"found in row =",ifrow |
---|
| 2312 | break |
---|
| 2313 | freqcoord = self.get_coordinate(ifrow) |
---|
| 2314 | freqs = self._getabcissa(ifrow) |
---|
| 2315 | minfreq = min(freqs) |
---|
| 2316 | maxfreq = max(freqs) |
---|
| 2317 | if len(freqs) == 1: |
---|
| 2318 | offset = 0.5 |
---|
| 2319 | elif frequnit.find('Hz') > 0: |
---|
[2349] | 2320 | offset = abs(freqcoord.to_frequency(1, |
---|
| 2321 | unit=frequnit) |
---|
| 2322 | -freqcoord.to_frequency(0, |
---|
| 2323 | unit=frequnit) |
---|
| 2324 | )*0.5 |
---|
[2013] | 2325 | elif frequnit.find('m/s') > 0: |
---|
[2349] | 2326 | offset = abs(freqcoord.to_velocity(1, |
---|
| 2327 | unit=frequnit) |
---|
| 2328 | -freqcoord.to_velocity(0, |
---|
| 2329 | unit=frequnit) |
---|
| 2330 | )*0.5 |
---|
[2013] | 2331 | else: |
---|
| 2332 | asaplog.post() |
---|
| 2333 | asaplog.push("Invalid frequency unit") |
---|
| 2334 | asaplog.post("ERROR") |
---|
| 2335 | del freqs, freqcoord, ifrow |
---|
| 2336 | for freq in freqmasklist: |
---|
| 2337 | selmask = freq or [minfreq, maxfreq] |
---|
| 2338 | if selmask[0] == None: |
---|
| 2339 | ## selection was "<freq[1]". |
---|
| 2340 | if selmask[1] < minfreq: |
---|
| 2341 | ## avoid adding region selection |
---|
| 2342 | selmask = None |
---|
| 2343 | else: |
---|
| 2344 | selmask = [minfreq,selmask[1]-offset] |
---|
| 2345 | elif selmask[1] == None: |
---|
| 2346 | ## selection was ">freq[0]" |
---|
| 2347 | if selmask[0] > maxfreq: |
---|
| 2348 | ## avoid adding region selection |
---|
| 2349 | selmask = None |
---|
| 2350 | else: |
---|
| 2351 | selmask = [selmask[0]+offset,maxfreq] |
---|
| 2352 | if selmask: |
---|
| 2353 | if not seldict.has_key(spwstr): |
---|
| 2354 | # new spw selection |
---|
| 2355 | seldict[spwstr] = [] |
---|
| 2356 | seldict[spwstr] += [selmask] |
---|
| 2357 | del minfreq,maxfreq,offset,freq,selmask |
---|
| 2358 | del spwstr |
---|
| 2359 | del freqmasklist |
---|
| 2360 | del valid_ifs |
---|
| 2361 | if len(seldict) == 0: |
---|
| 2362 | asaplog.post() |
---|
[2349] | 2363 | asaplog.push("No valid selection in the mask expression: " |
---|
| 2364 | +maskstring) |
---|
[2013] | 2365 | asaplog.post("WARN") |
---|
| 2366 | return None |
---|
| 2367 | msg = "Selected masklist:\n" |
---|
| 2368 | for sif, lmask in seldict.iteritems(): |
---|
| 2369 | msg += " IF"+sif+" - "+str(lmask)+"\n" |
---|
| 2370 | asaplog.push(msg) |
---|
| 2371 | return seldict |
---|
| 2372 | |
---|
[2611] | 2373 | @asaplog_post_dec |
---|
| 2374 | def parse_idx_selection(self, mode, selexpr): |
---|
| 2375 | """ |
---|
| 2376 | Parse CASA type mask selection syntax of SCANNO, IFNO, POLNO, |
---|
| 2377 | BEAMNO, and row number |
---|
| 2378 | |
---|
| 2379 | Parameters: |
---|
| 2380 | mode : which column to select. |
---|
| 2381 | ['scan',|'if'|'pol'|'beam'|'row'] |
---|
| 2382 | selexpr : A comma separated selection expression. |
---|
| 2383 | examples: |
---|
| 2384 | '' = all (returns []) |
---|
| 2385 | '<2,4~6,9' = indices less than 2, 4 to 6 and 9 |
---|
| 2386 | (returns [0,1,4,5,6,9]) |
---|
| 2387 | Returns: |
---|
| 2388 | A List of selected indices |
---|
| 2389 | """ |
---|
| 2390 | if selexpr == "": |
---|
| 2391 | return [] |
---|
| 2392 | valid_modes = {'s': 'scan', 'i': 'if', 'p': 'pol', |
---|
| 2393 | 'b': 'beam', 'r': 'row'} |
---|
| 2394 | smode = mode.lower()[0] |
---|
| 2395 | if not (smode in valid_modes.keys()): |
---|
| 2396 | msg = "Invalid mode '%s'. Valid modes are %s" %\ |
---|
| 2397 | (mode, str(valid_modes.values())) |
---|
| 2398 | asaplog.post() |
---|
| 2399 | asaplog.push(msg) |
---|
| 2400 | asaplog.post("ERROR") |
---|
| 2401 | mode = valid_modes[smode] |
---|
| 2402 | minidx = None |
---|
| 2403 | maxidx = None |
---|
| 2404 | if smode == 'r': |
---|
| 2405 | minidx = 0 |
---|
| 2406 | maxidx = self.nrow() |
---|
| 2407 | else: |
---|
| 2408 | idx = getattr(self,"get"+mode+"nos")() |
---|
| 2409 | minidx = min(idx) |
---|
| 2410 | maxidx = max(idx) |
---|
| 2411 | del idx |
---|
[2867] | 2412 | # split selexpr by "<spaces>,<spaces>" |
---|
| 2413 | comma_sep = re.compile('\s*,\s*') |
---|
| 2414 | sellist = comma_sep.split(selexpr) |
---|
[2611] | 2415 | idxlist = [] |
---|
| 2416 | for currselstr in sellist: |
---|
| 2417 | # single range (may include ~, < or >) |
---|
| 2418 | currlist = self._parse_selection(currselstr, typestr='integer', |
---|
| 2419 | minval=minidx,maxval=maxidx) |
---|
| 2420 | for thelist in currlist: |
---|
| 2421 | idxlist += range(thelist[0],thelist[1]+1) |
---|
[2932] | 2422 | # remove duplicated elements after first ones |
---|
| 2423 | for i in reversed(xrange(len(idxlist))): |
---|
| 2424 | if idxlist.index(idxlist[i]) < i: |
---|
| 2425 | idxlist.pop(i) |
---|
[2611] | 2426 | msg = "Selected %s: %s" % (mode.upper()+"NO", str(idxlist)) |
---|
| 2427 | asaplog.push(msg) |
---|
| 2428 | return idxlist |
---|
| 2429 | |
---|
[2349] | 2430 | def _parse_selection(self, selstr, typestr='float', offset=0., |
---|
[2351] | 2431 | minval=None, maxval=None): |
---|
[2013] | 2432 | """ |
---|
| 2433 | Parameters: |
---|
| 2434 | selstr : The Selection string, e.g., '<3;5~7;100~103;9' |
---|
| 2435 | typestr : The type of the values in returned list |
---|
| 2436 | ('integer' or 'float') |
---|
| 2437 | offset : The offset value to subtract from or add to |
---|
| 2438 | the boundary value if the selection string |
---|
[2611] | 2439 | includes '<' or '>' [Valid only for typestr='float'] |
---|
[2013] | 2440 | minval, maxval : The minimum/maximum values to set if the |
---|
| 2441 | selection string includes '<' or '>'. |
---|
| 2442 | The list element is filled with None by default. |
---|
| 2443 | Returns: |
---|
| 2444 | A list of min/max pair of selections. |
---|
| 2445 | Example: |
---|
[2611] | 2446 | _parse_selection('<3;5~7;9',typestr='int',minval=0) |
---|
| 2447 | --> returns [[0,2],[5,7],[9,9]] |
---|
| 2448 | _parse_selection('<3;5~7;9',typestr='float',offset=0.5,minval=0) |
---|
| 2449 | --> returns [[0.,2.5],[5.0,7.0],[9.,9.]] |
---|
[2013] | 2450 | """ |
---|
[2867] | 2451 | # split selstr by '<spaces>;<spaces>' |
---|
| 2452 | semi_sep = re.compile('\s*;\s*') |
---|
| 2453 | selgroups = semi_sep.split(selstr) |
---|
[2013] | 2454 | sellists = [] |
---|
| 2455 | if typestr.lower().startswith('int'): |
---|
| 2456 | formatfunc = int |
---|
[2611] | 2457 | offset = 1 |
---|
[2013] | 2458 | else: |
---|
| 2459 | formatfunc = float |
---|
| 2460 | |
---|
| 2461 | for currsel in selgroups: |
---|
[2867] | 2462 | if currsel.strip() == '*' or len(currsel.strip()) == 0: |
---|
| 2463 | minsel = minval |
---|
| 2464 | maxsel = maxval |
---|
[2013] | 2465 | if currsel.find('~') > 0: |
---|
[2611] | 2466 | # val0 <= x <= val1 |
---|
[2013] | 2467 | minsel = formatfunc(currsel.split('~')[0].strip()) |
---|
[2867] | 2468 | maxsel = formatfunc(currsel.split('~')[1].strip()) |
---|
[2611] | 2469 | elif currsel.strip().find('<=') > -1: |
---|
| 2470 | bound = currsel.split('<=') |
---|
| 2471 | try: # try "x <= val" |
---|
| 2472 | minsel = minval |
---|
| 2473 | maxsel = formatfunc(bound[1].strip()) |
---|
| 2474 | except ValueError: # now "val <= x" |
---|
| 2475 | minsel = formatfunc(bound[0].strip()) |
---|
| 2476 | maxsel = maxval |
---|
| 2477 | elif currsel.strip().find('>=') > -1: |
---|
| 2478 | bound = currsel.split('>=') |
---|
| 2479 | try: # try "x >= val" |
---|
| 2480 | minsel = formatfunc(bound[1].strip()) |
---|
| 2481 | maxsel = maxval |
---|
| 2482 | except ValueError: # now "val >= x" |
---|
| 2483 | minsel = minval |
---|
| 2484 | maxsel = formatfunc(bound[0].strip()) |
---|
| 2485 | elif currsel.strip().find('<') > -1: |
---|
| 2486 | bound = currsel.split('<') |
---|
| 2487 | try: # try "x < val" |
---|
| 2488 | minsel = minval |
---|
| 2489 | maxsel = formatfunc(bound[1].strip()) \ |
---|
| 2490 | - formatfunc(offset) |
---|
| 2491 | except ValueError: # now "val < x" |
---|
| 2492 | minsel = formatfunc(bound[0].strip()) \ |
---|
[2013] | 2493 | + formatfunc(offset) |
---|
[2611] | 2494 | maxsel = maxval |
---|
| 2495 | elif currsel.strip().find('>') > -1: |
---|
| 2496 | bound = currsel.split('>') |
---|
| 2497 | try: # try "x > val" |
---|
| 2498 | minsel = formatfunc(bound[1].strip()) \ |
---|
| 2499 | + formatfunc(offset) |
---|
| 2500 | maxsel = maxval |
---|
| 2501 | except ValueError: # now "val > x" |
---|
| 2502 | minsel = minval |
---|
| 2503 | maxsel = formatfunc(bound[0].strip()) \ |
---|
| 2504 | - formatfunc(offset) |
---|
[2013] | 2505 | else: |
---|
| 2506 | minsel = formatfunc(currsel) |
---|
| 2507 | maxsel = formatfunc(currsel) |
---|
| 2508 | sellists.append([minsel,maxsel]) |
---|
| 2509 | return sellists |
---|
| 2510 | |
---|
[1819] | 2511 | # def get_restfreqs(self): |
---|
| 2512 | # """ |
---|
| 2513 | # Get the restfrequency(s) stored in this scantable. |
---|
| 2514 | # The return value(s) are always of unit 'Hz' |
---|
| 2515 | # Parameters: |
---|
| 2516 | # none |
---|
| 2517 | # Returns: |
---|
| 2518 | # a list of doubles |
---|
| 2519 | # """ |
---|
| 2520 | # return list(self._getrestfreqs()) |
---|
| 2521 | |
---|
| 2522 | def get_restfreqs(self, ids=None): |
---|
[1846] | 2523 | """\ |
---|
[256] | 2524 | Get the restfrequency(s) stored in this scantable. |
---|
| 2525 | The return value(s) are always of unit 'Hz' |
---|
[1846] | 2526 | |
---|
[256] | 2527 | Parameters: |
---|
[1846] | 2528 | |
---|
[1819] | 2529 | ids: (optional) a list of MOLECULE_ID for that restfrequency(s) to |
---|
| 2530 | be retrieved |
---|
[1846] | 2531 | |
---|
[256] | 2532 | Returns: |
---|
[1846] | 2533 | |
---|
[1819] | 2534 | dictionary containing ids and a list of doubles for each id |
---|
[1846] | 2535 | |
---|
[256] | 2536 | """ |
---|
[1819] | 2537 | if ids is None: |
---|
[2349] | 2538 | rfreqs = {} |
---|
[1819] | 2539 | idlist = self.getmolnos() |
---|
| 2540 | for i in idlist: |
---|
[2349] | 2541 | rfreqs[i] = list(self._getrestfreqs(i)) |
---|
[1819] | 2542 | return rfreqs |
---|
| 2543 | else: |
---|
[2349] | 2544 | if type(ids) == list or type(ids) == tuple: |
---|
| 2545 | rfreqs = {} |
---|
[1819] | 2546 | for i in ids: |
---|
[2349] | 2547 | rfreqs[i] = list(self._getrestfreqs(i)) |
---|
[1819] | 2548 | return rfreqs |
---|
| 2549 | else: |
---|
| 2550 | return list(self._getrestfreqs(ids)) |
---|
[102] | 2551 | |
---|
[2349] | 2552 | @asaplog_post_dec |
---|
[931] | 2553 | def set_restfreqs(self, freqs=None, unit='Hz'): |
---|
[1846] | 2554 | """\ |
---|
[931] | 2555 | Set or replace the restfrequency specified and |
---|
[1938] | 2556 | if the 'freqs' argument holds a scalar, |
---|
[931] | 2557 | then that rest frequency will be applied to all the selected |
---|
| 2558 | data. If the 'freqs' argument holds |
---|
| 2559 | a vector, then it MUST be of equal or smaller length than |
---|
| 2560 | the number of IFs (and the available restfrequencies will be |
---|
| 2561 | replaced by this vector). In this case, *all* data have |
---|
| 2562 | the restfrequency set per IF according |
---|
| 2563 | to the corresponding value you give in the 'freqs' vector. |
---|
[1118] | 2564 | E.g. 'freqs=[1e9, 2e9]' would mean IF 0 gets restfreq 1e9 and |
---|
[931] | 2565 | IF 1 gets restfreq 2e9. |
---|
[1846] | 2566 | |
---|
[1395] | 2567 | You can also specify the frequencies via a linecatalog. |
---|
[1153] | 2568 | |
---|
[931] | 2569 | Parameters: |
---|
[1846] | 2570 | |
---|
[931] | 2571 | freqs: list of rest frequency values or string idenitfiers |
---|
[1855] | 2572 | |
---|
[931] | 2573 | unit: unit for rest frequency (default 'Hz') |
---|
[402] | 2574 | |
---|
[1846] | 2575 | |
---|
| 2576 | Example:: |
---|
| 2577 | |
---|
[1819] | 2578 | # set the given restfrequency for the all currently selected IFs |
---|
[931] | 2579 | scan.set_restfreqs(freqs=1.4e9) |
---|
[1845] | 2580 | # set restfrequencies for the n IFs (n > 1) in the order of the |
---|
| 2581 | # list, i.e |
---|
| 2582 | # IF0 -> 1.4e9, IF1 -> 1.41e9, IF3 -> 1.42e9 |
---|
| 2583 | # len(list_of_restfreqs) == nIF |
---|
| 2584 | # for nIF == 1 the following will set multiple restfrequency for |
---|
| 2585 | # that IF |
---|
[1819] | 2586 | scan.set_restfreqs(freqs=[1.4e9, 1.41e9, 1.42e9]) |
---|
[1845] | 2587 | # set multiple restfrequencies per IF. as a list of lists where |
---|
| 2588 | # the outer list has nIF elements, the inner s arbitrary |
---|
| 2589 | scan.set_restfreqs(freqs=[[1.4e9, 1.41e9], [1.67e9]]) |
---|
[391] | 2590 | |
---|
[1846] | 2591 | *Note*: |
---|
[1845] | 2592 | |
---|
[931] | 2593 | To do more sophisticate Restfrequency setting, e.g. on a |
---|
| 2594 | source and IF basis, use scantable.set_selection() before using |
---|
[1846] | 2595 | this function:: |
---|
[931] | 2596 | |
---|
[1846] | 2597 | # provided your scantable is called scan |
---|
| 2598 | selection = selector() |
---|
[2431] | 2599 | selection.set_name('ORION*') |
---|
[1846] | 2600 | selection.set_ifs([1]) |
---|
| 2601 | scan.set_selection(selection) |
---|
| 2602 | scan.set_restfreqs(freqs=86.6e9) |
---|
| 2603 | |
---|
[931] | 2604 | """ |
---|
| 2605 | varlist = vars() |
---|
[1157] | 2606 | from asap import linecatalog |
---|
| 2607 | # simple value |
---|
[1118] | 2608 | if isinstance(freqs, int) or isinstance(freqs, float): |
---|
[1845] | 2609 | self._setrestfreqs([freqs], [""], unit) |
---|
[1157] | 2610 | # list of values |
---|
[1118] | 2611 | elif isinstance(freqs, list) or isinstance(freqs, tuple): |
---|
[1157] | 2612 | # list values are scalars |
---|
[1118] | 2613 | if isinstance(freqs[-1], int) or isinstance(freqs[-1], float): |
---|
[1845] | 2614 | if len(freqs) == 1: |
---|
| 2615 | self._setrestfreqs(freqs, [""], unit) |
---|
| 2616 | else: |
---|
| 2617 | # allow the 'old' mode of setting mulitple IFs |
---|
| 2618 | savesel = self._getselection() |
---|
[2599] | 2619 | sel = self.get_selection() |
---|
[1845] | 2620 | iflist = self.getifnos() |
---|
| 2621 | if len(freqs)>len(iflist): |
---|
| 2622 | raise ValueError("number of elements in list of list " |
---|
| 2623 | "exeeds the current IF selections") |
---|
| 2624 | iflist = self.getifnos() |
---|
| 2625 | for i, fval in enumerate(freqs): |
---|
| 2626 | sel.set_ifs(iflist[i]) |
---|
| 2627 | self._setselection(sel) |
---|
| 2628 | self._setrestfreqs([fval], [""], unit) |
---|
| 2629 | self._setselection(savesel) |
---|
| 2630 | |
---|
| 2631 | # list values are dict, {'value'=, 'name'=) |
---|
[1157] | 2632 | elif isinstance(freqs[-1], dict): |
---|
[1845] | 2633 | values = [] |
---|
| 2634 | names = [] |
---|
| 2635 | for d in freqs: |
---|
| 2636 | values.append(d["value"]) |
---|
| 2637 | names.append(d["name"]) |
---|
| 2638 | self._setrestfreqs(values, names, unit) |
---|
[1819] | 2639 | elif isinstance(freqs[-1], list) or isinstance(freqs[-1], tuple): |
---|
[1157] | 2640 | savesel = self._getselection() |
---|
[2599] | 2641 | sel = self.get_selection() |
---|
[1322] | 2642 | iflist = self.getifnos() |
---|
[1819] | 2643 | if len(freqs)>len(iflist): |
---|
[1845] | 2644 | raise ValueError("number of elements in list of list exeeds" |
---|
| 2645 | " the current IF selections") |
---|
| 2646 | for i, fval in enumerate(freqs): |
---|
[1322] | 2647 | sel.set_ifs(iflist[i]) |
---|
[1259] | 2648 | self._setselection(sel) |
---|
[1845] | 2649 | self._setrestfreqs(fval, [""], unit) |
---|
[1157] | 2650 | self._setselection(savesel) |
---|
| 2651 | # freqs are to be taken from a linecatalog |
---|
[1153] | 2652 | elif isinstance(freqs, linecatalog): |
---|
| 2653 | savesel = self._getselection() |
---|
[2599] | 2654 | sel = self.get_selection() |
---|
[1153] | 2655 | for i in xrange(freqs.nrow()): |
---|
[1322] | 2656 | sel.set_ifs(iflist[i]) |
---|
[1153] | 2657 | self._setselection(sel) |
---|
[1845] | 2658 | self._setrestfreqs([freqs.get_frequency(i)], |
---|
| 2659 | [freqs.get_name(i)], "MHz") |
---|
[1153] | 2660 | # ensure that we are not iterating past nIF |
---|
| 2661 | if i == self.nif()-1: break |
---|
| 2662 | self._setselection(savesel) |
---|
[931] | 2663 | else: |
---|
| 2664 | return |
---|
| 2665 | self._add_history("set_restfreqs", varlist) |
---|
| 2666 | |
---|
[2349] | 2667 | @asaplog_post_dec |
---|
[1360] | 2668 | def shift_refpix(self, delta): |
---|
[1846] | 2669 | """\ |
---|
[1589] | 2670 | Shift the reference pixel of the Spectra Coordinate by an |
---|
| 2671 | integer amount. |
---|
[1846] | 2672 | |
---|
[1589] | 2673 | Parameters: |
---|
[1846] | 2674 | |
---|
[1589] | 2675 | delta: the amount to shift by |
---|
[1846] | 2676 | |
---|
| 2677 | *Note*: |
---|
| 2678 | |
---|
[1589] | 2679 | Be careful using this with broadband data. |
---|
[1846] | 2680 | |
---|
[1360] | 2681 | """ |
---|
[2349] | 2682 | varlist = vars() |
---|
[1731] | 2683 | Scantable.shift_refpix(self, delta) |
---|
[2349] | 2684 | s._add_history("shift_refpix", varlist) |
---|
[931] | 2685 | |
---|
[1862] | 2686 | @asaplog_post_dec |
---|
[2820] | 2687 | def history(self, filename=None, nrows=-1, start=0): |
---|
[1846] | 2688 | """\ |
---|
[1259] | 2689 | Print the history. Optionally to a file. |
---|
[1846] | 2690 | |
