multichannel_datasource.py 34.4 KB
 Patrick Godwin committed Feb 14, 2017 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 # # Copyright (C) 2017 Sydney J. Chamberlin, Patrick Godwin, Chad Hanna # # This program is free software; you can redistribute it and/or modify it # under the terms of the GNU General Public License as published by the # Free Software Foundation; either version 2 of the License, or (at your # option) any later version. # # This program is distributed in the hope that it will be useful, but # WITHOUT ANY WARRANTY; without even the implied warranty of # MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU General # Public License for more details. # # You should have received a copy of the GNU General Public License along # with this program; if not, write to the Free Software Foundation, Inc., # 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301, USA. # # ============================================================================= # # Preamble # # ============================================================================= # import sys import time  29 import itertools  Patrick Godwin committed Feb 27, 2017 30 31 import optparse from ConfigParser import SafeConfigParser  Patrick Godwin committed Feb 14, 2017 32 33 34 35 36 37 38 39 40 41  import gi gi.require_version('Gst', '1.0') from gi.repository import GObject from gi.repository import Gst GObject.threads_init() Gst.init(None) from gstlal import pipeparts from gstlal import datasource  Kipp Cannon committed Feb 14, 2019 42 43 from ligo.lw import utils as ligolw_utils from ligo.lw.utils import segments as ligolw_segments  Duncan Meacher committed Jul 11, 2018 44 from ligo import segments  Patrick Godwin committed Feb 14, 2017 45 46 47 import lal from lal import LIGOTimeGPS  48 import numpy  Patrick Godwin committed Feb 14, 2017 49 50 51 52 53  ## framexmit ports in use on the LDG # Look-up table to map instrument name to framexmit multicast address and # port #  Patrick Godwin committed Feb 27, 2017 54 # used in mkbasicmultisrc()  Patrick Godwin committed Feb 14, 2017 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 # # FIXME: this is only here temporarily while we test this approach to data # aquisition. obviously we can't hard-code this stuff # framexmit_ports = { "CIT": { "H1": ("224.3.2.1", 7096), "L1": ("224.3.2.2", 7097), "V1": ("224.3.2.3", 7098), } } # # misc useful functions #  Duncan Meacher committed Sep 21, 2017 71 72 73 74 75 76 77 78 79 def channel_dict_from_channel_list(channel_list): """! Given a list of channels, produce a dictionary keyed by channel names: The list here typically comes from an option parser with options that specify the "append" action. Examples:  Patrick Godwin committed Jul 05, 2018 80  >>> channel_dict_from_channel_list(["H1:AUX-CHANNEL-NAME_1:2048", "H1:AUX-CHANNEL-NAME-2:512"])  Duncan Meacher committed Sep 21, 2017 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95  {'H1:AUX-CHANNEL-NAME_1': {'qhigh': None, 'ifo': 'H1', 'flow': None, 'fsamp': 2048.0, 'fhigh': None, 'frametype': None}, 'H1:AUX-CHANNEL-NAME-2': {'qhigh': None, 'ifo': 'H1', 'flow': None, 'fsamp': 512.0, 'fhigh': None, 'frametype': None}} """ channel_dict = {} for channel in channel_list: ifo, channel_info, fsamp = channel.split(':') channel_name = ifo + ":" + channel_info channel_dict[channel_name] = {'fsamp': float(fsamp), 'ifo': ifo, 'flow': None, 'fhigh': None, 'qhigh' : None, 'frametype' : None} return channel_dict  Patrick Godwin committed Feb 14, 2017 96 97 98 99 100 101 102 def channel_dict_from_channel_file(channel_file): """! Given a file of channel names with sampling rates, produce a dictionary keyed by ifo: The file here comes from the output of a configuration file parser. """  Patrick Godwin committed Mar 29, 2017 103  channel_dict = {}  Patrick Godwin committed Feb 14, 2017 104 105  channel_list = open(channel_file) for channel in channel_list:  Patrick Godwin committed Mar 29, 2017 106  channel_name, fsamp = channel.split()  Patrick Godwin committed Feb 14, 2017 107  ifo, channel_info = channel.split(':')  Patrick Godwin committed Mar 29, 2017 108 109 110 111 112 113 114  channel_dict[channel_name] = {'fsamp': float(fsamp), 'ifo': ifo, 'flow': None, 'fhigh': None, 'qhigh' : None, 'frametype' : None}  Patrick Godwin committed Feb 14, 2017 115  channel_list.close()  Patrick Godwin committed Mar 29, 2017 116  return channel_dict  Patrick Godwin committed Feb 14, 2017 117   Patrick Godwin committed Feb 27, 2017 118 119 120 121 122 def