---|
[1348] | 2691 | Parameters: |
---|
[1846] | 2692 | |
---|
[1928] | 2693 | filename: The name of the file to save the history to. |
---|
[1846] | 2694 | |
---|
[1259] | 2695 | """ |
---|
[2820] | 2696 | n = self._historylength() |
---|
| 2697 | if nrows == -1: |
---|
| 2698 | nrows = n |
---|
| 2699 | if start+nrows > n: |
---|
| 2700 | nrows = nrows-start |
---|
| 2701 | if n > 1000 and nrows == n: |
---|
| 2702 | nrows = 1000 |
---|
| 2703 | start = n-1000 |
---|
| 2704 | asaplog.push("Warning: History has {0} entries. Displaying last " |
---|
| 2705 | "1000".format(n)) |
---|
| 2706 | hist = list(self._gethistory(nrows, start)) |
---|
[794] | 2707 | out = "-"*80 |
---|
[484] | 2708 | for h in hist: |
---|
[2820] | 2709 | if not h.strip(): |
---|
| 2710 | continue |
---|
| 2711 | if h.find("---") >-1: |
---|
| 2712 | continue |
---|
[489] | 2713 | else: |
---|
| 2714 | items = h.split("##") |
---|
| 2715 | date = items[0] |
---|
| 2716 | func = items[1] |
---|
| 2717 | items = items[2:] |
---|
[794] | 2718 | out += "\n"+date+"\n" |
---|
| 2719 | out += "Function: %s\n Parameters:" % (func) |
---|
[489] | 2720 | for i in items: |
---|
[1938] | 2721 | if i == '': |
---|
| 2722 | continue |
---|
[489] | 2723 | s = i.split("=") |
---|
[1118] | 2724 | out += "\n %s = %s" % (s[0], s[1]) |
---|
[2820] | 2725 | out = "\n".join([out, "*"*80]) |
---|
[1259] | 2726 | if filename is not None: |
---|
| 2727 | if filename is "": |
---|
| 2728 | filename = 'scantable_history.txt' |
---|
| 2729 | filename = os.path.expandvars(os.path.expanduser(filename)) |
---|
| 2730 | if not os.path.isdir(filename): |
---|
| 2731 | data = open(filename, 'w') |
---|
| 2732 | data.write(out) |
---|
| 2733 | data.close() |
---|
| 2734 | else: |
---|
| 2735 | msg = "Illegal file name '%s'." % (filename) |
---|
[1859] | 2736 | raise IOError(msg) |
---|
| 2737 | return page(out) |
---|
[2349] | 2738 | |
---|
[513] | 2739 | # |
---|
| 2740 | # Maths business |
---|
| 2741 | # |
---|
[1862] | 2742 | @asaplog_post_dec |
---|
[2818] | 2743 | def average_time(self, mask=None, scanav=False, weight='tint', align=False, |
---|
| 2744 | avmode="NONE"): |
---|
[1846] | 2745 | """\ |
---|
[2349] | 2746 | Return the (time) weighted average of a scan. Scans will be averaged |
---|
| 2747 | only if the source direction (RA/DEC) is within 1' otherwise |
---|
[1846] | 2748 | |
---|
| 2749 | *Note*: |
---|
| 2750 | |
---|
[1070] | 2751 | in channels only - align if necessary |
---|
[1846] | 2752 | |
---|
[513] | 2753 | Parameters: |
---|
[1846] | 2754 | |
---|
[513] | 2755 | mask: an optional mask (only used for 'var' and 'tsys' |
---|
| 2756 | weighting) |
---|
[1855] | 2757 | |
---|
[558] | 2758 | scanav: True averages each scan separately |
---|
| 2759 | False (default) averages all scans together, |
---|
[1855] | 2760 | |
---|
[1099] | 2761 | weight: Weighting scheme. |
---|
| 2762 | 'none' (mean no weight) |
---|
| 2763 | 'var' (1/var(spec) weighted) |
---|
| 2764 | 'tsys' (1/Tsys**2 weighted) |
---|
| 2765 | 'tint' (integration time weighted) |
---|
| 2766 | 'tintsys' (Tint/Tsys**2) |
---|
| 2767 | 'median' ( median averaging) |
---|
[535] | 2768 | The default is 'tint' |
---|
[1855] | 2769 | |
---|
[931] | 2770 | align: align the spectra in velocity before averaging. It takes |
---|
| 2771 | the time of the first spectrum as reference time. |
---|
[2818] | 2772 | avmode: 'SOURCE' - also select by source name - or |
---|
| 2773 | 'NONE' (default). Not applicable for scanav=True or |
---|
| 2774 | weight=median |
---|
[1846] | 2775 | |
---|
| 2776 | Example:: |
---|
| 2777 | |
---|
[513] | 2778 | # time average the scantable without using a mask |
---|
[710] | 2779 | newscan = scan.average_time() |
---|
[1846] | 2780 | |
---|
[513] | 2781 | """ |
---|
| 2782 | varlist = vars() |
---|
[1593] | 2783 | weight = weight or 'TINT' |
---|
| 2784 | mask = mask or () |
---|
[2818] | 2785 | scanav = (scanav and 'SCAN') or avmode.upper() |
---|
[1118] | 2786 | scan = (self, ) |
---|
[1859] | 2787 | |
---|
| 2788 | if align: |
---|
| 2789 | scan = (self.freq_align(insitu=False), ) |
---|
[2818] | 2790 | asaplog.push("Note: Alignment is don on a source-by-source basis") |
---|
| 2791 | asaplog.push("Note: Averaging (by default) is not") |
---|
| 2792 | # we need to set it to SOURCE averaging here |
---|
[1859] | 2793 | s = None |
---|
| 2794 | if weight.upper() == 'MEDIAN': |
---|
| 2795 | s = scantable(self._math._averagechannel(scan[0], 'MEDIAN', |
---|
| 2796 | scanav)) |
---|
| 2797 | else: |
---|
| 2798 | s = scantable(self._math._average(scan, mask, weight.upper(), |
---|
| 2799 | scanav)) |
---|
[1099] | 2800 | s._add_history("average_time", varlist) |
---|
[513] | 2801 | return s |
---|
[710] | 2802 | |
---|
[1862] | 2803 | @asaplog_post_dec |
---|
[876] | 2804 | def convert_flux(self, jyperk=None, eta=None, d=None, insitu=None): |
---|
[1846] | 2805 | """\ |
---|
[513] | 2806 | Return a scan where all spectra are converted to either |
---|
| 2807 | Jansky or Kelvin depending upon the flux units of the scan table. |
---|
| 2808 | By default the function tries to look the values up internally. |
---|
| 2809 | If it can't find them (or if you want to over-ride), you must |
---|
| 2810 | specify EITHER jyperk OR eta (and D which it will try to look up |
---|
| 2811 | also if you don't set it). jyperk takes precedence if you set both. |
---|
[1846] | 2812 | |
---|
[513] | 2813 | Parameters: |
---|
[1846] | 2814 | |
---|
[513] | 2815 | jyperk: the Jy / K conversion factor |
---|
[1855] | 2816 | |
---|
[513] | 2817 | eta: the aperture efficiency |
---|
[1855] | 2818 | |
---|
[1928] | 2819 | d: the geometric diameter (metres) |
---|
[1855] | 2820 | |
---|
[513] | 2821 | insitu: if False a new scantable is returned. |
---|
| 2822 | Otherwise, the scaling is done in-situ |
---|
| 2823 | The default is taken from .asaprc (False) |
---|
[1846] | 2824 | |
---|
[513] | 2825 | """ |
---|
| 2826 | if insitu is None: insitu = rcParams['insitu'] |
---|
[876] | 2827 | self._math._setinsitu(insitu) |
---|
[513] | 2828 | varlist = vars() |
---|
[1593] | 2829 | jyperk = jyperk or -1.0 |
---|
| 2830 | d = d or -1.0 |
---|
| 2831 | eta = eta or -1.0 |
---|
[876] | 2832 | s = scantable(self._math._convertflux(self, d, eta, jyperk)) |
---|
| 2833 | s._add_history("convert_flux", varlist) |
---|
| 2834 | if insitu: self._assign(s) |
---|
| 2835 | else: return s |
---|
[513] | 2836 | |
---|
[1862] | 2837 | @asaplog_post_dec |
---|
[876] | 2838 | def gain_el(self, poly=None, filename="", method="linear", insitu=None): |
---|
[1846] | 2839 | """\ |
---|
[513] | 2840 | Return a scan after applying a gain-elevation correction. |
---|
| 2841 | The correction can be made via either a polynomial or a |
---|
| 2842 | table-based interpolation (and extrapolation if necessary). |
---|
| 2843 | You specify polynomial coefficients, an ascii table or neither. |
---|
| 2844 | If you specify neither, then a polynomial correction will be made |
---|
| 2845 | with built in coefficients known for certain telescopes (an error |
---|
| 2846 | will occur if the instrument is not known). |
---|
| 2847 | The data and Tsys are *divided* by the scaling factors. |
---|
[1846] | 2848 | |
---|
[513] | 2849 | Parameters: |
---|
[1846] | 2850 | |
---|
[513] | 2851 | poly: Polynomial coefficients (default None) to compute a |
---|
| 2852 | gain-elevation correction as a function of |
---|
| 2853 | elevation (in degrees). |
---|
[1855] | 2854 | |
---|
[513] | 2855 | filename: The name of an ascii file holding correction factors. |
---|
| 2856 | The first row of the ascii file must give the column |
---|
| 2857 | names and these MUST include columns |
---|
[2431] | 2858 | 'ELEVATION' (degrees) and 'FACTOR' (multiply data |
---|
[513] | 2859 | by this) somewhere. |
---|
| 2860 | The second row must give the data type of the |
---|
| 2861 | column. Use 'R' for Real and 'I' for Integer. |
---|
| 2862 | An example file would be |
---|
| 2863 | (actual factors are arbitrary) : |
---|
| 2864 | |
---|
| 2865 | TIME ELEVATION FACTOR |
---|
| 2866 | R R R |
---|
| 2867 | 0.1 0 0.8 |
---|
| 2868 | 0.2 20 0.85 |
---|
| 2869 | 0.3 40 0.9 |
---|
| 2870 | 0.4 60 0.85 |
---|
| 2871 | 0.5 80 0.8 |
---|
| 2872 | 0.6 90 0.75 |
---|
[1855] | 2873 | |
---|
[513] | 2874 | method: Interpolation method when correcting from a table. |
---|
[2431] | 2875 | Values are 'nearest', 'linear' (default), 'cubic' |
---|
| 2876 | and 'spline' |
---|
[1855] | 2877 | |
---|
[513] | 2878 | insitu: if False a new scantable is returned. |
---|
| 2879 | Otherwise, the scaling is done in-situ |
---|
| 2880 | The default is taken from .asaprc (False) |
---|
[1846] | 2881 | |
---|
[513] | 2882 | """ |
---|
| 2883 | |
---|
| 2884 | if insitu is None: insitu = rcParams['insitu'] |
---|
[876] | 2885 | self._math._setinsitu(insitu) |
---|
[513] | 2886 | varlist = vars() |
---|
[1593] | 2887 | poly = poly or () |
---|
[513] | 2888 | from os.path import expandvars |
---|
| 2889 | filename = expandvars(filename) |
---|
[876] | 2890 | s = scantable(self._math._gainel(self, poly, filename, method)) |
---|
| 2891 | s._add_history("gain_el", varlist) |
---|
[1593] | 2892 | if insitu: |
---|
| 2893 | self._assign(s) |
---|
| 2894 | else: |
---|
| 2895 | return s |
---|
[710] | 2896 | |
---|
[1862] | 2897 | @asaplog_post_dec |
---|
[931] | 2898 | def freq_align(self, reftime=None, method='cubic', insitu=None): |
---|
[1846] | 2899 | """\ |
---|
[513] | 2900 | Return a scan where all rows have been aligned in frequency/velocity. |
---|
| 2901 | The alignment frequency frame (e.g. LSRK) is that set by function |
---|
| 2902 | set_freqframe. |
---|
[1846] | 2903 | |
---|
[513] | 2904 | Parameters: |
---|
[1855] | 2905 | |
---|
[513] | 2906 | reftime: reference time to align at. By default, the time of |
---|
| 2907 | the first row of data is used. |
---|
[1855] | 2908 | |
---|
[513] | 2909 | method: Interpolation method for regridding the spectra. |
---|
[2431] | 2910 | Choose from 'nearest', 'linear', 'cubic' (default) |
---|
| 2911 | and 'spline' |
---|
[1855] | 2912 | |
---|
[513] | 2913 | insitu: if False a new scantable is returned. |
---|
| 2914 | Otherwise, the scaling is done in-situ |
---|
| 2915 | The default is taken from .asaprc (False) |
---|
[1846] | 2916 | |
---|
[513] | 2917 | """ |
---|
[931] | 2918 | if insitu is None: insitu = rcParams["insitu"] |
---|
[2429] | 2919 | oldInsitu = self._math._insitu() |
---|
[876] | 2920 | self._math._setinsitu(insitu) |
---|
[513] | 2921 | varlist = vars() |
---|
[1593] | 2922 | reftime = reftime or "" |
---|
[931] | 2923 | s = scantable(self._math._freq_align(self, reftime, method)) |
---|
[876] | 2924 | s._add_history("freq_align", varlist) |
---|
[2429] | 2925 | self._math._setinsitu(oldInsitu) |
---|
[2349] | 2926 | if insitu: |
---|
| 2927 | self._assign(s) |
---|
| 2928 | else: |
---|
| 2929 | return s |
---|
[513] | 2930 | |
---|
[1862] | 2931 | @asaplog_post_dec |
---|
[1725] | 2932 | def opacity(self, tau=None, insitu=None): |
---|
[1846] | 2933 | """\ |
---|
[513] | 2934 | Apply an opacity correction. The data |
---|
| 2935 | and Tsys are multiplied by the correction factor. |
---|
[1846] | 2936 | |
---|
[513] | 2937 | Parameters: |
---|
[1855] | 2938 | |
---|
[1689] | 2939 | tau: (list of) opacity from which the correction factor is |
---|
[513] | 2940 | exp(tau*ZD) |
---|
[1689] | 2941 | where ZD is the zenith-distance. |
---|
| 2942 | If a list is provided, it has to be of length nIF, |
---|
| 2943 | nIF*nPol or 1 and in order of IF/POL, e.g. |
---|
| 2944 | [opif0pol0, opif0pol1, opif1pol0 ...] |
---|
[1725] | 2945 | if tau is `None` the opacities are determined from a |
---|
| 2946 | model. |
---|
[1855] | 2947 | |
---|
[513] | 2948 | insitu: if False a new scantable is returned. |
---|
| 2949 | Otherwise, the scaling is done in-situ |
---|
| 2950 | The default is taken from .asaprc (False) |
---|
[1846] | 2951 | |
---|
[513] | 2952 | """ |
---|
[2349] | 2953 | if insitu is None: |
---|
| 2954 | insitu = rcParams['insitu'] |
---|
[876] | 2955 | self._math._setinsitu(insitu) |
---|
[513] | 2956 | varlist = vars() |
---|
[1689] | 2957 | if not hasattr(tau, "__len__"): |
---|
| 2958 | tau = [tau] |
---|
[876] | 2959 | s = scantable(self._math._opacity(self, tau)) |
---|
| 2960 | s._add_history("opacity", varlist) |
---|
[2349] | 2961 | if insitu: |
---|
| 2962 | self._assign(s) |
---|
| 2963 | else: |
---|
| 2964 | return s |
---|
[513] | 2965 | |
---|
[1862] | 2966 | @asaplog_post_dec |
---|
[513] | 2967 | def bin(self, width=5, insitu=None): |
---|
[1846] | 2968 | """\ |
---|
[513] | 2969 | Return a scan where all spectra have been binned up. |
---|
[1846] | 2970 | |
---|
[1348] | 2971 | Parameters: |
---|
[1846] | 2972 | |
---|
[513] | 2973 | width: The bin width (default=5) in pixels |
---|
[1855] | 2974 | |
---|
[513] | 2975 | insitu: if False a new scantable is returned. |
---|
| 2976 | Otherwise, the scaling is done in-situ |
---|
| 2977 | The default is taken from .asaprc (False) |
---|
[1846] | 2978 | |
---|
[513] | 2979 | """ |
---|
[2349] | 2980 | if insitu is None: |
---|
| 2981 | insitu = rcParams['insitu'] |
---|
[876] | 2982 | self._math._setinsitu(insitu) |
---|
[513] | 2983 | varlist = vars() |
---|
[876] | 2984 | s = scantable(self._math._bin(self, width)) |
---|
[1118] | 2985 | s._add_history("bin", varlist) |
---|
[1589] | 2986 | if insitu: |
---|
| 2987 | self._assign(s) |
---|
| 2988 | else: |
---|
| 2989 | return s |
---|
[513] | 2990 | |
---|
[1862] | 2991 | @asaplog_post_dec |
---|
[2672] | 2992 | def reshape(self, first, last, insitu=None): |
---|
| 2993 | """Resize the band by providing first and last channel. |
---|
| 2994 | This will cut off all channels outside [first, last]. |
---|
| 2995 | """ |
---|
| 2996 | if insitu is None: |
---|
| 2997 | insitu = rcParams['insitu'] |
---|
| 2998 | varlist = vars() |
---|
| 2999 | if last < 0: |
---|
| 3000 | last = self.nchan()-1 + last |
---|
| 3001 | s = None |
---|
| 3002 | if insitu: |
---|
| 3003 | s = self |
---|
| 3004 | else: |
---|
| 3005 | s = self.copy() |
---|
| 3006 | s._reshape(first,last) |
---|
| 3007 | s._add_history("reshape", varlist) |
---|
| 3008 | if not insitu: |
---|
| 3009 | return s |
---|
| 3010 | |
---|
| 3011 | @asaplog_post_dec |
---|
[513] | 3012 | def resample(self, width=5, method='cubic', insitu=None): |
---|
[1846] | 3013 | """\ |
---|
[1348] | 3014 | Return a scan where all spectra have been binned up. |
---|
[1573] | 3015 | |
---|
[1348] | 3016 | Parameters: |
---|
[1846] | 3017 | |
---|
[513] | 3018 | width: The bin width (default=5) in pixels |
---|
[1855] | 3019 | |
---|
[513] | 3020 | method: Interpolation method when correcting from a table. |
---|
[2431] | 3021 | Values are 'nearest', 'linear', 'cubic' (default) |
---|
| 3022 | and 'spline' |
---|
[1855] | 3023 | |
---|
[513] | 3024 | insitu: if False a new scantable is returned. |
---|
| 3025 | Otherwise, the scaling is done in-situ |
---|
| 3026 | The default is taken from .asaprc (False) |
---|
[1846] | 3027 | |
---|
[513] | 3028 | """ |
---|
[2349] | 3029 | if insitu is None: |
---|
| 3030 | insitu = rcParams['insitu'] |
---|
[876] | 3031 | self._math._setinsitu(insitu) |
---|
[513] | 3032 | varlist = vars() |
---|
[876] | 3033 | s = scantable(self._math._resample(self, method, width)) |
---|
[1118] | 3034 | s._add_history("resample", varlist) |
---|
[2349] | 3035 | if insitu: |
---|
| 3036 | self._assign(s) |
---|
| 3037 | else: |
---|
| 3038 | return s |
---|
[513] | 3039 | |
---|
[1862] | 3040 | @asaplog_post_dec |
---|
[946] | 3041 | def average_pol(self, mask=None, weight='none'): |
---|
[1846] | 3042 | """\ |
---|
[946] | 3043 | Average the Polarisations together. |
---|
[1846] | 3044 | |
---|
[946] | 3045 | Parameters: |
---|
[1846] | 3046 | |
---|
[946] | 3047 | mask: An optional mask defining the region, where the |
---|
| 3048 | averaging will be applied. The output will have all |
---|
| 3049 | specified points masked. |
---|
[1855] | 3050 | |
---|
[946] | 3051 | weight: Weighting scheme. 'none' (default), 'var' (1/var(spec) |
---|
| 3052 | weighted), or 'tsys' (1/Tsys**2 weighted) |
---|
[1846] | 3053 | |
---|
[946] | 3054 | """ |
---|
| 3055 | varlist = vars() |
---|
[1593] | 3056 | mask = mask or () |
---|
[1010] | 3057 | s = scantable(self._math._averagepol(self, mask, weight.upper())) |
---|
[1118] | 3058 | s._add_history("average_pol", varlist) |
---|
[992] | 3059 | return s |
---|
[513] | 3060 | |
---|
[1862] | 3061 | @asaplog_post_dec |
---|
[1145] | 3062 | def average_beam(self, mask=None, weight='none'): |
---|
[1846] | 3063 | """\ |
---|
[1145] | 3064 | Average the Beams together. |
---|
[1846] | 3065 | |
---|
[1145] | 3066 | Parameters: |
---|
| 3067 | mask: An optional mask defining the region, where the |
---|
| 3068 | averaging will be applied. The output will have all |
---|
| 3069 | specified points masked. |
---|
[1855] | 3070 | |
---|
[1145] | 3071 | weight: Weighting scheme. 'none' (default), 'var' (1/var(spec) |
---|
| 3072 | weighted), or 'tsys' (1/Tsys**2 weighted) |
---|
[1846] | 3073 | |
---|
[1145] | 3074 | """ |
---|
| 3075 | varlist = vars() |
---|
[1593] | 3076 | mask = mask or () |
---|
[1145] | 3077 | s = scantable(self._math._averagebeams(self, mask, weight.upper())) |
---|
| 3078 | s._add_history("average_beam", varlist) |
---|
| 3079 | return s |
---|
| 3080 | |
---|
[1586] | 3081 | def parallactify(self, pflag): |
---|
[1846] | 3082 | """\ |
---|
[1843] | 3083 | Set a flag to indicate whether this data should be treated as having |
---|
[1617] | 3084 | been 'parallactified' (total phase == 0.0) |
---|
[1846] | 3085 | |
---|
[1617] | 3086 | Parameters: |
---|
[1855] | 3087 | |
---|
[1843] | 3088 | pflag: Bool indicating whether to turn this on (True) or |
---|
[1617] | 3089 | off (False) |
---|
[1846] | 3090 | |
---|
[1617] | 3091 | """ |
---|
[1586] | 3092 | varlist = vars() |
---|
| 3093 | self._parallactify(pflag) |
---|
| 3094 | self._add_history("parallactify", varlist) |
---|
| 3095 | |
---|
[1862] | 3096 | @asaplog_post_dec |
---|
[992] | 3097 | def convert_pol(self, poltype=None): |
---|
[1846] | 3098 | """\ |
---|
[992] | 3099 | Convert the data to a different polarisation type. |
---|
[1565] | 3100 | Note that you will need cross-polarisation terms for most conversions. |
---|
[1846] | 3101 | |
---|
[992] | 3102 | Parameters: |
---|
[1855] | 3103 | |
---|
[992] | 3104 | poltype: The new polarisation type. Valid types are: |
---|
[2431] | 3105 | 'linear', 'circular', 'stokes' and 'linpol' |
---|
[1846] | 3106 | |
---|
[992] | 3107 | """ |
---|
| 3108 | varlist = vars() |
---|
[1859] | 3109 | s = scantable(self._math._convertpol(self, poltype)) |
---|
[1118] | 3110 | s._add_history("convert_pol", varlist) |
---|
[992] | 3111 | return s |
---|
| 3112 | |
---|
[1862] | 3113 | @asaplog_post_dec |
---|
[2269] | 3114 | def smooth(self, kernel="hanning", width=5.0, order=2, plot=False, |
---|
| 3115 | insitu=None): |
---|
[1846] | 3116 | """\ |
---|
[513] | 3117 | Smooth the spectrum by the specified kernel (conserving flux). |
---|
[1846] | 3118 | |
---|
[513] | 3119 | Parameters: |
---|
[1846] | 3120 | |
---|
[513] | 3121 | kernel: The type of smoothing kernel. Select from |
---|
[1574] | 3122 | 'hanning' (default), 'gaussian', 'boxcar', 'rmedian' |
---|
| 3123 | or 'poly' |
---|
[1855] | 3124 | |
---|
[513] | 3125 | width: The width of the kernel in pixels. For hanning this is |
---|
| 3126 | ignored otherwise it defauls to 5 pixels. |
---|
| 3127 | For 'gaussian' it is the Full Width Half |
---|
| 3128 | Maximum. For 'boxcar' it is the full width. |
---|
[1574] | 3129 | For 'rmedian' and 'poly' it is the half width. |
---|
[1855] | 3130 | |
---|
[1574] | 3131 | order: Optional parameter for 'poly' kernel (default is 2), to |
---|
| 3132 | specify the order of the polnomial. Ignored by all other |
---|
| 3133 | kernels. |
---|
[1855] | 3134 | |
---|
[1819] | 3135 | plot: plot the original and the smoothed spectra. |
---|
| 3136 | In this each indivual fit has to be approved, by |
---|
| 3137 | typing 'y' or 'n' |
---|
[1855] | 3138 | |
---|
[513] | 3139 | insitu: if False a new scantable is returned. |
---|
| 3140 | Otherwise, the scaling is done in-situ |
---|
| 3141 | The default is taken from .asaprc (False) |
---|
[1846] | 3142 | |
---|
[513] | 3143 | """ |
---|
| 3144 | if insitu is None: insitu = rcParams['insitu'] |
---|
[876] | 3145 | self._math._setinsitu(insitu) |
---|
[513] | 3146 | varlist = vars() |
---|
[1819] | 3147 | |
---|
| 3148 | if plot: orgscan = self.copy() |
---|
| 3149 | |
---|
[1574] | 3150 | s = scantable(self._math._smooth(self, kernel.lower(), width, order)) |
---|
[876] | 3151 | s._add_history("smooth", varlist) |
---|
[1819] | 3152 | |
---|
[2610] | 3153 | action = 'H' |
---|
[1819] | 3154 | if plot: |
---|
[2150] | 3155 | from asap.asapplotter import new_asaplot |
---|
| 3156 | theplot = new_asaplot(rcParams['plotter.gui']) |
---|
[2535] | 3157 | from matplotlib import rc as rcp |
---|
| 3158 | rcp('lines', linewidth=1) |
---|
[2150] | 3159 | theplot.set_panels() |
---|
[1819] | 3160 | ylab=s._get_ordinate_label() |
---|
[2150] | 3161 | #theplot.palette(0,["#777777","red"]) |
---|
[1819] | 3162 | for r in xrange(s.nrow()): |
---|
| 3163 | xsm=s._getabcissa(r) |
---|
| 3164 | ysm=s._getspectrum(r) |
---|
| 3165 | xorg=orgscan._getabcissa(r) |
---|
| 3166 | yorg=orgscan._getspectrum(r) |
---|
[2610] | 3167 | if action != "N": #skip plotting if rejecting all |
---|
| 3168 | theplot.clear() |
---|
| 3169 | theplot.hold() |
---|
| 3170 | theplot.set_axes('ylabel',ylab) |
---|
| 3171 | theplot.set_axes('xlabel',s._getabcissalabel(r)) |
---|
| 3172 | theplot.set_axes('title',s._getsourcename(r)) |
---|
| 3173 | theplot.set_line(label='Original',color="#777777") |
---|
| 3174 | theplot.plot(xorg,yorg) |
---|
| 3175 | theplot.set_line(label='Smoothed',color="red") |
---|
| 3176 | theplot.plot(xsm,ysm) |
---|
| 3177 | ### Ugly part for legend |
---|
| 3178 | for i in [0,1]: |
---|
| 3179 | theplot.subplots[0]['lines'].append( |
---|
| 3180 | [theplot.subplots[0]['axes'].lines[i]] |
---|
| 3181 | ) |
---|
| 3182 | theplot.release() |
---|
| 3183 | ### Ugly part for legend |
---|
| 3184 | theplot.subplots[0]['lines']=[] |
---|
| 3185 | res = self._get_verify_action("Accept smoothing?",action) |
---|
| 3186 | #print "IF%d, POL%d: got result = %s" %(s.getif(r),s.getpol(r),res) |
---|
| 3187 | if r == 0: action = None |
---|
| 3188 | #res = raw_input("Accept smoothing ([y]/n): ") |
---|
[1819] | 3189 | if res.upper() == 'N': |
---|
[2610] | 3190 | # reject for the current rows |
---|
[1819] | 3191 | s._setspectrum(yorg, r) |
---|
[2610] | 3192 | elif res.upper() == 'R': |
---|
| 3193 | # reject all the following rows |
---|
| 3194 | action = "N" |
---|
| 3195 | s._setspectrum(yorg, r) |
---|
| 3196 | elif res.upper() == 'A': |
---|
| 3197 | # accept all the following rows |
---|
| 3198 | break |
---|
[2150] | 3199 | theplot.quit() |
---|
| 3200 | del theplot |
---|
[1819] | 3201 | del orgscan |
---|
| 3202 | |
---|
[876] | 3203 | if insitu: self._assign(s) |
---|
| 3204 | else: return s |
---|
[513] | 3205 | |
---|
[2186] | 3206 | @asaplog_post_dec |
---|
[2435] | 3207 | def regrid_channel(self, width=5, plot=False, insitu=None): |
---|
| 3208 | """\ |
---|
| 3209 | Regrid the spectra by the specified channel width |
---|
| 3210 | |
---|
| 3211 | Parameters: |
---|
| 3212 | |
---|
| 3213 | width: The channel width (float) of regridded spectra |
---|
| 3214 | in the current spectral unit. |
---|
| 3215 | |
---|
| 3216 | plot: [NOT IMPLEMENTED YET] |
---|
| 3217 | plot the original and the regridded spectra. |
---|
| 3218 | In this each indivual fit has to be approved, by |
---|
| 3219 | typing 'y' or 'n' |
---|
| 3220 | |
---|
| 3221 | insitu: if False a new scantable is returned. |
---|
| 3222 | Otherwise, the scaling is done in-situ |
---|
| 3223 | The default is taken from .asaprc (False) |
---|
| 3224 | |
---|
| 3225 | """ |
---|
| 3226 | if insitu is None: insitu = rcParams['insitu'] |
---|
| 3227 | varlist = vars() |
---|
| 3228 | |
---|
| 3229 | if plot: |
---|
| 3230 | asaplog.post() |
---|
| 3231 | asaplog.push("Verification plot is not implemtnetd yet.") |
---|
| 3232 | asaplog.post("WARN") |
---|
| 3233 | |
---|
| 3234 | s = self.copy() |
---|
| 3235 | s._regrid_specchan(width) |
---|
| 3236 | |
---|
| 3237 | s._add_history("regrid_channel", varlist) |
---|
| 3238 | |
---|
| 3239 | # if plot: |
---|
| 3240 | # from asap.asapplotter import new_asaplot |
---|
| 3241 | # theplot = new_asaplot(rcParams['plotter.gui']) |
---|
[2535] | 3242 | # from matplotlib import rc as rcp |
---|
| 3243 | # rcp('lines', linewidth=1) |
---|
[2435] | 3244 | # theplot.set_panels() |
---|
| 3245 | # ylab=s._get_ordinate_label() |
---|
| 3246 | # #theplot.palette(0,["#777777","red"]) |
---|
| 3247 | # for r in xrange(s.nrow()): |
---|
| 3248 | # xsm=s._getabcissa(r) |
---|
| 3249 | # ysm=s._getspectrum(r) |
---|
| 3250 | # xorg=orgscan._getabcissa(r) |
---|
| 3251 | # yorg=orgscan._getspectrum(r) |
---|
| 3252 | # theplot.clear() |
---|
| 3253 | # theplot.hold() |
---|
| 3254 | # theplot.set_axes('ylabel',ylab) |
---|
| 3255 | # theplot.set_axes('xlabel',s._getabcissalabel(r)) |
---|
| 3256 | # theplot.set_axes('title',s._getsourcename(r)) |
---|
| 3257 | # theplot.set_line(label='Original',color="#777777") |
---|
| 3258 | # theplot.plot(xorg,yorg) |
---|
| 3259 | # theplot.set_line(label='Smoothed',color="red") |
---|
| 3260 | # theplot.plot(xsm,ysm) |
---|
| 3261 | # ### Ugly part for legend |
---|
| 3262 | # for i in [0,1]: |
---|
| 3263 | # theplot.subplots[0]['lines'].append( |
---|
| 3264 | # [theplot.subplots[0]['axes'].lines[i]] |
---|
| 3265 | # ) |
---|
| 3266 | # theplot.release() |
---|
| 3267 | # ### Ugly part for legend |
---|
| 3268 | # theplot.subplots[0]['lines']=[] |
---|
| 3269 | # res = raw_input("Accept smoothing ([y]/n): ") |
---|
| 3270 | # if res.upper() == 'N': |
---|
| 3271 | # s._setspectrum(yorg, r) |
---|
| 3272 | # theplot.quit() |
---|
| 3273 | # del theplot |
---|
| 3274 | # del orgscan |
---|
| 3275 | |
---|
| 3276 | if insitu: self._assign(s) |
---|
| 3277 | else: return s |
---|
| 3278 | |
---|
| 3279 | @asaplog_post_dec |
---|
[2186] | 3280 | def _parse_wn(self, wn): |
---|
| 3281 | if isinstance(wn, list) or isinstance(wn, tuple): |
---|
| 3282 | return wn |
---|
| 3283 | elif isinstance(wn, int): |
---|
| 3284 | return [ wn ] |
---|
| 3285 | elif isinstance(wn, str): |
---|
[2277] | 3286 | if '-' in wn: # case 'a-b' : return [a,a+1,...,b-1,b] |
---|
[2186] | 3287 | val = wn.split('-') |
---|
| 3288 | val = [int(val[0]), int(val[1])] |
---|
| 3289 | val.sort() |
---|
| 3290 | res = [i for i in xrange(val[0], val[1]+1)] |
---|
[2277] | 3291 | elif wn[:2] == '<=' or wn[:2] == '=<': # cases '<=a','=<a' : return [0,1,...,a-1,a] |
---|
[2186] | 3292 | val = int(wn[2:])+1 |
---|
| 3293 | res = [i for i in xrange(val)] |
---|
[2277] | 3294 | elif wn[-2:] == '>=' or wn[-2:] == '=>': # cases 'a>=','a=>' : return [0,1,...,a-1,a] |
---|
[2186] | 3295 | val = int(wn[:-2])+1 |
---|
| 3296 | res = [i for i in xrange(val)] |
---|
[2277] | 3297 | elif wn[0] == '<': # case '<a' : return [0,1,...,a-2,a-1] |
---|
[2186] | 3298 | val = int(wn[1:]) |
---|
| 3299 | res = [i for i in xrange(val)] |
---|
[2277] | 3300 | elif wn[-1] == '>': # case 'a>' : return [0,1,...,a-2,a-1] |
---|
[2186] | 3301 | val = int(wn[:-1]) |
---|
| 3302 | res = [i for i in xrange(val)] |
---|
[2411] | 3303 | elif wn[:2] == '>=' or wn[:2] == '=>': # cases '>=a','=>a' : return [a,-999], which is |
---|
| 3304 | # then interpreted in C++ |
---|
| 3305 | # side as [a,a+1,...,a_nyq] |
---|
| 3306 | # (CAS-3759) |
---|
[2186] | 3307 | val = int(wn[2:]) |
---|
[2411] | 3308 | res = [val, -999] |
---|
| 3309 | #res = [i for i in xrange(val, self.nchan()/2+1)] |
---|
| 3310 | elif wn[-2:] == '<=' or wn[-2:] == '=<': # cases 'a<=','a=<' : return [a,-999], which is |
---|
| 3311 | # then interpreted in C++ |
---|
| 3312 | # side as [a,a+1,...,a_nyq] |
---|
| 3313 | # (CAS-3759) |
---|
[2186] | 3314 | val = int(wn[:-2]) |
---|
[2411] | 3315 | res = [val, -999] |
---|
| 3316 | #res = [i for i in xrange(val, self.nchan()/2+1)] |
---|
| 3317 | elif wn[0] == '>': # case '>a' : return [a+1,-999], which is |
---|
| 3318 | # then interpreted in C++ |
---|
| 3319 | # side as [a+1,a+2,...,a_nyq] |
---|
| 3320 | # (CAS-3759) |
---|
[2186] | 3321 | val = int(wn[1:])+1 |
---|
[2411] | 3322 | res = [val, -999] |
---|
| 3323 | #res = [i for i in xrange(val, self.nchan()/2+1)] |
---|
| 3324 | elif wn[-1] == '<': # case 'a<' : return [a+1,-999], which is |
---|
| 3325 | # then interpreted in C++ |
---|
| 3326 | # side as [a+1,a+2,...,a_nyq] |
---|
| 3327 | # (CAS-3759) |
---|
[2186] | 3328 | val = int(wn[:-1])+1 |
---|
[2411] | 3329 | res = [val, -999] |
---|
| 3330 | #res = [i for i in xrange(val, self.nchan()/2+1)] |
---|
[2012] | 3331 | |
---|
[2186] | 3332 | return res |
---|
| 3333 | else: |
---|
| 3334 | msg = 'wrong value given for addwn/rejwn' |
---|
| 3335 | raise RuntimeError(msg) |
---|
| 3336 | |
---|
[2713] | 3337 | @asaplog_post_dec |
---|
[2810] | 3338 | def apply_bltable(self, insitu=None, retfitres=None, inbltable=None, outbltable=None, overwrite=None): |
---|
[2767] | 3339 | """\ |
---|
| 3340 | Subtract baseline based on parameters written in Baseline Table. |
---|
| 3341 | |
---|
| 3342 | Parameters: |
---|
[2809] | 3343 | insitu: if True, baseline fitting/subtraction is done |
---|
[2810] | 3344 | in-situ. If False, a new scantable with |
---|
| 3345 | baseline subtracted is returned. Actually, |
---|
| 3346 | format of the returned value depends on both |
---|
| 3347 | insitu and retfitres (see below). |
---|
[2767] | 3348 | The default is taken from .asaprc (False) |
---|
[2810] | 3349 | retfitres: if True, the results of baseline fitting (i.e., |
---|
| 3350 | coefficients and rms) are returned. |
---|
| 3351 | default is False. |
---|
| 3352 | The format of the returned value of this |
---|
| 3353 | function varies as follows: |
---|
| 3354 | (1) in case insitu=True and retfitres=True: |
---|
| 3355 | fitting result. |
---|
| 3356 | (2) in case insitu=True and retfitres=False: |
---|
| 3357 | None. |
---|
| 3358 | (3) in case insitu=False and retfitres=True: |
---|
| 3359 | a dictionary containing a new scantable |
---|
| 3360 | (with baseline subtracted) and the fitting |
---|
| 3361 | results. |
---|
| 3362 | (4) in case insitu=False and retfitres=False: |
---|
| 3363 | a new scantable (with baseline subtracted). |
---|
[2767] | 3364 | inbltable: name of input baseline table. The row number of |
---|
| 3365 | scantable and that of inbltable must be |
---|
| 3366 | identical. |
---|
| 3367 | outbltable: name of output baseline table where baseline |
---|
| 3368 | parameters and fitting results recorded. |
---|
| 3369 | default is ''(no output). |
---|
[2809] | 3370 | overwrite: if True when an existing baseline table is |
---|
| 3371 | specified for outbltable, overwrites it. |
---|
| 3372 | Otherwise there is no harm. |
---|
[2767] | 3373 | default is False. |
---|
| 3374 | """ |
---|
| 3375 | |
---|
| 3376 | try: |
---|
| 3377 | varlist = vars() |
---|
[2810] | 3378 | if retfitres is None: retfitres = False |
---|
[2767] | 3379 | if inbltable is None: raise ValueError("bltable missing.") |
---|
| 3380 | if outbltable is None: outbltable = '' |
---|
| 3381 | if overwrite is None: overwrite = False |
---|
| 3382 | |
---|
| 3383 | if insitu is None: insitu = rcParams['insitu'] |
---|
| 3384 | if insitu: |
---|
| 3385 | workscan = self |
---|
| 3386 | else: |
---|
| 3387 | workscan = self.copy() |
---|
| 3388 | |
---|
| 3389 | sres = workscan._apply_bltable(inbltable, |
---|
[2810] | 3390 | retfitres, |
---|
[2767] | 3391 | outbltable, |
---|
| 3392 | os.path.exists(outbltable), |
---|
| 3393 | overwrite) |
---|
[2810] | 3394 | if retfitres: res = parse_fitresult(sres) |
---|
[2767] | 3395 | |
---|
| 3396 | workscan._add_history('apply_bltable', varlist) |
---|
| 3397 | |
---|
| 3398 | if insitu: |
---|
| 3399 | self._assign(workscan) |
---|
[2810] | 3400 | if retfitres: |
---|
| 3401 | return res |
---|
| 3402 | else: |
---|
| 3403 | return None |
---|
[2767] | 3404 | else: |
---|
[2810] | 3405 | if retfitres: |
---|
| 3406 | return {'scantable': workscan, 'fitresults': res} |
---|
| 3407 | else: |
---|
| 3408 | return workscan |
---|
[2767] | 3409 | |
---|
| 3410 | except RuntimeError, e: |
---|
| 3411 | raise_fitting_failure_exception(e) |
---|
| 3412 | |
---|
| 3413 | @asaplog_post_dec |
---|
[2810] | 3414 | def sub_baseline(self, insitu=None, retfitres=None, blinfo=None, bltable=None, overwrite=None): |
---|
[2767] | 3415 | """\ |
---|
| 3416 | Subtract baseline based on parameters written in the input list. |
---|
| 3417 | |
---|
| 3418 | Parameters: |
---|
[2809] | 3419 | insitu: if True, baseline fitting/subtraction is done |
---|
[2810] | 3420 | in-situ. If False, a new scantable with |
---|
| 3421 | baseline subtracted is returned. Actually, |
---|
| 3422 | format of the returned value depends on both |
---|
| 3423 | insitu and retfitres (see below). |
---|
[2767] | 3424 | The default is taken from .asaprc (False) |
---|
[2810] | 3425 | retfitres: if True, the results of baseline fitting (i.e., |
---|
| 3426 | coefficients and rms) are returned. |
---|
| 3427 | default is False. |
---|
| 3428 | The format of the returned value of this |
---|
| 3429 | function varies as follows: |
---|
| 3430 | (1) in case insitu=True and retfitres=True: |
---|
| 3431 | fitting result. |
---|
| 3432 | (2) in case insitu=True and retfitres=False: |
---|
| 3433 | None. |
---|
| 3434 | (3) in case insitu=False and retfitres=True: |
---|
| 3435 | a dictionary containing a new scantable |
---|
| 3436 | (with baseline subtracted) and the fitting |
---|
| 3437 | results. |
---|
| 3438 | (4) in case insitu=False and retfitres=False: |
---|
| 3439 | a new scantable (with baseline subtracted). |
---|
[2767] | 3440 | blinfo: baseline parameter set stored in a dictionary |
---|
| 3441 | or a list of dictionary. Each dictionary |
---|