channel_dict_from_channel_ini(options): """! Given a channel list INI, produces a dictionary keyed by ifo, filtered by frame type. """  Patrick Godwin committed Mar 29, 2017 123  channel_dict = {}  Patrick Godwin committed Feb 27, 2017 124   125  # frame types considered  126  included_frame_types = set(("H1_R", "L1_R", "H1_lldetchar", "L1_lldetchar"))  127   Patrick Godwin committed Feb 27, 2017 128  # known/permissible values of safety and fidelity flags  129  known_safety = set(("safe", "unsafe", "unsafeabove2kHz", "unknown"))  Patrick Godwin committed Feb 27, 2017 130 131 132 133 134 135 136 137 138  known_fidelity = set(("clean", "flat", "glitchy", "unknown")) # read in channel list config = SafeConfigParser() config.read(options.channel_list) # filter out channels by frame type sections = [] for name in config.sections():  139  if config.get(name, 'frametype') in included_frame_types:  Patrick Godwin committed Feb 27, 2017 140 141  sections.append(name)  Patrick Godwin committed Sep 12, 2018 142 143 144 145 146 147 148 149  # specify which channels are considered if options.section_include: section_include = [section.replace('_', ' ') for section in options.section_include] else: section_include = sections channel_include = options.safe_channel_include + options.unsafe_channel_include  Patrick Godwin committed Feb 27, 2017 150 151  # generate dictionary of channels for name in sections:  Patrick Godwin committed Sep 12, 2018 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198  # extract low frequency, high Q flow = config.getfloat(name, 'flow') qhigh = config.getfloat(name, 'qhigh') # figure out whether to use Nyquist for each channel or a specific limit fhigh = config.get(name, 'fhigh') use_nyquist = fhigh == "Nyquist" if not use_nyquist: fhigh = float(fhigh) # set up each channel for channel in config.get(name, 'channels').strip().split('\n'): # parse out expected format for each channel channel = channel.split() if len(channel)==2: # backward compatibility with old format channel, fsamp = channel fsamp = int(fsamp) safety = "unknown" fidelity = "unknown" elif len(channel)==4: # expected format channel, fsamp, safety, fidelity = channel fsamp = int(fsamp) else: raise SyntaxError( 'could not parse channel : %s'%(''.join(channel)) ) #----------------------------------------- ### check that safety and fidelity are permissible values assert safety in known_safety, 'safety=%s is not understood. Must be one of %s'%(safety, ", ".join(known_safety)) assert fidelity in known_fidelity, 'fidelity=%s is not understood. Must be one of %s'%(fidelity, ", ".join(known_fidelity)) # conditions on whether or not we want to include this channel if name in section_include or channel in channel_include: if (safety in options.safety_include and fidelity not in options.fidelity_exclude) or channel in options.unsafe_channel_include: # add ifo, channel name & omicron parameters to dict channel_name = channel ifo,_ = channel.split(':') if use_nyquist: fhigh = fsamp/2. channel_dict[channel_name] = {'fsamp': fsamp, 'ifo': ifo, 'flow': flow, 'fhigh': fhigh, 'qhigh' : qhigh}  Patrick Godwin committed Feb 14, 2017 199   Patrick Godwin committed Mar 29, 2017 200  return channel_dict  Patrick Godwin committed Feb 27, 2017 201   202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262  def partition_channels_to_equal_subsets(channel_dict, max_streams, min_sample_rate, max_sample_rate): """! Given a channel dictionary, will produce partitions of channel subsets where the number of channels in each partition is equal (except possibly the last partition). This is given max_streams, and well as max and min sample rates enforced to determine the number of streams that a particular channel will generate. Returns a list of disjoint channel lists. """ # determine how many streams a single channel will produce when split up into multiple frequency bands # and separate them based on this criterion channel_streams = {} for channel in channel_dict.keys(): sample_rate = int(channel_dict[channel]['fsamp']) max_rate = min(max_sample_rate, sample_rate) min_rate = min(min_sample_rate, max_rate) n_rates = int(numpy.log2(max_rate/min_rate) + 1) channel_streams.setdefault(n_rates, []).append((n_rates, channel)) # find relative probabilities in each bin total = sum((len(channel_streams[n]) for n in channel_streams.keys())) p_relative = {n: (len(channel_streams[n]) / float(total)) for n in channel_streams.keys()} # figure out total number of channels needed per subset num_streams = {n: int(numpy.ceil(p_relative[n] * max_streams)) for n