| 3442 | corresponds to each spectrum and must contain |
---|
| 3443 | the following keys and values: |
---|
| 3444 | 'row': row number, |
---|
| 3445 | 'blfunc': function name. available ones include |
---|
| 3446 | 'poly', 'chebyshev', 'cspline' and |
---|
| 3447 | 'sinusoid', |
---|
| 3448 | 'order': maximum order of polynomial. needed |
---|
| 3449 | if blfunc='poly' or 'chebyshev', |
---|
| 3450 | 'npiece': number or piecewise polynomial. |
---|
| 3451 | needed if blfunc='cspline', |
---|
| 3452 | 'nwave': a list of sinusoidal wave numbers. |
---|
| 3453 | needed if blfunc='sinusoid', and |
---|
| 3454 | 'masklist': min-max windows for channel mask. |
---|
| 3455 | the specified ranges will be used |
---|
| 3456 | for fitting. |
---|
| 3457 | bltable: name of output baseline table where baseline |
---|
| 3458 | parameters and fitting results recorded. |
---|
| 3459 | default is ''(no output). |
---|
[2809] | 3460 | overwrite: if True when an existing baseline table is |
---|
| 3461 | specified for bltable, overwrites it. |
---|
| 3462 | Otherwise there is no harm. |
---|
[2767] | 3463 | default is False. |
---|
| 3464 | |
---|
| 3465 | Example: |
---|
| 3466 | sub_baseline(blinfo=[{'row':0, 'blfunc':'poly', 'order':5, |
---|
| 3467 | 'masklist':[[10,350],[352,510]]}, |
---|
| 3468 | {'row':1, 'blfunc':'cspline', 'npiece':3, |
---|
| 3469 | 'masklist':[[3,16],[19,404],[407,511]]} |
---|
| 3470 | ]) |
---|
| 3471 | |
---|
| 3472 | the first spectrum (row=0) will be fitted with polynomial |
---|
| 3473 | of order=5 and the next one (row=1) will be fitted with cubic |
---|
| 3474 | spline consisting of 3 pieces. |
---|
| 3475 | """ |
---|
| 3476 | |
---|
| 3477 | try: |
---|
| 3478 | varlist = vars() |
---|
[2810] | 3479 | if retfitres is None: retfitres = False |
---|
[2767] | 3480 | if blinfo is None: blinfo = [] |
---|
| 3481 | if bltable is None: bltable = '' |
---|
| 3482 | if overwrite is None: overwrite = False |
---|
| 3483 | |
---|
| 3484 | if insitu is None: insitu = rcParams['insitu'] |
---|
| 3485 | if insitu: |
---|
| 3486 | workscan = self |
---|
| 3487 | else: |
---|
| 3488 | workscan = self.copy() |
---|
| 3489 | |
---|
| 3490 | nrow = workscan.nrow() |
---|
| 3491 | |
---|
| 3492 | in_blinfo = pack_blinfo(blinfo=blinfo, maxirow=nrow) |
---|
| 3493 | |
---|
| 3494 | print "in_blinfo=< "+ str(in_blinfo)+" >" |
---|
| 3495 | |
---|
| 3496 | sres = workscan._sub_baseline(in_blinfo, |
---|
[2810] | 3497 | retfitres, |
---|
[2767] | 3498 | bltable, |
---|
| 3499 | os.path.exists(bltable), |
---|
| 3500 | overwrite) |
---|
[2810] | 3501 | if retfitres: res = parse_fitresult(sres) |
---|
[2767] | 3502 | |
---|
| 3503 | workscan._add_history('sub_baseline', varlist) |
---|
| 3504 | |
---|
| 3505 | if insitu: |
---|
| 3506 | self._assign(workscan) |
---|
[2810] | 3507 | if retfitres: |
---|
| 3508 | return res |
---|
| 3509 | else: |
---|
| 3510 | return None |
---|
[2767] | 3511 | else: |
---|
[2810] | 3512 | if retfitres: |
---|
| 3513 | return {'scantable': workscan, 'fitresults': res} |
---|
| 3514 | else: |
---|
| 3515 | return workscan |
---|
[2767] | 3516 | |
---|
| 3517 | except RuntimeError, e: |
---|
| 3518 | raise_fitting_failure_exception(e) |
---|
| 3519 | |
---|
| 3520 | @asaplog_post_dec |
---|
[2713] | 3521 | def calc_aic(self, value=None, blfunc=None, order=None, mask=None, |
---|
| 3522 | whichrow=None, uselinefinder=None, edge=None, |
---|
| 3523 | threshold=None, chan_avg_limit=None): |
---|
| 3524 | """\ |
---|
| 3525 | Calculates and returns model selection criteria for a specified |
---|
| 3526 | baseline model and a given spectrum data. |
---|
| 3527 | Available values include Akaike Information Criterion (AIC), the |
---|
| 3528 | corrected Akaike Information Criterion (AICc) by Sugiura(1978), |
---|
| 3529 | Bayesian Information Criterion (BIC) and the Generalised Cross |
---|
| 3530 | Validation (GCV). |
---|
[2186] | 3531 | |
---|
[2713] | 3532 | Parameters: |
---|
| 3533 | value: name of model selection criteria to calculate. |
---|
| 3534 | available ones include 'aic', 'aicc', 'bic' and |
---|
| 3535 | 'gcv'. default is 'aicc'. |
---|
| 3536 | blfunc: baseline function name. available ones include |
---|
| 3537 | 'chebyshev', 'cspline' and 'sinusoid'. |
---|
| 3538 | default is 'chebyshev'. |
---|
| 3539 | order: parameter for basline function. actually stands for |
---|
| 3540 | order of polynomial (order) for 'chebyshev', |
---|
| 3541 | number of spline pieces (npiece) for 'cspline' and |
---|
| 3542 | maximum wave number for 'sinusoid', respectively. |
---|
| 3543 | default is 5 (which is also the default order value |
---|
| 3544 | for [auto_]chebyshev_baseline()). |
---|
| 3545 | mask: an optional mask. default is []. |
---|
| 3546 | whichrow: row number. default is 0 (the first row) |
---|
| 3547 | uselinefinder: use sd.linefinder() to flag out line regions |
---|
| 3548 | default is True. |
---|
| 3549 | edge: an optional number of channel to drop at |
---|
| 3550 | the edge of spectrum. If only one value is |
---|
| 3551 | specified, the same number will be dropped |
---|
| 3552 | from both sides of the spectrum. Default |
---|
| 3553 | is to keep all channels. Nested tuples |
---|
| 3554 | represent individual edge selection for |
---|
| 3555 | different IFs (a number of spectral channels |
---|
| 3556 | can be different) |
---|
| 3557 | default is (0, 0). |
---|
| 3558 | threshold: the threshold used by line finder. It is |
---|
| 3559 | better to keep it large as only strong lines |
---|
| 3560 | affect the baseline solution. |
---|
| 3561 | default is 3. |
---|
| 3562 | chan_avg_limit: a maximum number of consequtive spectral |
---|
| 3563 | channels to average during the search of |
---|
| 3564 | weak and broad lines. The default is no |
---|
| 3565 | averaging (and no search for weak lines). |
---|
| 3566 | If such lines can affect the fitted baseline |
---|
| 3567 | (e.g. a high order polynomial is fitted), |
---|
| 3568 | increase this parameter (usually values up |
---|
| 3569 | to 8 are reasonable). Most users of this |
---|
| 3570 | method should find the default value sufficient. |
---|
| 3571 | default is 1. |
---|
| 3572 | |
---|
| 3573 | Example: |
---|
| 3574 | aic = scan.calc_aic(blfunc='chebyshev', order=5, whichrow=0) |
---|
| 3575 | """ |
---|
| 3576 | |
---|
| 3577 | try: |
---|
| 3578 | varlist = vars() |
---|
| 3579 | |
---|
| 3580 | if value is None: value = 'aicc' |
---|
| 3581 | if blfunc is None: blfunc = 'chebyshev' |
---|
| 3582 | if order is None: order = 5 |
---|
| 3583 | if mask is None: mask = [] |
---|
| 3584 | if whichrow is None: whichrow = 0 |
---|
| 3585 | if uselinefinder is None: uselinefinder = True |
---|
| 3586 | if edge is None: edge = (0, 0) |
---|
| 3587 | if threshold is None: threshold = 3 |
---|
| 3588 | if chan_avg_limit is None: chan_avg_limit = 1 |
---|
| 3589 | |
---|
| 3590 | return self._calc_aic(value, blfunc, order, mask, |
---|
| 3591 | whichrow, uselinefinder, edge, |
---|
| 3592 | threshold, chan_avg_limit) |
---|
| 3593 | |
---|
| 3594 | except RuntimeError, e: |
---|
| 3595 | raise_fitting_failure_exception(e) |
---|
| 3596 | |
---|
[1862] | 3597 | @asaplog_post_dec |
---|
[2771] | 3598 | def sinusoid_baseline(self, mask=None, applyfft=None, |
---|
[2269] | 3599 | fftmethod=None, fftthresh=None, |
---|
[2771] | 3600 | addwn=None, rejwn=None, |
---|
| 3601 | insitu=None, |
---|
| 3602 | clipthresh=None, clipniter=None, |
---|
| 3603 | plot=None, |
---|
| 3604 | getresidual=None, |
---|
| 3605 | showprogress=None, minnrow=None, |
---|
| 3606 | outlog=None, |
---|
[2767] | 3607 | blfile=None, csvformat=None, |
---|
| 3608 | bltable=None): |
---|
[2047] | 3609 | """\ |
---|
[2349] | 3610 | Return a scan which has been baselined (all rows) with sinusoidal |
---|
| 3611 | functions. |
---|
| 3612 | |
---|
[2047] | 3613 | Parameters: |
---|
[2186] | 3614 | mask: an optional mask |
---|
| 3615 | applyfft: if True use some method, such as FFT, to find |
---|
| 3616 | strongest sinusoidal components in the wavenumber |
---|
| 3617 | domain to be used for baseline fitting. |
---|
| 3618 | default is True. |
---|
| 3619 | fftmethod: method to find the strong sinusoidal components. |
---|
| 3620 | now only 'fft' is available and it is the default. |
---|
| 3621 | fftthresh: the threshold to select wave numbers to be used for |
---|
| 3622 | fitting from the distribution of amplitudes in the |
---|
| 3623 | wavenumber domain. |
---|
| 3624 | both float and string values accepted. |
---|
| 3625 | given a float value, the unit is set to sigma. |
---|
| 3626 | for string values, allowed formats include: |
---|
[2349] | 3627 | 'xsigma' or 'x' (= x-sigma level. e.g., |
---|
| 3628 | '3sigma'), or |
---|
[2186] | 3629 | 'topx' (= the x strongest ones, e.g. 'top5'). |
---|
| 3630 | default is 3.0 (unit: sigma). |
---|
| 3631 | addwn: the additional wave numbers to be used for fitting. |
---|
| 3632 | list or integer value is accepted to specify every |
---|
| 3633 | wave numbers. also string value can be used in case |
---|
| 3634 | you need to specify wave numbers in a certain range, |
---|
| 3635 | e.g., 'a-b' (= a, a+1, a+2, ..., b-1, b), |
---|
| 3636 | '<a' (= 0,1,...,a-2,a-1), |
---|
| 3637 | '>=a' (= a, a+1, ... up to the maximum wave |
---|
| 3638 | number corresponding to the Nyquist |
---|
| 3639 | frequency for the case of FFT). |
---|
[2411] | 3640 | default is [0]. |
---|
[2186] | 3641 | rejwn: the wave numbers NOT to be used for fitting. |
---|
| 3642 | can be set just as addwn but has higher priority: |
---|
| 3643 | wave numbers which are specified both in addwn |
---|
| 3644 | and rejwn will NOT be used. default is []. |
---|
[2771] | 3645 | insitu: if False a new scantable is returned. |
---|
| 3646 | Otherwise, the scaling is done in-situ |
---|
| 3647 | The default is taken from .asaprc (False) |
---|
[2081] | 3648 | clipthresh: Clipping threshold. (default is 3.0, unit: sigma) |
---|
[2349] | 3649 | clipniter: maximum number of iteration of 'clipthresh'-sigma |
---|
| 3650 | clipping (default is 0) |
---|
[2081] | 3651 | plot: *** CURRENTLY UNAVAILABLE, ALWAYS FALSE *** |
---|
| 3652 | plot the fit and the residual. In this each |
---|
| 3653 | indivual fit has to be approved, by typing 'y' |
---|
| 3654 | or 'n' |
---|
| 3655 | getresidual: if False, returns best-fit values instead of |
---|
| 3656 | residual. (default is True) |
---|
[2189] | 3657 | showprogress: show progress status for large data. |
---|
| 3658 | default is True. |
---|
| 3659 | minnrow: minimum number of input spectra to show. |
---|
| 3660 | default is 1000. |
---|
[2081] | 3661 | outlog: Output the coefficients of the best-fit |
---|
| 3662 | function to logger (default is False) |
---|
| 3663 | blfile: Name of a text file in which the best-fit |
---|
| 3664 | parameter values to be written |
---|
[2186] | 3665 | (default is '': no file/logger output) |
---|
[2641] | 3666 | csvformat: if True blfile is csv-formatted, default is False. |
---|
[2767] | 3667 | bltable: name of a baseline table where fitting results |
---|
| 3668 | (coefficients, rms, etc.) are to be written. |
---|
| 3669 | if given, fitting results will NOT be output to |
---|
| 3670 | scantable (insitu=True) or None will be |
---|
| 3671 | returned (insitu=False). |
---|
| 3672 | (default is "": no table output) |
---|
[2047] | 3673 | |
---|
| 3674 | Example: |
---|
[2349] | 3675 | # return a scan baselined by a combination of sinusoidal curves |
---|
| 3676 | # having wave numbers in spectral window up to 10, |
---|
[2047] | 3677 | # also with 3-sigma clipping, iteration up to 4 times |
---|
[2186] | 3678 | bscan = scan.sinusoid_baseline(addwn='<=10',clipthresh=3.0,clipniter=4) |
---|
[2081] | 3679 | |
---|
| 3680 | Note: |
---|
| 3681 | The best-fit parameter values output in logger and/or blfile are now |
---|
| 3682 | based on specunit of 'channel'. |
---|
[2047] | 3683 | """ |
---|
| 3684 | |
---|
[2186] | 3685 | try: |
---|
| 3686 | varlist = vars() |
---|
[2047] | 3687 | |
---|
[2186] | 3688 | if insitu is None: insitu = rcParams['insitu'] |
---|
| 3689 | if insitu: |
---|
| 3690 | workscan = self |
---|
| 3691 | else: |
---|
| 3692 | workscan = self.copy() |
---|
| 3693 | |
---|
[2410] | 3694 | if mask is None: mask = [] |
---|
[2186] | 3695 | if applyfft is None: applyfft = True |
---|
| 3696 | if fftmethod is None: fftmethod = 'fft' |
---|
| 3697 | if fftthresh is None: fftthresh = 3.0 |
---|
[2411] | 3698 | if addwn is None: addwn = [0] |
---|
[2186] | 3699 | if rejwn is None: rejwn = [] |
---|
| 3700 | if clipthresh is None: clipthresh = 3.0 |
---|
| 3701 | if clipniter is None: clipniter = 0 |
---|
| 3702 | if plot is None: plot = False |
---|
| 3703 | if getresidual is None: getresidual = True |
---|
[2189] | 3704 | if showprogress is None: showprogress = True |
---|
| 3705 | if minnrow is None: minnrow = 1000 |
---|
[2186] | 3706 | if outlog is None: outlog = False |
---|
| 3707 | if blfile is None: blfile = '' |
---|
[2641] | 3708 | if csvformat is None: csvformat = False |
---|
[2767] | 3709 | if bltable is None: bltable = '' |
---|
[2047] | 3710 | |
---|
[2767] | 3711 | sapplyfft = 'true' if applyfft else 'false' |
---|
| 3712 | fftinfo = ','.join([sapplyfft, fftmethod.lower(), str(fftthresh).lower()]) |
---|
[2641] | 3713 | |
---|
[2767] | 3714 | scsvformat = 'T' if csvformat else 'F' |
---|
| 3715 | |
---|
[2081] | 3716 | #CURRENTLY, PLOT=true is UNAVAILABLE UNTIL sinusoidal fitting is implemented as a fitter method. |
---|
[2767] | 3717 | workscan._sinusoid_baseline(mask, |
---|
| 3718 | fftinfo, |
---|
| 3719 | #applyfft, fftmethod.lower(), |
---|
| 3720 | #str(fftthresh).lower(), |
---|
[2349] | 3721 | workscan._parse_wn(addwn), |
---|
[2643] | 3722 | workscan._parse_wn(rejwn), |
---|
| 3723 | clipthresh, clipniter, |
---|
| 3724 | getresidual, |
---|
[2349] | 3725 | pack_progress_params(showprogress, |
---|
[2641] | 3726 | minnrow), |
---|
[2767] | 3727 | outlog, scsvformat+blfile, |
---|
| 3728 | bltable) |
---|
[2186] | 3729 | workscan._add_history('sinusoid_baseline', varlist) |
---|
[2767] | 3730 | |
---|
| 3731 | if bltable == '': |
---|
| 3732 | if insitu: |
---|
| 3733 | self._assign(workscan) |
---|
| 3734 | else: |
---|
| 3735 | return workscan |
---|
[2047] | 3736 | else: |
---|
[2767] | 3737 | if not insitu: |
---|
| 3738 | return None |
---|
[2047] | 3739 | |
---|
| 3740 | except RuntimeError, e: |
---|
[2186] | 3741 | raise_fitting_failure_exception(e) |
---|
[2047] | 3742 | |
---|
| 3743 | |
---|
[2186] | 3744 | @asaplog_post_dec |
---|
[2771] | 3745 | def auto_sinusoid_baseline(self, mask=None, applyfft=None, |
---|
[2349] | 3746 | fftmethod=None, fftthresh=None, |
---|
[2771] | 3747 | addwn=None, rejwn=None, |
---|
| 3748 | insitu=None, |
---|
| 3749 | clipthresh=None, clipniter=None, |
---|
| 3750 | edge=None, threshold=None, chan_avg_limit=None, |
---|
| 3751 | plot=None, |
---|
| 3752 | getresidual=None, |
---|
| 3753 | showprogress=None, minnrow=None, |
---|
| 3754 | outlog=None, |
---|
[2767] | 3755 | blfile=None, csvformat=None, |
---|
| 3756 | bltable=None): |
---|
[2047] | 3757 | """\ |
---|
[2349] | 3758 | Return a scan which has been baselined (all rows) with sinusoidal |
---|
| 3759 | functions. |
---|
[2047] | 3760 | Spectral lines are detected first using linefinder and masked out |
---|
| 3761 | to avoid them affecting the baseline solution. |
---|
| 3762 | |
---|
| 3763 | Parameters: |
---|
[2189] | 3764 | mask: an optional mask retreived from scantable |
---|
| 3765 | applyfft: if True use some method, such as FFT, to find |
---|
| 3766 | strongest sinusoidal components in the wavenumber |
---|
| 3767 | domain to be used for baseline fitting. |
---|
| 3768 | default is True. |
---|
| 3769 | fftmethod: method to find the strong sinusoidal components. |
---|
| 3770 | now only 'fft' is available and it is the default. |
---|
| 3771 | fftthresh: the threshold to select wave numbers to be used for |
---|
| 3772 | fitting from the distribution of amplitudes in the |
---|
| 3773 | wavenumber domain. |
---|
| 3774 | both float and string values accepted. |
---|
| 3775 | given a float value, the unit is set to sigma. |
---|
| 3776 | for string values, allowed formats include: |
---|
[2349] | 3777 | 'xsigma' or 'x' (= x-sigma level. e.g., |
---|
| 3778 | '3sigma'), or |
---|
[2189] | 3779 | 'topx' (= the x strongest ones, e.g. 'top5'). |
---|
| 3780 | default is 3.0 (unit: sigma). |
---|
| 3781 | addwn: the additional wave numbers to be used for fitting. |
---|
| 3782 | list or integer value is accepted to specify every |
---|
| 3783 | wave numbers. also string value can be used in case |
---|
| 3784 | you need to specify wave numbers in a certain range, |
---|
| 3785 | e.g., 'a-b' (= a, a+1, a+2, ..., b-1, b), |
---|
| 3786 | '<a' (= 0,1,...,a-2,a-1), |
---|
| 3787 | '>=a' (= a, a+1, ... up to the maximum wave |
---|
| 3788 | number corresponding to the Nyquist |
---|
| 3789 | frequency for the case of FFT). |
---|
[2411] | 3790 | default is [0]. |
---|
[2189] | 3791 | rejwn: the wave numbers NOT to be used for fitting. |
---|
| 3792 | can be set just as addwn but has higher priority: |
---|
| 3793 | wave numbers which are specified both in addwn |
---|
| 3794 | and rejwn will NOT be used. default is []. |
---|
[2771] | 3795 | insitu: if False a new scantable is returned. |
---|
| 3796 | Otherwise, the scaling is done in-situ |
---|
| 3797 | The default is taken from .asaprc (False) |
---|
[2189] | 3798 | clipthresh: Clipping threshold. (default is 3.0, unit: sigma) |
---|
[2349] | 3799 | clipniter: maximum number of iteration of 'clipthresh'-sigma |
---|
| 3800 | clipping (default is 0) |
---|
[2189] | 3801 | edge: an optional number of channel to drop at |
---|
| 3802 | the edge of spectrum. If only one value is |
---|
| 3803 | specified, the same number will be dropped |
---|
| 3804 | from both sides of the spectrum. Default |
---|
| 3805 | is to keep all channels. Nested tuples |
---|
| 3806 | represent individual edge selection for |
---|
| 3807 | different IFs (a number of spectral channels |
---|
| 3808 | can be different) |
---|
| 3809 | threshold: the threshold used by line finder. It is |
---|
| 3810 | better to keep it large as only strong lines |
---|
| 3811 | affect the baseline solution. |
---|
| 3812 | chan_avg_limit: a maximum number of consequtive spectral |
---|
| 3813 | channels to average during the search of |
---|
| 3814 | weak and broad lines. The default is no |
---|
| 3815 | averaging (and no search for weak lines). |
---|
| 3816 | If such lines can affect the fitted baseline |
---|
| 3817 | (e.g. a high order polynomial is fitted), |
---|
| 3818 | increase this parameter (usually values up |
---|
| 3819 | to 8 are reasonable). Most users of this |
---|
| 3820 | method should find the default value sufficient. |
---|
| 3821 | plot: *** CURRENTLY UNAVAILABLE, ALWAYS FALSE *** |
---|
| 3822 | plot the fit and the residual. In this each |
---|
| 3823 | indivual fit has to be approved, by typing 'y' |
---|
| 3824 | or 'n' |
---|
| 3825 | getresidual: if False, returns best-fit values instead of |
---|
| 3826 | residual. (default is True) |
---|
| 3827 | showprogress: show progress status for large data. |
---|
| 3828 | default is True. |
---|
| 3829 | minnrow: minimum number of input spectra to show. |
---|
| 3830 | default is 1000. |
---|
| 3831 | outlog: Output the coefficients of the best-fit |
---|
| 3832 | function to logger (default is False) |
---|
| 3833 | blfile: Name of a text file in which the best-fit |
---|
| 3834 | parameter values to be written |
---|
| 3835 | (default is "": no file/logger output) |
---|
[2641] | 3836 | csvformat: if True blfile is csv-formatted, default is False. |
---|
[2767] | 3837 | bltable: name of a baseline table where fitting results |
---|
| 3838 | (coefficients, rms, etc.) are to be written. |
---|
| 3839 | if given, fitting results will NOT be output to |
---|
| 3840 | scantable (insitu=True) or None will be |
---|
| 3841 | returned (insitu=False). |
---|
| 3842 | (default is "": no table output) |
---|
[2047] | 3843 | |
---|
| 3844 | Example: |
---|
[2186] | 3845 | bscan = scan.auto_sinusoid_baseline(addwn='<=10', insitu=False) |
---|
[2081] | 3846 | |
---|
| 3847 | Note: |
---|
| 3848 | The best-fit parameter values output in logger and/or blfile are now |
---|
| 3849 | based on specunit of 'channel'. |
---|
[2047] | 3850 | """ |
---|
| 3851 | |
---|
[2186] | 3852 | try: |
---|
| 3853 | varlist = vars() |
---|
[2047] | 3854 | |
---|
[2186] | 3855 | if insitu is None: insitu = rcParams['insitu'] |
---|
| 3856 | if insitu: |
---|
| 3857 | workscan = self |
---|
[2047] | 3858 | else: |
---|
[2186] | 3859 | workscan = self.copy() |
---|
| 3860 | |
---|
[2410] | 3861 | if mask is None: mask = [] |
---|
[2186] | 3862 | if applyfft is None: applyfft = True |
---|
| 3863 | if fftmethod is None: fftmethod = 'fft' |
---|
| 3864 | if fftthresh is None: fftthresh = 3.0 |
---|
[2411] | 3865 | if addwn is None: addwn = [0] |
---|
[2186] | 3866 | if rejwn is None: rejwn = [] |
---|
| 3867 | if clipthresh is None: clipthresh = 3.0 |
---|
| 3868 | if clipniter is None: clipniter = 0 |
---|
| 3869 | if edge is None: edge = (0,0) |
---|
| 3870 | if threshold is None: threshold = 3 |
---|
| 3871 | if chan_avg_limit is None: chan_avg_limit = 1 |
---|
| 3872 | if plot is None: plot = False |
---|
| 3873 | if getresidual is None: getresidual = True |
---|
[2189] | 3874 | if showprogress is None: showprogress = True |
---|
| 3875 | if minnrow is None: minnrow = 1000 |
---|
[2186] | 3876 | if outlog is None: outlog = False |
---|
| 3877 | if blfile is None: blfile = '' |
---|
[2641] | 3878 | if csvformat is None: csvformat = False |
---|
[2767] | 3879 | if bltable is None: bltable = '' |
---|
[2047] | 3880 | |
---|
[2767] | 3881 | sapplyfft = 'true' if applyfft else 'false' |
---|
| 3882 | fftinfo = ','.join([sapplyfft, fftmethod.lower(), str(fftthresh).lower()]) |
---|
[2641] | 3883 | |
---|
[2767] | 3884 | scsvformat = 'T' if csvformat else 'F' |
---|
| 3885 | |
---|
[2277] | 3886 | #CURRENTLY, PLOT=true is UNAVAILABLE UNTIL sinusoidal fitting is implemented as a fitter method. |
---|
[2767] | 3887 | workscan._auto_sinusoid_baseline(mask, |
---|
| 3888 | fftinfo, |
---|
[2349] | 3889 | workscan._parse_wn(addwn), |
---|
| 3890 | workscan._parse_wn(rejwn), |
---|
| 3891 | clipthresh, clipniter, |
---|
| 3892 | normalise_edge_param(edge), |
---|
| 3893 | threshold, chan_avg_limit, |
---|
| 3894 | getresidual, |
---|
| 3895 | pack_progress_params(showprogress, |
---|
| 3896 | minnrow), |
---|
[2767] | 3897 | outlog, scsvformat+blfile, bltable) |
---|
[2047] | 3898 | workscan._add_history("auto_sinusoid_baseline", varlist) |
---|
[2767] | 3899 | |
---|
| 3900 | if bltable == '': |
---|
| 3901 | if insitu: |
---|
| 3902 | self._assign(workscan) |
---|
| 3903 | else: |
---|
| 3904 | return workscan |
---|
[2047] | 3905 | else: |
---|
[2767] | 3906 | if not insitu: |
---|
| 3907 | return None |
---|
[2047] | 3908 | |
---|
| 3909 | except RuntimeError, e: |
---|
[2186] | 3910 | raise_fitting_failure_exception(e) |
---|
[2047] | 3911 | |
---|
| 3912 | @asaplog_post_dec |
---|
[2771] | 3913 | def cspline_baseline(self, mask=None, npiece=None, insitu=None, |
---|
[2349] | 3914 | clipthresh=None, clipniter=None, plot=None, |
---|
| 3915 | getresidual=None, showprogress=None, minnrow=None, |
---|
[2767] | 3916 | outlog=None, blfile=None, csvformat=None, |
---|
| 3917 | bltable=None): |
---|
[1846] | 3918 | """\ |
---|
[2349] | 3919 | Return a scan which has been baselined (all rows) by cubic spline |
---|
| 3920 | function (piecewise cubic polynomial). |
---|
| 3921 | |
---|
[513] | 3922 | Parameters: |
---|
[2771] | 3923 | mask: An optional mask |
---|
| 3924 | npiece: Number of pieces. (default is 2) |
---|
[2189] | 3925 | insitu: If False a new scantable is returned. |
---|
| 3926 | Otherwise, the scaling is done in-situ |
---|
| 3927 | The default is taken from .asaprc (False) |
---|
| 3928 | clipthresh: Clipping threshold. (default is 3.0, unit: sigma) |
---|
[2349] | 3929 | clipniter: maximum number of iteration of 'clipthresh'-sigma |
---|
| 3930 | clipping (default is 0) |
---|
[2189] | 3931 | plot: *** CURRENTLY UNAVAILABLE, ALWAYS FALSE *** |
---|
| 3932 | plot the fit and the residual. In this each |
---|
| 3933 | indivual fit has to be approved, by typing 'y' |
---|
| 3934 | or 'n' |
---|
| 3935 | getresidual: if False, returns best-fit values instead of |
---|
| 3936 | residual. (default is True) |
---|
| 3937 | showprogress: show progress status for large data. |
---|
| 3938 | default is True. |
---|
| 3939 | minnrow: minimum number of input spectra to show. |
---|
| 3940 | default is 1000. |
---|
| 3941 | outlog: Output the coefficients of the best-fit |
---|
| 3942 | function to logger (default is False) |
---|
| 3943 | blfile: Name of a text file in which the best-fit |
---|
| 3944 | parameter values to be written |
---|
| 3945 | (default is "": no file/logger output) |
---|
[2641] | 3946 | csvformat: if True blfile is csv-formatted, default is False. |
---|
[2767] | 3947 | bltable: name of a baseline table where fitting results |
---|
| 3948 | (coefficients, rms, etc.) are to be written. |
---|
| 3949 | if given, fitting results will NOT be output to |
---|
| 3950 | scantable (insitu=True) or None will be |
---|
| 3951 | returned (insitu=False). |
---|
| 3952 | (default is "": no table output) |
---|
[1846] | 3953 | |
---|
[2012] | 3954 | Example: |
---|
[2349] | 3955 | # return a scan baselined by a cubic spline consisting of 2 pieces |
---|
| 3956 | # (i.e., 1 internal knot), |
---|
[2012] | 3957 | # also with 3-sigma clipping, iteration up to 4 times |
---|
| 3958 | bscan = scan.cspline_baseline(npiece=2,clipthresh=3.0,clipniter=4) |
---|
[2081] | 3959 | |
---|
| 3960 | Note: |
---|
| 3961 | The best-fit parameter values output in logger and/or blfile are now |
---|
| 3962 | based on specunit of 'channel'. |
---|
[2012] | 3963 | """ |
---|
| 3964 | |
---|
[2186] | 3965 | try: |
---|
| 3966 | varlist = vars() |
---|
| 3967 | |
---|
| 3968 | if insitu is None: insitu = rcParams['insitu'] |
---|
| 3969 | if insitu: |
---|
| 3970 | workscan = self |
---|
| 3971 | else: |
---|
| 3972 | workscan = self.copy() |
---|
[1855] | 3973 | |
---|
[2410] | 3974 | if mask is None: mask = [] |
---|
[2189] | 3975 | if npiece is None: npiece = 2 |
---|
| 3976 | if clipthresh is None: clipthresh = 3.0 |
---|
| 3977 | if clipniter is None: clipniter = 0 |
---|
| 3978 | if plot is None: plot = False |
---|
| 3979 | if getresidual is None: getresidual = True |
---|
| 3980 | if showprogress is None: showprogress = True |
---|
| 3981 | if minnrow is None: minnrow = 1000 |
---|
| 3982 | if outlog is None: outlog = False |
---|
| 3983 | if blfile is None: blfile = '' |
---|
[2767] | 3984 | if csvformat is None: csvformat = False |
---|
| 3985 | if bltable is None: bltable = '' |
---|
[1855] | 3986 | |
---|
[2767] | 3987 | scsvformat = 'T' if csvformat else 'F' |
---|
[2641] | 3988 | |
---|
[2012] | 3989 | #CURRENTLY, PLOT=true UNAVAILABLE UNTIL cubic spline fitting is implemented as a fitter method. |
---|
[2767] | 3990 | workscan._cspline_baseline(mask, npiece, |
---|
| 3991 | clipthresh, clipniter, |
---|
[2349] | 3992 | getresidual, |
---|
| 3993 | pack_progress_params(showprogress, |
---|
[2641] | 3994 | minnrow), |
---|
[2767] | 3995 | outlog, scsvformat+blfile, |
---|
| 3996 | bltable) |
---|
[2012] | 3997 | workscan._add_history("cspline_baseline", varlist) |
---|
[2767] | 3998 | |
---|
| 3999 | if bltable == '': |
---|
| 4000 | if insitu: |
---|
| 4001 | self._assign(workscan) |
---|
| 4002 | else: |
---|
| 4003 | return workscan |
---|
[2012] | 4004 | else: |
---|
[2767] | 4005 | if not insitu: |
---|
| 4006 | return None |
---|
[2012] | 4007 | |
---|
| 4008 | except RuntimeError, e: |
---|
[2186] | 4009 | raise_fitting_failure_exception(e) |
---|
[1855] | 4010 | |
---|
[2186] | 4011 | @asaplog_post_dec |
---|
[2771] | 4012 | def auto_cspline_baseline(self, mask=None, npiece=None, insitu=None, |
---|
[2349] | 4013 | clipthresh=None, clipniter=None, |
---|
| 4014 | edge=None, threshold=None, chan_avg_limit=None, |
---|
| 4015 | getresidual=None, plot=None, |
---|
| 4016 | showprogress=None, minnrow=None, outlog=None, |
---|
[2767] | 4017 | blfile=None, csvformat=None, bltable=None): |
---|
[2012] | 4018 | """\ |
---|
| 4019 | Return a scan which has been baselined (all rows) by cubic spline |
---|
| 4020 | function (piecewise cubic polynomial). |
---|
| 4021 | Spectral lines are detected first using linefinder and masked out |
---|
| 4022 | to avoid them affecting the baseline solution. |
---|
| 4023 | |
---|
| 4024 | Parameters: |
---|
[2771] | 4025 | mask: an optional mask retreived from scantable |
---|
| 4026 | npiece: Number of pieces. (default is 2) |
---|
[2189] | 4027 | insitu: if False a new scantable is returned. |
---|
| 4028 | Otherwise, the scaling is done in-situ |
---|
| 4029 | The default is taken from .asaprc (False) |
---|
| 4030 | clipthresh: Clipping threshold. (default is 3.0, unit: sigma) |
---|
[2349] | 4031 | clipniter: maximum number of iteration of 'clipthresh'-sigma |
---|
| 4032 | clipping (default is 0) |
---|
[2189] | 4033 | edge: an optional number of channel to drop at |
---|
| 4034 | the edge of spectrum. If only one value is |
---|
| 4035 | specified, the same number will be dropped |
---|
| 4036 | from both sides of the spectrum. Default |
---|
| 4037 | is to keep all channels. Nested tuples |
---|
| 4038 | represent individual edge selection for |
---|
| 4039 | different IFs (a number of spectral channels |
---|
| 4040 | can be different) |
---|
| 4041 | threshold: the threshold used by line finder. It is |
---|
| 4042 | better to keep it large as only strong lines |
---|
| 4043 | affect the baseline solution. |
---|
| 4044 | chan_avg_limit: a maximum number of consequtive spectral |
---|
| 4045 | channels to average during the search of |
---|
| 4046 | weak and broad lines. The default is no |
---|
| 4047 | averaging (and no search for weak lines). |
---|
| 4048 | If such lines can affect the fitted baseline |
---|
| 4049 | (e.g. a high order polynomial is fitted), |
---|
| 4050 | increase this parameter (usually values up |
---|
| 4051 | to 8 are reasonable). Most users of this |
---|
| 4052 | method should find the default value sufficient. |
---|
| 4053 | plot: *** CURRENTLY UNAVAILABLE, ALWAYS FALSE *** |
---|
| 4054 | plot the fit and the residual. In this each |
---|
| 4055 | indivual fit has to be approved, by typing 'y' |
---|
| 4056 | or 'n' |
---|
| 4057 | getresidual: if False, returns best-fit values instead of |
---|
| 4058 | residual. (default is True) |
---|
| 4059 | showprogress: show progress status for large data. |
---|
| 4060 | default is True. |
---|
| 4061 | minnrow: minimum number of input spectra to show. |
---|
| 4062 | default is 1000. |
---|
| 4063 | outlog: Output the coefficients of the best-fit |
---|
| 4064 | function to logger (default is False) |
---|
| 4065 | blfile: Name of a text file in which the best-fit |
---|
| 4066 | parameter values to be written |
---|
| 4067 | (default is "": no file/logger output) |
---|
[2641] | 4068 | csvformat: if True blfile is csv-formatted, default is False. |
---|
[2767] | 4069 | bltable: name of a baseline table where fitting results |
---|
| 4070 | (coefficients, rms, etc.) are to be written. |
---|
| 4071 | if given, fitting results will NOT be output to |
---|
| 4072 | scantable (insitu=True) or None will be |
---|
| 4073 | returned (insitu=False). |
---|
| 4074 | (default is "": no table output) |
---|
[1846] | 4075 | |
---|
[1907] | 4076 | Example: |
---|
[2012] | 4077 | bscan = scan.auto_cspline_baseline(npiece=3, insitu=False) |
---|
[2081] | 4078 | |
---|
| 4079 | Note: |
---|
| 4080 | The best-fit parameter values output in logger and/or blfile are now |
---|
| 4081 | based on specunit of 'channel'. |
---|
[2012] | 4082 | """ |
---|
[1846] | 4083 | |
---|
[2186] | 4084 | try: |
---|
| 4085 | varlist = vars() |
---|
[2012] | 4086 | |
---|
[2186] | 4087 | if insitu is None: insitu = rcParams['insitu'] |
---|
| 4088 | if insitu: |
---|
| 4089 | workscan = self |
---|
[1391] | 4090 | else: |
---|
[2186] | 4091 | workscan = self.copy() |
---|
| 4092 | |
---|
[2410] | 4093 | #if mask is None: mask = [True for i in xrange(workscan.nchan())] |
---|
| 4094 | if mask is None: mask = [] |
---|
[2186] | 4095 | if npiece is None: npiece = 2 |
---|
| 4096 | if clipthresh is None: clipthresh = 3.0 |
---|
| 4097 | if clipniter is None: clipniter = 0 |
---|
| 4098 | if edge is None: edge = (0, 0) |
---|
| 4099 | if threshold is None: threshold = 3 |
---|
| 4100 | if chan_avg_limit is None: chan_avg_limit = 1 |
---|
| 4101 | if plot is None: plot = False |
---|
| 4102 | if getresidual is None: getresidual = True |
---|
[2189] | 4103 | if showprogress is None: showprogress = True |
---|
| 4104 | if minnrow is None: minnrow = 1000 |
---|
[2186] | 4105 | if outlog is None: outlog = False |
---|
| 4106 | if blfile is None: blfile = '' |
---|
[2641] | 4107 | if csvformat is None: csvformat = False |
---|
[2767] | 4108 | if bltable is None: bltable = '' |
---|
[1819] | 4109 | |
---|
[2767] | 4110 | scsvformat = 'T' if csvformat else 'F' |
---|
[2641] | 4111 | |
---|
[2277] | 4112 | #CURRENTLY, PLOT=true UNAVAILABLE UNTIL cubic spline fitting is implemented as a fitter method. |
---|
[2767] | 4113 | workscan._auto_cspline_baseline(mask, npiece, |
---|
| 4114 | clipthresh, clipniter, |
---|
[2269] | 4115 | normalise_edge_param(edge), |
---|
| 4116 | threshold, |
---|
| 4117 | chan_avg_limit, getresidual, |
---|
| 4118 | pack_progress_params(showprogress, |
---|
| 4119 | minnrow), |
---|
[2767] | 4120 | outlog, |
---|
| 4121 | scsvformat+blfile, |
---|
| 4122 | bltable) |
---|
[2012] | 4123 | workscan._add_history("auto_cspline_baseline", varlist) |
---|
[2767] | 4124 | |
---|
| 4125 | if bltable == '': |
---|
| 4126 | if insitu: |
---|
| 4127 | self._assign(workscan) |
---|
| 4128 | else: |
---|
| 4129 | return workscan |
---|
[1856] | 4130 | else: |
---|