in channel_streams.keys()} num_channels = {n: int(numpy.ceil(num_streams[n] / float(n))) for n in num_streams.keys()} # force less sampling from the lowest bins (16Hz and 32Hz) at the beginning # to reduce the number of channels in each subset rates = sorted(channel_streams.keys()) if rates[0] == 1: num_channels[1] = 1 max_channels = sum((num_channels[n] for n in num_channels.keys())) # generate a round-robin type way to sample from rates2sample = itertools.cycle(n for n in channel_streams.keys() for i in range(int(numpy.round(p_relative[n] * max_channels)))) # generate channel subsets subsets = [] total = sum((len(channel_streams[n]) for n in channel_streams.keys())) while total > 0: subset = [] while len(subset) < max_channels and total > 0: rate = next(rates2sample) while not channel_streams[rate]: # recalculate probabilities and rates2sample p_relative = {n: (len(channel_streams[n]) / float(total)) for n in channel_streams.keys()} rates2sample = itertools.cycle(n for n in channel_streams.keys() for i in range(int(numpy.round(p_relative[n] * max_channels)))) rate = next(rates2sample) subset.append(channel_streams[rate].pop()[1]) total -= 1 subsets.append(subset) return subsets  263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 def partition_channels_to_subsets(channel_dict, max_streams, min_sample_rate, max_sample_rate): """! Given a channel dictionary, will produce roughly equal partitions of channel subsets, given max_streams, and well as max and min sample rates enforced to determine the number of streams that a particular channel will generate. Returns a list of disjoint channel lists. """ # determine how many streams a single channel will produce when split up into multiple frequency bands channel_streams = [] for channel in channel_dict.keys(): sample_rate = int(channel_dict[channel]['fsamp']) max_rate = min(max_sample_rate, sample_rate) min_rate = min(min_sample_rate, max_rate) n_rates = int(numpy.log2(max_rate/min_rate) + 1) channel_streams.append((n_rates, channel)) return [subset for subset in partition_list(channel_streams, max_streams)] def partition_list(lst, target_sum): """! Partition list to roughly equal partitioned chunks based on a target sum, given a list with items in the form (int, value), where ints are used to determine partitions. Returns a sublist with items value. """ total_sum = sum(item[0] for item in lst) chunks = numpy.ceil(total_sum/float(target_sum)) avg_sum = total_sum/float(chunks) chunks_yielded = 0 chunk = [] chunksum = 0 sum_of_seen = 0 for i, item in enumerate(lst): # if only one chunk left to process, yield rest of list if chunks - chunks_yielded == 1: yield chunk + [x[1] for x in lst[i:]] raise StopIteration to_yield = chunks - chunks_yielded chunks_left = len(lst) - i # yield remaining list in single item chunks if to_yield > chunks_left: if chunk: yield chunk for x in lst[i:]: yield [x[1]] raise StopIteration sum_of_seen += item[0] # if target sum is less than the average, add another item to chunk if chunksum < avg_sum: chunk.append(item[1]) chunksum += item[0] # else, yield the chunk, and update expected sum since this chunk isn't perfectly partitioned else: yield chunk avg_sum = (total_sum - sum_of_seen)/(to_yield - 1) chunks_yielded += 1 chunksum = item[0] chunk = [item[1]]  Patrick Godwin committed Feb 27, 2017 332   Patrick Godwin committed Feb 14, 2017 333 334 335 336 337 338 339 340 341 342 343 class DataSourceInfo(object): """! Hold the data associated with data source command lines. """ ## See datasource.append_options() def __init__(self, options): """! Initialize a DataSourceInfo class instance from command line options specified by append_options() """ ## A list of possible, valid data sources ("frames", "framexmit", "lvshm", "white", "silence")  344 345  self.data_sources = set(("framexmit", "lvshm", "frames", "white", "silence", "white_live")) self.live_sources = set(("framexmit", "lvshm", "white_live"))  Patrick Godwin committed Feb 14, 2017 346 347 348 349 350  assert self.live_sources <= self.data_sources # Sanity check the options if options.data_source not in self.data_sources: raise ValueError("--data-source must be one of %s" % ", ".join(self.data_sources))  Patrick Godwin committed Feb 27, 2017 351 352 353 354 355 356  if options.data_source == "frames" and options.frame_cache is None: raise ValueError("--frame-cache must be specified when using --data-source=frames") if options.frame_segments_file is not None and options.data_source != "frames": raise ValueError("can only give --frame-segments-file