[2767] | 4131 | if not insitu: |
---|
| 4132 | return None |
---|
[2012] | 4133 | |
---|
| 4134 | except RuntimeError, e: |
---|
[2186] | 4135 | raise_fitting_failure_exception(e) |
---|
[513] | 4136 | |
---|
[1931] | 4137 | @asaplog_post_dec |
---|
[2771] | 4138 | def chebyshev_baseline(self, mask=None, order=None, insitu=None, |
---|
[2645] | 4139 | clipthresh=None, clipniter=None, plot=None, |
---|
| 4140 | getresidual=None, showprogress=None, minnrow=None, |
---|
[2767] | 4141 | outlog=None, blfile=None, csvformat=None, |
---|
| 4142 | bltable=None): |
---|
[2645] | 4143 | """\ |
---|
| 4144 | Return a scan which has been baselined (all rows) by Chebyshev polynomials. |
---|
| 4145 | |
---|
| 4146 | Parameters: |
---|
[2771] | 4147 | mask: An optional mask |
---|
| 4148 | order: the maximum order of Chebyshev polynomial (default is 5) |
---|
[2645] | 4149 | insitu: If False a new scantable is returned. |
---|
| 4150 | Otherwise, the scaling is done in-situ |
---|
| 4151 | The default is taken from .asaprc (False) |
---|
| 4152 | clipthresh: Clipping threshold. (default is 3.0, unit: sigma) |
---|
| 4153 | clipniter: maximum number of iteration of 'clipthresh'-sigma |
---|
| 4154 | clipping (default is 0) |
---|
| 4155 | plot: *** CURRENTLY UNAVAILABLE, ALWAYS FALSE *** |
---|
| 4156 | plot the fit and the residual. In this each |
---|
| 4157 | indivual fit has to be approved, by typing 'y' |
---|
| 4158 | or 'n' |
---|
| 4159 | getresidual: if False, returns best-fit values instead of |
---|
| 4160 | residual. (default is True) |
---|
| 4161 | showprogress: show progress status for large data. |
---|
| 4162 | default is True. |
---|
| 4163 | minnrow: minimum number of input spectra to show. |
---|
| 4164 | default is 1000. |
---|
| 4165 | outlog: Output the coefficients of the best-fit |
---|
| 4166 | function to logger (default is False) |
---|
| 4167 | blfile: Name of a text file in which the best-fit |
---|
| 4168 | parameter values to be written |
---|
| 4169 | (default is "": no file/logger output) |
---|
| 4170 | csvformat: if True blfile is csv-formatted, default is False. |
---|
[2767] | 4171 | bltable: name of a baseline table where fitting results |
---|
| 4172 | (coefficients, rms, etc.) are to be written. |
---|
| 4173 | if given, fitting results will NOT be output to |
---|
| 4174 | scantable (insitu=True) or None will be |
---|
| 4175 | returned (insitu=False). |
---|
| 4176 | (default is "": no table output) |
---|
[2645] | 4177 | |
---|
| 4178 | Example: |
---|
| 4179 | # return a scan baselined by a cubic spline consisting of 2 pieces |
---|
| 4180 | # (i.e., 1 internal knot), |
---|
| 4181 | # also with 3-sigma clipping, iteration up to 4 times |
---|
| 4182 | bscan = scan.cspline_baseline(npiece=2,clipthresh=3.0,clipniter=4) |
---|
| 4183 | |
---|
| 4184 | Note: |
---|
| 4185 | The best-fit parameter values output in logger and/or blfile are now |
---|
| 4186 | based on specunit of 'channel'. |
---|
| 4187 | """ |
---|
| 4188 | |
---|
| 4189 | try: |
---|
| 4190 | varlist = vars() |
---|
| 4191 | |
---|
| 4192 | if insitu is None: insitu = rcParams['insitu'] |
---|
| 4193 | if insitu: |
---|
| 4194 | workscan = self |
---|
| 4195 | else: |
---|
| 4196 | workscan = self.copy() |
---|
| 4197 | |
---|
| 4198 | if mask is None: mask = [] |
---|
| 4199 | if order is None: order = 5 |
---|
| 4200 | if clipthresh is None: clipthresh = 3.0 |
---|
| 4201 | if clipniter is None: clipniter = 0 |
---|
| 4202 | if plot is None: plot = False |
---|
| 4203 | if getresidual is None: getresidual = True |
---|
| 4204 | if showprogress is None: showprogress = True |
---|
| 4205 | if minnrow is None: minnrow = 1000 |
---|
| 4206 | if outlog is None: outlog = False |
---|
| 4207 | if blfile is None: blfile = '' |
---|
[2767] | 4208 | if csvformat is None: csvformat = False |
---|
| 4209 | if bltable is None: bltable = '' |
---|
[2645] | 4210 | |
---|
[2767] | 4211 | scsvformat = 'T' if csvformat else 'F' |
---|
[2645] | 4212 | |
---|
| 4213 | #CURRENTLY, PLOT=true UNAVAILABLE UNTIL cubic spline fitting is implemented as a fitter method. |
---|
[2767] | 4214 | workscan._chebyshev_baseline(mask, order, |
---|
| 4215 | clipthresh, clipniter, |
---|
[2645] | 4216 | getresidual, |
---|
| 4217 | pack_progress_params(showprogress, |
---|
| 4218 | minnrow), |
---|
[2767] | 4219 | outlog, scsvformat+blfile, |
---|
| 4220 | bltable) |
---|
[2645] | 4221 | workscan._add_history("chebyshev_baseline", varlist) |
---|
[2767] | 4222 | |
---|
| 4223 | if bltable == '': |
---|
| 4224 | if insitu: |
---|
| 4225 | self._assign(workscan) |
---|
| 4226 | else: |
---|
| 4227 | return workscan |
---|
[2645] | 4228 | else: |
---|
[2767] | 4229 | if not insitu: |
---|
| 4230 | return None |
---|
[2645] | 4231 | |
---|
| 4232 | except RuntimeError, e: |
---|
| 4233 | raise_fitting_failure_exception(e) |
---|
| 4234 | |
---|
| 4235 | @asaplog_post_dec |
---|
[2771] | 4236 | def auto_chebyshev_baseline(self, mask=None, order=None, insitu=None, |
---|
[2645] | 4237 | clipthresh=None, clipniter=None, |
---|
| 4238 | edge=None, threshold=None, chan_avg_limit=None, |
---|
| 4239 | getresidual=None, plot=None, |
---|
| 4240 | showprogress=None, minnrow=None, outlog=None, |
---|
[2767] | 4241 | blfile=None, csvformat=None, bltable=None): |
---|
[2645] | 4242 | """\ |
---|
| 4243 | Return a scan which has been baselined (all rows) by Chebyshev polynomials. |
---|
| 4244 | Spectral lines are detected first using linefinder and masked out |
---|
| 4245 | to avoid them affecting the baseline solution. |
---|
| 4246 | |
---|
| 4247 | Parameters: |
---|
[2771] | 4248 | mask: an optional mask retreived from scantable |
---|
| 4249 | order: the maximum order of Chebyshev polynomial (default is 5) |
---|
[2645] | 4250 | insitu: if False a new scantable is returned. |
---|
| 4251 | Otherwise, the scaling is done in-situ |
---|
| 4252 | The default is taken from .asaprc (False) |
---|
| 4253 | clipthresh: Clipping threshold. (default is 3.0, unit: sigma) |
---|
| 4254 | clipniter: maximum number of iteration of 'clipthresh'-sigma |
---|
| 4255 | clipping (default is 0) |
---|
| 4256 | edge: an optional number of channel to drop at |
---|
| 4257 | the edge of spectrum. If only one value is |
---|
| 4258 | specified, the same number will be dropped |
---|
| 4259 | from both sides of the spectrum. Default |
---|
| 4260 | is to keep all channels. Nested tuples |
---|
| 4261 | represent individual edge selection for |
---|
| 4262 | different IFs (a number of spectral channels |
---|
| 4263 | can be different) |
---|
| 4264 | threshold: the threshold used by line finder. It is |
---|
| 4265 | better to keep it large as only strong lines |
---|
| 4266 | affect the baseline solution. |
---|
| 4267 | chan_avg_limit: a maximum number of consequtive spectral |
---|
| 4268 | channels to average during the search of |
---|
| 4269 | weak and broad lines. The default is no |
---|
| 4270 | averaging (and no search for weak lines). |
---|
| 4271 | If such lines can affect the fitted baseline |
---|
| 4272 | (e.g. a high order polynomial is fitted), |
---|
| 4273 | increase this parameter (usually values up |
---|
| 4274 | to 8 are reasonable). Most users of this |
---|
| 4275 | method should find the default value sufficient. |
---|
| 4276 | plot: *** CURRENTLY UNAVAILABLE, ALWAYS FALSE *** |
---|
| 4277 | plot the fit and the residual. In this each |
---|
| 4278 | indivual fit has to be approved, by typing 'y' |
---|
| 4279 | or 'n' |
---|
| 4280 | getresidual: if False, returns best-fit values instead of |
---|
| 4281 | residual. (default is True) |
---|
| 4282 | showprogress: show progress status for large data. |
---|
| 4283 | default is True. |
---|
| 4284 | minnrow: minimum number of input spectra to show. |
---|
| 4285 | default is 1000. |
---|
| 4286 | outlog: Output the coefficients of the best-fit |
---|
| 4287 | function to logger (default is False) |
---|
| 4288 | blfile: Name of a text file in which the best-fit |
---|
| 4289 | parameter values to be written |
---|
| 4290 | (default is "": no file/logger output) |
---|
| 4291 | csvformat: if True blfile is csv-formatted, default is False. |
---|
[2767] | 4292 | bltable: name of a baseline table where fitting results |
---|
| 4293 | (coefficients, rms, etc.) are to be written. |
---|
| 4294 | if given, fitting results will NOT be output to |
---|
| 4295 | scantable (insitu=True) or None will be |
---|
| 4296 | returned (insitu=False). |
---|
| 4297 | (default is "": no table output) |
---|
[2645] | 4298 | |
---|
| 4299 | Example: |
---|
| 4300 | bscan = scan.auto_cspline_baseline(npiece=3, insitu=False) |
---|
| 4301 | |
---|
| 4302 | Note: |
---|
| 4303 | The best-fit parameter values output in logger and/or blfile are now |
---|
| 4304 | based on specunit of 'channel'. |
---|
| 4305 | """ |
---|
| 4306 | |
---|
| 4307 | try: |
---|
| 4308 | varlist = vars() |
---|
| 4309 | |
---|
| 4310 | if insitu is None: insitu = rcParams['insitu'] |
---|
| 4311 | if insitu: |
---|
| 4312 | workscan = self |
---|
| 4313 | else: |
---|
| 4314 | workscan = self.copy() |
---|
| 4315 | |
---|
| 4316 | if mask is None: mask = [] |
---|
| 4317 | if order is None: order = 5 |
---|
| 4318 | if clipthresh is None: clipthresh = 3.0 |
---|
| 4319 | if clipniter is None: clipniter = 0 |
---|
| 4320 | if edge is None: edge = (0, 0) |
---|
| 4321 | if threshold is None: threshold = 3 |
---|
| 4322 | if chan_avg_limit is None: chan_avg_limit = 1 |
---|
| 4323 | if plot is None: plot = False |
---|
| 4324 | if getresidual is None: getresidual = True |
---|
| 4325 | if showprogress is None: showprogress = True |
---|
| 4326 | if minnrow is None: minnrow = 1000 |
---|
| 4327 | if outlog is None: outlog = False |
---|
| 4328 | if blfile is None: blfile = '' |
---|
| 4329 | if csvformat is None: csvformat = False |
---|
[2767] | 4330 | if bltable is None: bltable = '' |
---|
[2645] | 4331 | |
---|
[2767] | 4332 | scsvformat = 'T' if csvformat else 'F' |
---|
[2645] | 4333 | |
---|
| 4334 | #CURRENTLY, PLOT=true UNAVAILABLE UNTIL cubic spline fitting is implemented as a fitter method. |
---|
[2767] | 4335 | workscan._auto_chebyshev_baseline(mask, order, |
---|
| 4336 | clipthresh, clipniter, |
---|
[2645] | 4337 | normalise_edge_param(edge), |
---|
| 4338 | threshold, |
---|
| 4339 | chan_avg_limit, getresidual, |
---|
| 4340 | pack_progress_params(showprogress, |
---|
| 4341 | minnrow), |
---|
[2767] | 4342 | outlog, scsvformat+blfile, |
---|
| 4343 | bltable) |
---|
[2645] | 4344 | workscan._add_history("auto_chebyshev_baseline", varlist) |
---|
[2767] | 4345 | |
---|
| 4346 | if bltable == '': |
---|
| 4347 | if insitu: |
---|
| 4348 | self._assign(workscan) |
---|
| 4349 | else: |
---|
| 4350 | return workscan |
---|
[2645] | 4351 | else: |
---|
[2767] | 4352 | if not insitu: |
---|
| 4353 | return None |
---|
[2645] | 4354 | |
---|
| 4355 | except RuntimeError, e: |
---|
| 4356 | raise_fitting_failure_exception(e) |
---|
| 4357 | |
---|
| 4358 | @asaplog_post_dec |
---|
[2771] | 4359 | def poly_baseline(self, mask=None, order=None, insitu=None, |
---|
[2767] | 4360 | clipthresh=None, clipniter=None, plot=None, |
---|
[2269] | 4361 | getresidual=None, showprogress=None, minnrow=None, |
---|
[2767] | 4362 | outlog=None, blfile=None, csvformat=None, |
---|
| 4363 | bltable=None): |
---|
[1907] | 4364 | """\ |
---|
| 4365 | Return a scan which has been baselined (all rows) by a polynomial. |
---|
| 4366 | Parameters: |
---|
[2771] | 4367 | mask: an optional mask |
---|
| 4368 | order: the order of the polynomial (default is 0) |
---|
[2189] | 4369 | insitu: if False a new scantable is returned. |
---|
| 4370 | Otherwise, the scaling is done in-situ |
---|
| 4371 | The default is taken from .asaprc (False) |
---|
[2767] | 4372 | clipthresh: Clipping threshold. (default is 3.0, unit: sigma) |
---|
| 4373 | clipniter: maximum number of iteration of 'clipthresh'-sigma |
---|
| 4374 | clipping (default is 0) |
---|
[2189] | 4375 | plot: plot the fit and the residual. In this each |
---|
| 4376 | indivual fit has to be approved, by typing 'y' |
---|
| 4377 | or 'n' |
---|
| 4378 | getresidual: if False, returns best-fit values instead of |
---|
| 4379 | residual. (default is True) |
---|
| 4380 | showprogress: show progress status for large data. |
---|
| 4381 | default is True. |
---|
| 4382 | minnrow: minimum number of input spectra to show. |
---|
| 4383 | default is 1000. |
---|
| 4384 | outlog: Output the coefficients of the best-fit |
---|
| 4385 | function to logger (default is False) |
---|
| 4386 | blfile: Name of a text file in which the best-fit |
---|
| 4387 | parameter values to be written |
---|
| 4388 | (default is "": no file/logger output) |
---|
[2641] | 4389 | csvformat: if True blfile is csv-formatted, default is False. |
---|
[2767] | 4390 | bltable: name of a baseline table where fitting results |
---|
| 4391 | (coefficients, rms, etc.) are to be written. |
---|
| 4392 | if given, fitting results will NOT be output to |
---|
| 4393 | scantable (insitu=True) or None will be |
---|
| 4394 | returned (insitu=False). |
---|
| 4395 | (default is "": no table output) |
---|
[2012] | 4396 | |
---|
[1907] | 4397 | Example: |
---|
| 4398 | # return a scan baselined by a third order polynomial, |
---|
| 4399 | # not using a mask |
---|
| 4400 | bscan = scan.poly_baseline(order=3) |
---|
| 4401 | """ |
---|
[1931] | 4402 | |
---|
[2186] | 4403 | try: |
---|
| 4404 | varlist = vars() |
---|
[1931] | 4405 | |
---|
[2269] | 4406 | if insitu is None: |
---|
| 4407 | insitu = rcParams["insitu"] |
---|
[2186] | 4408 | if insitu: |
---|
| 4409 | workscan = self |
---|
| 4410 | else: |
---|
| 4411 | workscan = self.copy() |
---|
[1907] | 4412 | |
---|
[2410] | 4413 | if mask is None: mask = [] |
---|
[2189] | 4414 | if order is None: order = 0 |
---|
[2767] | 4415 | if clipthresh is None: clipthresh = 3.0 |
---|
| 4416 | if clipniter is None: clipniter = 0 |
---|
[2189] | 4417 | if plot is None: plot = False |
---|
| 4418 | if getresidual is None: getresidual = True |
---|
| 4419 | if showprogress is None: showprogress = True |
---|
| 4420 | if minnrow is None: minnrow = 1000 |
---|
| 4421 | if outlog is None: outlog = False |
---|
[2767] | 4422 | if blfile is None: blfile = '' |
---|
[2641] | 4423 | if csvformat is None: csvformat = False |
---|
[2767] | 4424 | if bltable is None: bltable = '' |
---|
[1907] | 4425 | |
---|
[2767] | 4426 | scsvformat = 'T' if csvformat else 'F' |
---|
[2641] | 4427 | |
---|
[2012] | 4428 | if plot: |
---|
[2269] | 4429 | outblfile = (blfile != "") and \ |
---|
[2349] | 4430 | os.path.exists(os.path.expanduser( |
---|
| 4431 | os.path.expandvars(blfile)) |
---|
| 4432 | ) |
---|
[2269] | 4433 | if outblfile: |
---|
| 4434 | blf = open(blfile, "a") |
---|
[2012] | 4435 | |
---|
[1907] | 4436 | f = fitter() |
---|
| 4437 | f.set_function(lpoly=order) |
---|
[2186] | 4438 | |
---|
| 4439 | rows = xrange(workscan.nrow()) |
---|
| 4440 | #if len(rows) > 0: workscan._init_blinfo() |
---|
[2610] | 4441 | |
---|
| 4442 | action = "H" |
---|
[1907] | 4443 | for r in rows: |
---|
| 4444 | f.x = workscan._getabcissa(r) |
---|
| 4445 | f.y = workscan._getspectrum(r) |
---|
[2541] | 4446 | if mask: |
---|
| 4447 | f.mask = mask_and(mask, workscan._getmask(r)) # (CAS-1434) |
---|
| 4448 | else: # mask=None |
---|
| 4449 | f.mask = workscan._getmask(r) |
---|
| 4450 | |
---|
[1907] | 4451 | f.data = None |
---|
| 4452 | f.fit() |
---|
[2541] | 4453 | |
---|
[2610] | 4454 | if action != "Y": # skip plotting when accepting all |
---|
| 4455 | f.plot(residual=True) |
---|
| 4456 | #accept_fit = raw_input("Accept fit ( [y]/n ): ") |
---|
| 4457 | #if accept_fit.upper() == "N": |
---|
| 4458 | # #workscan._append_blinfo(None, None, None) |
---|
| 4459 | # continue |
---|
| 4460 | accept_fit = self._get_verify_action("Accept fit?",action) |
---|
| 4461 | if r == 0: action = None |
---|
[1907] | 4462 | if accept_fit.upper() == "N": |
---|
| 4463 | continue |
---|
[2610] | 4464 | elif accept_fit.upper() == "R": |
---|
| 4465 | break |
---|
| 4466 | elif accept_fit.upper() == "A": |
---|
| 4467 | action = "Y" |
---|
[2012] | 4468 | |
---|
| 4469 | blpars = f.get_parameters() |
---|
| 4470 | masklist = workscan.get_masklist(f.mask, row=r, silent=True) |
---|
| 4471 | #workscan._append_blinfo(blpars, masklist, f.mask) |
---|
[2269] | 4472 | workscan._setspectrum((f.fitter.getresidual() |
---|
| 4473 | if getresidual else f.fitter.getfit()), r) |
---|
[1907] | 4474 | |
---|
[2012] | 4475 | if outblfile: |
---|
| 4476 | rms = workscan.get_rms(f.mask, r) |
---|
[2269] | 4477 | dataout = \ |
---|
| 4478 | workscan.format_blparams_row(blpars["params"], |
---|
| 4479 | blpars["fixed"], |
---|
| 4480 | rms, str(masklist), |
---|
[2641] | 4481 | r, True, csvformat) |
---|
[2012] | 4482 | blf.write(dataout) |
---|
| 4483 | |
---|
[1907] | 4484 | f._p.unmap() |
---|
| 4485 | f._p = None |
---|
[2012] | 4486 | |
---|
[2349] | 4487 | if outblfile: |
---|
| 4488 | blf.close() |
---|
[1907] | 4489 | else: |
---|
[2767] | 4490 | workscan._poly_baseline(mask, order, |
---|
| 4491 | clipthresh, clipniter, # |
---|
| 4492 | getresidual, |
---|
[2269] | 4493 | pack_progress_params(showprogress, |
---|
| 4494 | minnrow), |
---|
[2767] | 4495 | outlog, scsvformat+blfile, |
---|
| 4496 | bltable) # |
---|
[1907] | 4497 | |
---|
| 4498 | workscan._add_history("poly_baseline", varlist) |
---|
| 4499 | |
---|
| 4500 | if insitu: |
---|
| 4501 | self._assign(workscan) |
---|
| 4502 | else: |
---|
| 4503 | return workscan |
---|
| 4504 | |
---|
[1919] | 4505 | except RuntimeError, e: |
---|
[2186] | 4506 | raise_fitting_failure_exception(e) |
---|
[1907] | 4507 | |
---|
[2186] | 4508 | @asaplog_post_dec |
---|
[2771] | 4509 | def auto_poly_baseline(self, mask=None, order=None, insitu=None, |
---|
[2767] | 4510 | clipthresh=None, clipniter=None, |
---|
| 4511 | edge=None, threshold=None, chan_avg_limit=None, |
---|
| 4512 | getresidual=None, plot=None, |
---|
| 4513 | showprogress=None, minnrow=None, outlog=None, |
---|
| 4514 | blfile=None, csvformat=None, bltable=None): |
---|
[1846] | 4515 | """\ |
---|
[1931] | 4516 | Return a scan which has been baselined (all rows) by a polynomial. |
---|
[880] | 4517 | Spectral lines are detected first using linefinder and masked out |
---|
| 4518 | to avoid them affecting the baseline solution. |
---|
| 4519 | |
---|
| 4520 | Parameters: |
---|
[2771] | 4521 | mask: an optional mask retreived from scantable |
---|
| 4522 | order: the order of the polynomial (default is 0) |
---|
[2189] | 4523 | insitu: if False a new scantable is returned. |
---|
| 4524 | Otherwise, the scaling is done in-situ |
---|
| 4525 | The default is taken from .asaprc (False) |
---|
[2767] | 4526 | clipthresh: Clipping threshold. (default is 3.0, unit: sigma) |
---|
| 4527 | clipniter: maximum number of iteration of 'clipthresh'-sigma |
---|
| 4528 | clipping (default is 0) |
---|
[2189] | 4529 | edge: an optional number of channel to drop at |
---|
| 4530 | the edge of spectrum. If only one value is |
---|
| 4531 | specified, the same number will be dropped |
---|
| 4532 | from both sides of the spectrum. Default |
---|
| 4533 | is to keep all channels. Nested tuples |
---|
| 4534 | represent individual edge selection for |
---|
| 4535 | different IFs (a number of spectral channels |
---|
| 4536 | can be different) |
---|
| 4537 | threshold: the threshold used by line finder. It is |
---|
| 4538 | better to keep it large as only strong lines |
---|
| 4539 | affect the baseline solution. |
---|
| 4540 | chan_avg_limit: a maximum number of consequtive spectral |
---|
| 4541 | channels to average during the search of |
---|
| 4542 | weak and broad lines. The default is no |
---|
| 4543 | averaging (and no search for weak lines). |
---|
| 4544 | If such lines can affect the fitted baseline |
---|
| 4545 | (e.g. a high order polynomial is fitted), |
---|
| 4546 | increase this parameter (usually values up |
---|
| 4547 | to 8 are reasonable). Most users of this |
---|
| 4548 | method should find the default value sufficient. |
---|
| 4549 | plot: plot the fit and the residual. In this each |
---|
| 4550 | indivual fit has to be approved, by typing 'y' |
---|
| 4551 | or 'n' |
---|
| 4552 | getresidual: if False, returns best-fit values instead of |
---|
| 4553 | residual. (default is True) |
---|
| 4554 | showprogress: show progress status for large data. |
---|
| 4555 | default is True. |
---|
| 4556 | minnrow: minimum number of input spectra to show. |
---|
| 4557 | default is 1000. |
---|
| 4558 | outlog: Output the coefficients of the best-fit |
---|
| 4559 | function to logger (default is False) |
---|
| 4560 | blfile: Name of a text file in which the best-fit |
---|
| 4561 | parameter values to be written |
---|
| 4562 | (default is "": no file/logger output) |
---|
[2641] | 4563 | csvformat: if True blfile is csv-formatted, default is False. |
---|
[2767] | 4564 | bltable: name of a baseline table where fitting results |
---|
| 4565 | (coefficients, rms, etc.) are to be written. |
---|
| 4566 | if given, fitting results will NOT be output to |
---|
| 4567 | scantable (insitu=True) or None will be |
---|
| 4568 | returned (insitu=False). |
---|
| 4569 | (default is "": no table output) |
---|
[1846] | 4570 | |
---|
[2012] | 4571 | Example: |
---|
| 4572 | bscan = scan.auto_poly_baseline(order=7, insitu=False) |
---|
| 4573 | """ |
---|
[880] | 4574 | |
---|
[2186] | 4575 | try: |
---|
| 4576 | varlist = vars() |
---|
[1846] | 4577 | |
---|
[2269] | 4578 | if insitu is None: |
---|
| 4579 | insitu = rcParams['insitu'] |
---|
[2186] | 4580 | if insitu: |
---|
| 4581 | workscan = self |
---|
| 4582 | else: |
---|
| 4583 | workscan = self.copy() |
---|
[1846] | 4584 | |
---|
[2410] | 4585 | if mask is None: mask = [] |
---|
[2186] | 4586 | if order is None: order = 0 |
---|
[2767] | 4587 | if clipthresh is None: clipthresh = 3.0 |
---|
| 4588 | if clipniter is None: clipniter = 0 |
---|
[2186] | 4589 | if edge is None: edge = (0, 0) |
---|
| 4590 | if threshold is None: threshold = 3 |
---|
| 4591 | if chan_avg_limit is None: chan_avg_limit = 1 |
---|
| 4592 | if plot is None: plot = False |
---|
| 4593 | if getresidual is None: getresidual = True |
---|
[2189] | 4594 | if showprogress is None: showprogress = True |
---|
| 4595 | if minnrow is None: minnrow = 1000 |
---|
[2186] | 4596 | if outlog is None: outlog = False |
---|
| 4597 | if blfile is None: blfile = '' |
---|
[2641] | 4598 | if csvformat is None: csvformat = False |
---|
[2767] | 4599 | if bltable is None: bltable = '' |
---|
[1846] | 4600 | |
---|
[2767] | 4601 | scsvformat = 'T' if csvformat else 'F' |
---|
[2641] | 4602 | |
---|
[2186] | 4603 | edge = normalise_edge_param(edge) |
---|
[880] | 4604 | |
---|
[2012] | 4605 | if plot: |
---|
[2269] | 4606 | outblfile = (blfile != "") and \ |
---|
| 4607 | os.path.exists(os.path.expanduser(os.path.expandvars(blfile))) |
---|
[2012] | 4608 | if outblfile: blf = open(blfile, "a") |
---|
| 4609 | |
---|
[2186] | 4610 | from asap.asaplinefind import linefinder |
---|
[2012] | 4611 | fl = linefinder() |
---|
[2269] | 4612 | fl.set_options(threshold=threshold, avg_limit=chan_avg_limit) |
---|
[2012] | 4613 | fl.set_scan(workscan) |
---|
[2186] | 4614 | |
---|
[2012] | 4615 | f = fitter() |
---|
| 4616 | f.set_function(lpoly=order) |
---|
[880] | 4617 | |
---|
[2186] | 4618 | rows = xrange(workscan.nrow()) |
---|
| 4619 | #if len(rows) > 0: workscan._init_blinfo() |
---|
[2610] | 4620 | |
---|
| 4621 | action = "H" |
---|
[2012] | 4622 | for r in rows: |
---|
[2186] | 4623 | idx = 2*workscan.getif(r) |
---|
[2541] | 4624 | if mask: |
---|
| 4625 | msk = mask_and(mask, workscan._getmask(r)) # (CAS-1434) |
---|
| 4626 | else: # mask=None |
---|
| 4627 | msk = workscan._getmask(r) |
---|
| 4628 | fl.find_lines(r, msk, edge[idx:idx+2]) |
---|
[907] | 4629 | |
---|
[2012] | 4630 | f.x = workscan._getabcissa(r) |
---|
| 4631 | f.y = workscan._getspectrum(r) |
---|
| 4632 | f.mask = fl.get_mask() |
---|
| 4633 | f.data = None |
---|
| 4634 | f.fit() |
---|
| 4635 | |
---|
[2610] | 4636 | if action != "Y": # skip plotting when accepting all |
---|
| 4637 | f.plot(residual=True) |
---|
| 4638 | #accept_fit = raw_input("Accept fit ( [y]/n ): ") |
---|
| 4639 | accept_fit = self._get_verify_action("Accept fit?",action) |
---|
| 4640 | if r == 0: action = None |
---|
[2012] | 4641 | if accept_fit.upper() == "N": |
---|
| 4642 | #workscan._append_blinfo(None, None, None) |
---|
| 4643 | continue |
---|
[2610] | 4644 | elif accept_fit.upper() == "R": |
---|
| 4645 | break |
---|
| 4646 | elif accept_fit.upper() == "A": |
---|
| 4647 | action = "Y" |
---|
[2012] | 4648 | |
---|
| 4649 | blpars = f.get_parameters() |
---|
| 4650 | masklist = workscan.get_masklist(f.mask, row=r, silent=True) |
---|
| 4651 | #workscan._append_blinfo(blpars, masklist, f.mask) |
---|
[2349] | 4652 | workscan._setspectrum( |
---|
| 4653 | (f.fitter.getresidual() if getresidual |
---|
| 4654 | else f.fitter.getfit()), r |
---|
| 4655 | ) |
---|
[2012] | 4656 | |
---|
| 4657 | if outblfile: |
---|
| 4658 | rms = workscan.get_rms(f.mask, r) |
---|
[2269] | 4659 | dataout = \ |
---|
| 4660 | workscan.format_blparams_row(blpars["params"], |
---|
| 4661 | blpars["fixed"], |
---|
| 4662 | rms, str(masklist), |
---|
[2641] | 4663 | r, True, csvformat) |
---|
[2012] | 4664 | blf.write(dataout) |
---|
| 4665 | |
---|
| 4666 | f._p.unmap() |
---|
| 4667 | f._p = None |
---|
| 4668 | |
---|
| 4669 | if outblfile: blf.close() |
---|
| 4670 | else: |
---|
[2767] | 4671 | workscan._auto_poly_baseline(mask, order, |
---|
| 4672 | clipthresh, clipniter, |
---|
| 4673 | edge, threshold, |
---|
[2269] | 4674 | chan_avg_limit, getresidual, |
---|
| 4675 | pack_progress_params(showprogress, |
---|
| 4676 | minnrow), |
---|
[2767] | 4677 | outlog, scsvformat+blfile, |
---|
| 4678 | bltable) |
---|
| 4679 | workscan._add_history("auto_poly_baseline", varlist) |
---|
[2012] | 4680 | |
---|
[2767] | 4681 | if bltable == '': |
---|
| 4682 | if insitu: |
---|
| 4683 | self._assign(workscan) |
---|
| 4684 | else: |
---|
| 4685 | return workscan |
---|
[2012] | 4686 | else: |
---|
[2767] | 4687 | if not insitu: |
---|
| 4688 | return None |
---|
[2012] | 4689 | |
---|
| 4690 | except RuntimeError, e: |
---|
[2186] | 4691 | raise_fitting_failure_exception(e) |
---|
[2012] | 4692 | |
---|
| 4693 | def _init_blinfo(self): |
---|
| 4694 | """\ |
---|
| 4695 | Initialise the following three auxiliary members: |
---|
| 4696 | blpars : parameters of the best-fit baseline, |
---|
| 4697 | masklists : mask data (edge positions of masked channels) and |
---|
| 4698 | actualmask : mask data (in boolean list), |
---|
| 4699 | to keep for use later (including output to logger/text files). |
---|
| 4700 | Used by poly_baseline() and auto_poly_baseline() in case of |
---|
| 4701 | 'plot=True'. |
---|
| 4702 | """ |
---|
| 4703 | self.blpars = [] |
---|
| 4704 | self.masklists = [] |
---|
| 4705 | self.actualmask = [] |
---|
| 4706 | return |
---|
[880] | 4707 | |
---|
[2012] | 4708 | def _append_blinfo(self, data_blpars, data_masklists, data_actualmask): |
---|
| 4709 | """\ |
---|
| 4710 | Append baseline-fitting related info to blpars, masklist and |
---|
| 4711 | actualmask. |
---|
| 4712 | """ |
---|
| 4713 | self.blpars.append(data_blpars) |
---|
| 4714 | self.masklists.append(data_masklists) |
---|
| 4715 | self.actualmask.append(data_actualmask) |
---|
| 4716 | return |
---|
| 4717 | |
---|
[1862] | 4718 | @asaplog_post_dec |
---|
[914] | 4719 | def rotate_linpolphase(self, angle): |
---|
[1846] | 4720 | """\ |
---|
[914] | 4721 | Rotate the phase of the complex polarization O=Q+iU correlation. |
---|
| 4722 | This is always done in situ in the raw data. So if you call this |
---|
| 4723 | function more than once then each call rotates the phase further. |
---|
[1846] | 4724 | |
---|
[914] | 4725 | Parameters: |
---|
[1846] | 4726 | |
---|
[914] | 4727 | angle: The angle (degrees) to rotate (add) by. |
---|
[1846] | 4728 | |
---|
| 4729 | Example:: |
---|
| 4730 | |
---|
[914] | 4731 | scan.rotate_linpolphase(2.3) |
---|
[1846] | 4732 | |
---|
[914] | 4733 | """ |
---|
| 4734 | varlist = vars() |
---|
[936] | 4735 | self._math._rotate_linpolphase(self, angle) |
---|
[914] | 4736 | self._add_history("rotate_linpolphase", varlist) |
---|
| 4737 | return |
---|
[710] | 4738 | |
---|
[1862] | 4739 | @asaplog_post_dec |
---|
[914] | 4740 | def rotate_xyphase(self, angle): |
---|
[1846] | 4741 | """\ |
---|
[914] | 4742 | Rotate the phase of the XY correlation. This is always done in situ |
---|
| 4743 | in the data. So if you call this function more than once |
---|
| 4744 | then each call rotates the phase further. |
---|
[1846] | 4745 | |
---|
[914] | 4746 | Parameters: |
---|
[1846] | 4747 | |
---|
[914] | 4748 | angle: The angle (degrees) to rotate (add) by. |
---|
[1846] | 4749 | |
---|
| 4750 | Example:: |
---|
| 4751 | |
---|
[914] | 4752 | scan.rotate_xyphase(2.3) |
---|
[1846] | 4753 | |
---|
[914] | 4754 | """ |
---|
| 4755 | varlist = vars() |
---|
[936] | 4756 | self._math._rotate_xyphase(self, angle) |
---|
[914] | 4757 | self._add_history("rotate_xyphase", varlist) |
---|
| 4758 | return |
---|
| 4759 | |
---|
[1862] | 4760 | @asaplog_post_dec |
---|
[914] | 4761 | def swap_linears(self): |
---|
[1846] | 4762 | """\ |
---|
[1573] | 4763 | Swap the linear polarisations XX and YY, or better the first two |
---|
[1348] | 4764 | polarisations as this also works for ciculars. |
---|
[914] | 4765 | """ |
---|
| 4766 | varlist = vars() |
---|
[936] | 4767 | self._math._swap_linears(self) |
---|
[914] | 4768 | self._add_history("swap_linears", varlist) |
---|
| 4769 | return |
---|
| 4770 | |
---|
[1862] | 4771 | @asaplog_post_dec |
---|
[914] | 4772 | def invert_phase(self): |
---|
[1846] | 4773 | """\ |
---|
[914] | 4774 | Invert the phase of the complex polarisation |
---|
| 4775 | """ |
---|
| 4776 | varlist = vars() |
---|
[936] | 4777 | self._math._invert_phase(self) |
---|
[914] | 4778 | self._add_history("invert_phase", varlist) |
---|
| 4779 | return |
---|
| 4780 | |
---|
[1862] | 4781 | @asaplog_post_dec |
---|
[876] | 4782 | def add(self, offset, insitu=None): |
---|
[1846] | 4783 | """\ |
---|
[513] | 4784 | Return a scan where all spectra have the offset added |
---|
[1846] | 4785 | |
---|
[513] | 4786 | Parameters: |
---|
[1846] | 4787 | |
---|
[513] | 4788 | offset: the offset |
---|
[1855] | 4789 | |
---|
[513] | 4790 | insitu: if False a new scantable is returned. |
---|
| 4791 | Otherwise, the scaling is done in-situ |
---|
| 4792 | The default is taken from .asaprc (False) |
---|
[1846] | 4793 | |
---|
[513] | 4794 | """ |
---|
| 4795 | if insitu is None: insitu = rcParams['insitu'] |
---|
[876] | 4796 | self._math._setinsitu(insitu) |
---|
[513] | 4797 | varlist = vars() |
---|
[2954] | 4798 | s = scantable(self._math._unaryop(self, offset, "ADD", False, False)) |
---|
[1118] | 4799 | s._add_history("add", varlist) |
---|
[876] | 4800 | if insitu: |
---|
| 4801 | self._assign(s) |
---|
| 4802 | else: |
---|
[513] | 4803 | return s |
---|
| 4804 | |
---|
[1862] | 4805 | @asaplog_post_dec |
---|
[2952] | 4806 | def scale(self, factor, tsys=True, insitu=None, skip_flaggedrow=False): |
---|
[1846] | 4807 | """\ |
---|
| 4808 | |
---|
[1938] | 4809 | Return a scan where all spectra are scaled by the given 'factor' |
---|
[1846] | 4810 | |
---|
[513] | 4811 | Parameters: |
---|
[1846] | 4812 | |
---|
[1819] | 4813 | factor: the scaling factor (float or 1D float list) |
---|
[1855] | 4814 | |
---|
[513] | 4815 | insitu: if False a new scantable is returned. |
---|
| 4816 | Otherwise, the scaling is done in-situ |
---|
| 4817 | The default is taken from .asaprc (False) |
---|
[1855] | 4818 | |
---|
[513] | 4819 | tsys: if True (default) then apply the operation to Tsys |
---|
| 4820 | as well as the data |
---|
[1846] | 4821 | |
---|
[2952] | 4822 | skip_flaggedrow: if True, scaling is NOT executed for |
---|
| 4823 | row-flagged spectra. default is False. |
---|
[513] | 4824 | """ |
---|
| 4825 | if insitu is None: insitu = rcParams['insitu'] |
---|
[876] | 4826 | self._math._setinsitu(insitu) |
---|
[513] | 4827 | varlist = vars() |
---|
[1819] | 4828 | s = None |
---|
| 4829 | import numpy |
---|
| 4830 | if isinstance(factor, list) or isinstance(factor, numpy.ndarray): |
---|
[2320] | 4831 | if isinstance(factor[0], list) or isinstance(factor[0], |
---|
| 4832 | numpy.ndarray): |
---|
[1819] | 4833 | from asapmath import _array2dOp |
---|
[2952] | 4834 | s = _array2dOp( self, factor, "MUL", tsys, insitu, skip_flaggedrow ) |
---|
[1819] | 4835 | else: |
---|
[2320] | 4836 | s = scantable( self._math._arrayop( self, factor, |
---|
[2952] | 4837 | "MUL", tsys, skip_flaggedrow ) ) |
---|
[1819] | 4838 | else: |
---|
[2952] | 4839 | s = scantable(self._math._unaryop(self, factor, "MUL", tsys, skip_flaggedrow)) |
---|
[1118] | 4840 | s._add_history("scale", varlist) |
---|
[876] | 4841 | if insitu: |
---|
| 4842 | self._assign(s) |
---|
| 4843 | else: |
---|
[513] | 4844 | return s |
---|
| 4845 | |
---|
[2349] | 4846 | @preserve_selection |
---|
| 4847 | def set_sourcetype(self, match, matchtype="pattern", |
---|
[1504] | 4848 | sourcetype="reference"): |
---|
[1846] | 4849 | """\ |
---|
[1502] | 4850 | Set the type of the source to be an source or reference scan |
---|
[1846] | 4851 | using the provided pattern. |
---|
| 4852 | |
---|
[1502] | 4853 | Parameters: |
---|
[1846] | 4854 | |
---|
[1504] | 4855 | match: a Unix style pattern, regular expression or selector |
---|
[1855] | 4856 | |
---|
[1504] | 4857 | matchtype: 'pattern' (default) UNIX style pattern or |
---|
| 4858 | 'regex' regular expression |
---|
[1855] | 4859 | |
---|
[1502] | 4860 | sourcetype: the type of the source to use (source/reference) |
---|
[1846] | 4861 | |
---|
[1502] | 4862 | """ |
---|
| 4863 | varlist = vars() |
---|
| 4864 | stype = -1 |
---|
[2480] | 4865 | if sourcetype.lower().startswith("r") or sourcetype.lower() == "off": |
---|
[1502] | 4866 | stype = 1 |
---|
[2480] | 4867 | elif sourcetype.lower().startswith("s") or sourcetype.lower() == "on": |
---|
[1502] | 4868 | stype = 0 |
---|
[1504] | 4869 | else: |
---|
[2480] | 4870 | raise ValueError("Illegal sourcetype use s(ource)/on or r(eference)/off") |
---|
[1504] | 4871 | if matchtype.lower().startswith("p"): |
---|
| 4872 | matchtype = "pattern" |
---|
| 4873 | elif matchtype.lower().startswith("r"): |
---|
| 4874 | matchtype = "regex" |
---|
| 4875 | else: |
---|
| 4876 | raise ValueError("Illegal matchtype, use p(attern) or r(egex)") |
---|
[1502] | 4877 | sel = selector() |
---|
| 4878 | if isinstance(match, selector): |
---|
| 4879 | sel = match |
---|
| 4880 | else: |
---|
[2480] | 4881 | sel.set_query("SRCNAME=%s('%s')" % (matchtype, match)) |
---|
| 4882 | self.set_selection(sel) |
---|
[1502] | 4883 | self._setsourcetype(stype) |
---|
[1573] | 4884 | self._add_history("set_sourcetype", varlist) |
---|
[1502] | 4885 | |
---|
[2818] | 4886 | |
---|
| 4887 | def set_sourcename(self, name): |
---|
| 4888 | varlist = vars() |
---|
| 4889 | self._setsourcename(name) |
---|
| 4890 | self._add_history("set_sourcename", varlist) |
---|
| 4891 | |
---|
[1862] | 4892 | @asaplog_post_dec |
---|
[1857] | 4893 | @preserve_selection |
---|
[1819] | 4894 | def auto_quotient(self, preserve=True, mode='paired', verify=False): |
---|
[1846] | 4895 | """\ |
---|
[670] | 4896 | This function allows to build quotients automatically. |
---|
[1819] | 4897 | It assumes the observation to have the same number of |
---|
[670] | 4898 | "ons" and "offs" |
---|
[1846] | 4899 | |
---|
[670] | 4900 | Parameters: |
---|
[1846] | 4901 | |
---|
[710] | 4902 | preserve: you can preserve (default) the continuum or |
---|
| 4903 | remove it. The equations used are |
---|
[1857] | 4904 | |
---|
[670] | 4905 | preserve: Output = Toff * (on/off) - Toff |
---|
[1857] | 4906 | |
---|
[1070] | 4907 | remove: Output = Toff * (on/off) - Ton |
---|
[1855] | 4908 | |
---|
[1573] | 4909 | mode: the on/off detection mode |
---|
[1348] | 4910 | 'paired' (default) |
---|
| 4911 | identifies 'off' scans by the |
---|
| 4912 | trailing '_R' (Mopra/Parkes) or |
---|
| 4913 | '_e'/'_w' (Tid) and matches |
---|
| 4914 | on/off pairs from the observing pattern |
---|
[1502] | 4915 | 'time' |
---|
| 4916 | finds the closest off in time |
---|
[1348] | 4917 | |
---|
[1857] | 4918 | .. todo:: verify argument is not implemented |
---|
| 4919 | |
---|
[670] | 4920 | """ |
---|
[1857] | 4921 | varlist = vars() |
---|
[1348] | 4922 | modes = ["time", "paired"] |
---|
[670] | 4923 | if not mode in modes: |
---|
[876] | 4924 | msg = "please provide valid mode. Valid modes are %s" % (modes) |
---|
| 4925 | raise ValueError(msg) |
---|
[1348] | 4926 | s = None |
---|
| 4927 | if mode.lower() == "paired": |
---|
[2840] | 4928 | from asap._asap import srctype |
---|
[1857] | 4929 | sel = self.get_selection() |
---|
[2840] | 4930 | #sel.set_query("SRCTYPE==psoff") |
---|
| 4931 | sel.set_types(srctype.psoff) |
---|
[1356] | 4932 | self.set_selection(sel) |
---|
[1348] | 4933 | offs = self.copy() |
---|
[2840] | 4934 | #sel.set_query("SRCTYPE==pson") |
---|
| 4935 | sel.set_types(srctype.pson) |
---|
[1356] | 4936 | self.set_selection(sel) |
---|
[1348] | 4937 | ons = self.copy() |
---|
| 4938 | s = scantable(self._math._quotient(ons, offs, preserve)) |
---|
| 4939 | elif mode.lower() == "time": |
---|
| 4940 | s = scantable(self._math._auto_quotient(self, mode, preserve)) |
---|
[1118] | 4941 | s._add_history("auto_quotient", varlist) |
---|
[876] | 4942 | return s |
---|
[710] | 4943 | |
---|
[1862] | 4944 | @asaplog_post_dec |
---|
[1145] | 4945 | def mx_quotient(self, mask = None, weight='median', preserve=True): |
---|
[1846] | 4946 | """\ |
---|
[1143] | 4947 | Form a quotient using "off" beams when observing in "MX" mode. |
---|
[1846] | 4948 | |
---|
[1143] | 4949 | Parameters: |
---|
[1846] | 4950 | |
---|
[1145] | 4951 | mask: an optional mask to be used when weight == 'stddev' |
---|
[1855] | 4952 | |
---|
[1143] | 4953 | weight: How to average the off beams. Default is 'median'. |
---|
[1855] | 4954 | |
---|
[1145] | 4955 | preserve: you can preserve (default) the continuum or |
---|
[1855] | 4956 | remove it. The equations used are: |
---|
[1846] | 4957 | |
---|
[1855] | 4958 | preserve: Output = Toff * (on/off) - Toff |
---|
| 4959 | |
---|
| 4960 | remove: Output = Toff * (on/off) - Ton |
---|
| 4961 | |
---|
[1217] | 4962 | """ |
---|
[1593] | 4963 | mask = mask or () |
---|
[1141] | 4964 | varlist = vars() |
---|
| 4965 | on = scantable(self._math._mx_extract(self, 'on')) |
---|
[1143] | 4966 | preoff = scantable(self._math._mx_extract(self, 'off')) |
---|
| 4967 | off = preoff.average_time(mask=mask, weight=weight, scanav=False) |
---|
[1217] | 4968 | from asapmath import quotient |
---|
[1145] | 4969 | q = quotient(on, off, preserve) |
---|
[1143] | 4970 | q._add_history("mx_quotient", varlist) |
---|
[1217] | 4971 | return q |
---|
[513] | 4972 | |
---|
[1862] | 4973 | @asaplog_post_dec |
---|
[718] | 4974 | def freq_switch(self, insitu=None): |
---|
[1846] | 4975 | """\ |
---|
[718] | 4976 | Apply frequency switching to the data. |
---|
[1846] | 4977 | |
---|
[718] | 4978 | Parameters: |
---|
[1846] | 4979 | |
---|
[718] | 4980 | insitu: if False a new scantable is returned. |
---|
| 4981 | Otherwise, the swictching is done in-situ |
---|
| 4982 | The default is taken from .asaprc (False) |
---|
[1846] | 4983 | |
---|
[718] | 4984 | """ |
---|
| 4985 | if insitu is None: insitu = rcParams['insitu'] |
---|
[876] | 4986 | self._math._setinsitu(insitu) |
---|
[718] | 4987 | varlist = vars() |
---|
[876] | 4988 | s = scantable(self._math._freqswitch(self)) |
---|
[1118] | 4989 | s._add_history("freq_switch", varlist) |
---|
[1856] | 4990 | if insitu: |
---|
| 4991 | self._assign(s) |
---|
| 4992 | else: |
---|
| 4993 | return s |
---|
[718] | 4994 | |
---|
[1862] | 4995 | @asaplog_post_dec |
---|
[780] | 4996 | def recalc_azel(self): |
---|
[1846] | 4997 | """Recalculate the azimuth and elevation for each position.""" |
---|
[780] | 4998 | varlist = vars() |
---|
[876] | 4999 | self._recalcazel() |
---|
[780] | 5000 | self._add_history("recalc_azel", varlist) |
---|
| 5001 | return |
---|
| 5002 | |
---|
[1862] | 5003 | @asaplog_post_dec |
---|
[513] | 5004 | def __add__(self, other): |
---|
[2574] | 5005 | """ |
---|
| 5006 | implicit on all axes and on Tsys |
---|
| 5007 | """ |
---|
[513] | 5008 | varlist = vars() |
---|
[2574] | 5009 | s = self.__op( other, "ADD" ) |
---|
[513] | 5010 | s._add_history("operator +", varlist) |
---|
| 5011 | return s |
---|
| 5012 | |
---|
[1862] | 5013 | @asaplog_post_dec |
---|
[513] | 5014 | def __sub__(self, other): |
---|
| 5015 | """ |
---|
| 5016 | implicit on all axes and on Tsys |
---|
| 5017 | """ |
---|
| 5018 | varlist = vars() |
---|
[2574] | 5019 | s = self.__op( other, "SUB" ) |
---|
[513] | 5020 | s._add_history("operator -", varlist) |
---|
| 5021 | return s |
---|
[710] | 5022 | |
---|
[1862] | 5023 | @asaplog_post_dec |
---|
[513] | 5024 | def __mul__(self, other): |
---|
| 5025 | """ |
---|
| 5026 | implicit on all axes and on Tsys |
---|
| 5027 | """ |
---|
| 5028 | varlist = vars() |
---|
[2574] | 5029 | s = self.__op( other, "MUL" ) ; |
---|
[513] | 5030 | s._add_history("operator *", varlist) |
---|
| 5031 | return s |
---|
| 5032 | |
---|
[710] | 5033 | |
---|
[1862] | 5034 | @asaplog_post_dec |
---|
[513] | 5035 | def __div__(self, other): |
---|
| 5036 | """ |
---|
| 5037 | implicit on all axes and on Tsys |
---|
| 5038 | """ |
---|
| 5039 | varlist = vars() |
---|
[2574] | 5040 | s = self.__op( other, "DIV" ) |
---|
| 5041 | s._add_history("operator /", varlist) |
---|
| 5042 | return s |
---|
| 5043 | |
---|
| 5044 | @asaplog_post_dec |
---|
| 5045 | def __op( self, other, mode ): |
---|
[513] | 5046 | s = None |
---|
| 5047 | if isinstance(other, scantable): |
---|
[2574] | 5048 | s = scantable(self._math._binaryop(self, other, mode)) |
---|
[513] | 5049 | elif isinstance(other, float): |
---|
| 5050 | if other == 0.0: |
---|
[718] | 5051 | raise ZeroDivisionError("Dividing by zero is not recommended") |
---|
[2954] | 5052 | s = scantable(self._math._unaryop(self, other, mode, False, False)) |
---|
[2144] | 5053 | elif isinstance(other, list) or isinstance(other, numpy.ndarray): |
---|
[2349] | 5054 | if isinstance(other[0], list) \ |
---|
| 5055 | or isinstance(other[0], numpy.ndarray): |
---|
[2144] | 5056 | from asapmath import _array2dOp |
---|
[2574] | 5057 | s = _array2dOp( self, other, mode, False ) |
---|
[2144] | 5058 | else: |
---|
[2574] | 5059 | s = scantable( self._math._arrayop( self, other, |
---|
[2954] | 5060 | mode, False, False ) ) |
---|
[513] | 5061 | else: |
---|
[718] | 5062 | raise TypeError("Other input is not a scantable or float value") |
---|
[513] | 5063 | return s |
---|
| 5064 | |
---|
[1862] | 5065 | @asaplog_post_dec |
---|
[530] | 5066 | def get_fit(self, row=0): |
---|
[1846] | 5067 | """\ |
---|
[530] | 5068 | Print or return the stored fits for a row in the scantable |
---|
[1846] | 5069 | |
---|
[530] | 5070 | Parameters: |
---|
[1846] | 5071 | |
---|
[530] | 5072 | row: the row which the fit has been applied to. |
---|
[1846] | 5073 | |
---|
[530] | 5074 | """ |
---|
| 5075 | if row > self.nrow(): |
---|
| 5076 | return |
---|
[976] | 5077 | from asap.asapfit import asapfit |
---|
[530] | 5078 | fit = asapfit(self._getfit(row)) |
---|
[1859] | 5079 | asaplog.push( '%s' %(fit) ) |
---|
| 5080 | return fit.as_dict() |
---|
[530] | 5081 | |
---|
[2349] | 5082 | @preserve_selection |
---|
[1483] | 5083 | def flag_nans(self): |
---|
[1846] | 5084 | """\ |
---|
[1483] | 5085 | Utility function to flag NaN values in the scantable. |
---|
| 5086 | """ |
---|
| 5087 | import numpy |
---|
| 5088 | basesel = self.get_selection() |
---|
| 5089 | for i in range(self.nrow()): |
---|
[1589] | 5090 | sel = self.get_row_selector(i) |
---|
| 5091 | self.set_selection(basesel+sel) |
---|
[1483] | 5092 | nans = numpy.isnan(self._getspectrum(0)) |
---|
[2877] | 5093 | if numpy.any(nans): |
---|
| 5094 | bnans = [ bool(v) for v in nans] |
---|
| 5095 | self.flag(bnans) |
---|
| 5096 | |
---|
| 5097 | self.set_selection(basesel) |
---|
[1483] | 5098 | |
---|
[1588] | 5099 | def get_row_selector(self, rowno): |
---|
[1992] | 5100 | return selector(rows=[rowno]) |
---|
[1573] | 5101 | |
---|
[484] | 5102 | def _add_history(self, funcname, parameters): |
---|
[1435] | 5103 | if not rcParams['scantable.history']: |
---|
| 5104 | return |
---|
[484] | 5105 | # create date |
---|
| 5106 | sep = "##" |
---|
| 5107 | from datetime import datetime |
---|
| 5108 | dstr = datetime.now().strftime('%Y/%m/%d %H:%M:%S') |
---|
| 5109 | hist = dstr+sep |
---|
| 5110 | hist += funcname+sep#cdate+sep |
---|
[2349] | 5111 | if parameters.has_key('self'): |
---|
| 5112 | del parameters['self'] |
---|
[1118] | 5113 | for k, v in parameters.iteritems(): |
---|
[484] | 5114 | if type(v) is dict: |
---|
[1118] | 5115 | for k2, v2 in v.iteritems(): |
---|
[484] | 5116 | hist += k2 |
---|
| 5117 | hist += "=" |
---|
[1118] | 5118 | if isinstance(v2, scantable): |
---|
[484] | 5119 | hist += 'scantable' |
---|
| 5120 | elif k2 == 'mask': |
---|
[1118] | 5121 | if isinstance(v2, list) or isinstance(v2, tuple): |
---|
[513] | 5122 | hist += str(self._zip_mask(v2)) |
---|
| 5123 | else: |
---|
| 5124 | hist += str(v2) |
---|
[484] | 5125 | else: |
---|
[513] | 5126 | hist += str(v2) |
---|
[484] | 5127 | else: |
---|
| 5128 | hist += k |
---|
| 5129 | hist += "=" |
---|
[1118] | 5130 | if isinstance(v, scantable): |
---|
[484] | 5131 | hist += 'scantable' |
---|
| 5132 | elif k == 'mask': |
---|
[1118] | 5133 | if isinstance(v, list) or isinstance(v, tuple): |
---|
[513] | 5134 | hist += str(self._zip_mask(v)) |
---|
| 5135 | else: |
---|
| 5136 | hist += str(v) |
---|
[484] | 5137 | else: |
---|
| 5138 | hist += str(v) |
---|
| 5139 | hist += sep |
---|
| 5140 | hist = hist[:-2] # remove trailing '##' |
---|
| 5141 | self._addhistory(hist) |
---|
| 5142 | |
---|
[710] | 5143 | |
---|
[484] | 5144 | def _zip_mask(self, mask): |
---|
| 5145 | mask = list(mask) |
---|
| 5146 | i = 0 |
---|
| 5147 | segments = [] |
---|
| 5148 | while mask[i:].count(1): |
---|
| 5149 | i += mask[i:].index(1) |
---|
| 5150 | if mask[i:].count(0): |
---|
| 5151 | j = i + mask[i:].index(0) |
---|
| 5152 | else: |
---|
[710] | 5153 | j = len(mask) |
---|
[1118] | 5154 | segments.append([i, j]) |
---|
[710] | 5155 | i = j |
---|
[484] | 5156 | return segments |
---|
[714] | 5157 | |
---|
[626] | 5158 | def _get_ordinate_label(self): |
---|
| 5159 | fu = "("+self.get_fluxunit()+")" |
---|
| 5160 | import re |
---|
| 5161 | lbl = "Intensity" |
---|
[1118] | 5162 | if re.match(".K.", fu): |
---|
[626] | 5163 | lbl = "Brightness Temperature "+ fu |
---|
[1118] | 5164 | elif re.match(".Jy.", fu): |
---|
[626] | 5165 | lbl = "Flux density "+ fu |
---|
| 5166 | return lbl |
---|
[710] | 5167 | |
---|
[876] | 5168 | def _check_ifs(self): |
---|
[2349] | 5169 | # return len(set([self.nchan(i) for i in self.getifnos()])) == 1 |
---|
[1986] | 5170 | nchans = [self.nchan(i) for i in self.getifnos()] |
---|
[2004] | 5171 | nchans = filter(lambda t: t > 0, nchans) |
---|
[876] | 5172 | return (sum(nchans)/len(nchans) == nchans[0]) |
---|
[976] | 5173 | |
---|
[1862] | 5174 | @asaplog_post_dec |
---|
[1916] | 5175 | def _fill(self, names, unit, average, opts={}): |
---|
[976] | 5176 | first = True |
---|
| 5177 | fullnames = [] |
---|
| 5178 | for name in names: |
---|
| 5179 | name = os.path.expandvars(name) |
---|
| 5180 | name = os.path.expanduser(name) |
---|
| 5181 | if not os.path.exists(name): |
---|
| 5182 | msg = "File '%s' does not exists" % (name) |
---|
| 5183 | raise IOError(msg) |
---|
| 5184 | fullnames.append(name) |
---|
| 5185 | if average: |
---|
| 5186 | asaplog.push('Auto averaging integrations') |
---|
[1079] | 5187 | stype = int(rcParams['scantable.storage'].lower() == 'disk') |
---|
[976] | 5188 | for name in fullnames: |
---|
[1073] | 5189 | tbl = Scantable(stype) |
---|
[2004] | 5190 | if is_ms( name ): |
---|
| 5191 | r = msfiller( tbl ) |
---|
| 5192 | else: |
---|
| 5193 | r = filler( tbl ) |
---|
[976] | 5194 | msg = "Importing %s..." % (name) |
---|
[1118] | 5195 | asaplog.push(msg, False) |
---|
[2349] | 5196 | r.open(name, opts) |
---|
[2480] | 5197 | rx = rcParams['scantable.reference'] |
---|
| 5198 | r.setreferenceexpr(rx) |
---|
[1843] | 5199 | r.fill() |
---|
[976] | 5200 | if average: |
---|
[1118] | 5201 | tbl = self._math._average((tbl, ), (), 'NONE', 'SCAN') |
---|
[976] | 5202 | if not first: |
---|
[2902] | 5203 | tbl = self._math._merge([self, tbl]) |
---|
[976] | 5204 | Scantable.__init__(self, tbl) |
---|
[1843] | 5205 | r.close() |
---|
[1118] | 5206 | del r, tbl |
---|
[976] | 5207 | first = False |
---|
[1861] | 5208 | #flush log |
---|
| 5209 | asaplog.post() |
---|
[976] | 5210 | if unit is not None: |
---|
| 5211 | self.set_fluxunit(unit) |
---|
[1824] | 5212 | if not is_casapy(): |
---|
| 5213 | self.set_freqframe(rcParams['scantable.freqframe']) |
---|
[976] | 5214 | |
---|
[2610] | 5215 | def _get_verify_action( self, msg, action=None ): |
---|
| 5216 | valid_act = ['Y', 'N', 'A', 'R'] |
---|
| 5217 | if not action or not isinstance(action, str): |
---|
| 5218 | action = raw_input("%s [Y/n/a/r] (h for help): " % msg) |
---|
| 5219 | if action == '': |
---|
| 5220 | return "Y" |
---|
| 5221 | elif (action.upper()[0] in valid_act): |
---|
| 5222 | return action.upper()[0] |
---|
| 5223 | elif (action.upper()[0] in ['H','?']): |
---|
| 5224 | print "Available actions of verification [Y|n|a|r]" |
---|
| 5225 | print " Y : Yes for current data (default)" |
---|
| 5226 | print " N : No for current data" |
---|
| 5227 | print " A : Accept all in the following and exit from verification" |
---|
| 5228 | print " R : Reject all in the following and exit from verification" |
---|
| 5229 | print " H or ?: help (show this message)" |
---|
| 5230 | return self._get_verify_action(msg) |
---|
| 5231 | else: |
---|
| 5232 | return 'Y' |
---|
[2012] | 5233 | |
---|
[1402] | 5234 | def __getitem__(self, key): |
---|
| 5235 | if key < 0: |
---|
| 5236 | key += self.nrow() |
---|
| 5237 | if key >= self.nrow(): |
---|
| 5238 | raise IndexError("Row index out of range.") |
---|
| 5239 | return self._getspectrum(key) |
---|
| 5240 | |
---|
| 5241 | def __setitem__(self, key, value): |
---|
| 5242 | if key < 0: |
---|
| 5243 | key += self.nrow() |
---|
| 5244 | if key >= self.nrow(): |
---|
| 5245 | raise IndexError("Row index out of range.") |
---|
| 5246 | if not hasattr(value, "__len__") or \ |
---|
| 5247 | len(value) > self.nchan(self.getif(key)): |
---|
| 5248 | raise ValueError("Spectrum length doesn't match.") |
---|
| 5249 | return self._setspectrum(value, key) |
---|
| 5250 | |
---|
| 5251 | def __len__(self): |
---|
| 5252 | return self.nrow() |
---|
| 5253 | |
---|
| 5254 | def __iter__(self): |
---|
| 5255 | for i in range(len(self)): |
---|
| 5256 | yield self[i] |
---|