if --data-source=frames") if options.frame_segments_name is not None and options.frame_segments_file is None: raise ValueError("can only specify --frame-segments-name if --frame-segments-file is given")  Duncan Meacher committed Sep 21, 2017 357 358 359 360  if not (options.channel_list or options.channel_name): raise ValueError("must specify a channel list in the form --channel-list=/path/to/file or --channel-name=H1:AUX-CHANNEL-NAME:RATE --channel-name=H1:SOMETHING-ELSE:RATE") if (options.channel_list and options.channel_name): raise ValueError("must specify a channel list in the form --channel-list=/path/to/file or --channel-name=H1:AUX-CHANNEL-NAME:RATE --channel-name=H1:SOMETHING-ELSE:RATE")  Patrick Godwin committed Feb 14, 2017 361   Patrick Godwin committed Feb 27, 2017 362 363 364  ## Generate a dictionary of requested channels from channel INI file # known/permissible values of safety and fidelity flags  365  self.known_safety = set(("safe", "unsafe", "unsafeabove2kHz", "unknown"))  Patrick Godwin committed Feb 27, 2017 366 367  self.known_fidelity = set(("clean", "flat", "glitchy", "unknown"))  368 369 370 371 372 373  # ensure safety and fidelity options are valid options.safety_include = set(options.safety_include) options.fidelity_exclude = set(options.fidelity_exclude) for safety in options.safety_include: assert safety in self.known_safety, '--safety-include=%s is not understood. Must be one of %s'%(safety, ", ".join(self.known_safety))  Patrick Godwin committed Feb 27, 2017 374 375 376 377  for fidelity in options.fidelity_exclude: assert fidelity in self.known_fidelity, '--fidelity-exclude=%s is not understood. Must be one of %s'%(fidelity, ", ".join(self.known_fidelity))  378  # dictionary of the requested channels, e.g., {"H1:LDAS-STRAIN": 16384, "H1:ODC-LARM": 2048}  Duncan Meacher committed Sep 21, 2017 379 380 381 382 383 384 385 386 387 388  if options.channel_list: name, self.extension = options.channel_list.rsplit('.', 1) if self.extension == 'ini': self.channel_dict = channel_dict_from_channel_ini(options) else: self.channel_dict = channel_dict_from_channel_file(options.channel_list) elif options.channel_name: self.extension = 'none' self.channel_dict = channel_dict_from_channel_list(options.channel_name)  389  # set instrument; it is assumed all channels from a given channel list are from the same instrument  Patrick Godwin committed Mar 29, 2017 390 391  self.instrument = self.channel_dict[next(iter(self.channel_dict))]['ifo']  392 393 394 395 396 397 398 399  # set the maximum number of streams to be run by a single pipeline. self.max_streams = options.max_streams # set the frequency ranges considered by channels with splitting into multiple frequency bands. # If channel sampling rate doesn't fall within this range, it will not be split into multiple bands. self.max_sample_rate = options.max_sample_rate self.min_sample_rate = options.min_sample_rate  400 401 402 403 404  # split up channels requested into partitions for serial processing if options.equal_subsets: self.channel_subsets = partition_channels_to_equal_subsets(self.channel_dict, self.max_streams, self.min_sample_rate, self.max_sample_rate) else: self.channel_subsets = partition_channels_to_subsets(self.channel_dict, self.max_streams, self.min_sample_rate, self.max_sample_rate)  405   Patrick Godwin committed Feb 14, 2017 406 407 408 409 410  ## A dictionary for shared memory partition, e.g., {"H1": "LHO_Data", "H2": "LHO_Data", "L1": "LLO_Data", "V1": "VIRGO_Data"} self.shm_part_dict = {"H1": "LHO_Data", "H2": "LHO_Data", "L1": "LLO_Data", "V1": "VIRGO_Data"} if options.shared_memory_partition is not None: self.shm_part_dict.update( datasource.channel_dict_from_channel_list(options.shared_memory_partition) )  Patrick Godwin committed Jul 02, 2018 411 412 413 414  ## options for shared memory self.shm_assumed_duration = options.shared_memory_assumed_duration self.shm_block_size = options.shared_memory_block_size # NOTE: should this be incorporated into options.block_size? currently only used for offline data sources  Patrick Godwin committed Feb 14, 2017 415 416 417 418 419 420 421 422  ## A dictionary of framexmit addresses self.framexmit_addr = framexmit_ports["CIT"] if options.framexmit_addr is not None: self.framexmit_addr.update( datasource.framexmit_dict_from_framexmit_list(options.framexmit_addr) ) self.framexmit_iface = options.framexmit_iface ## Analysis segment. Default is None self.seg = None  Patrick Godwin committed May 16, 2017 423 424 425  ## Set latency output self.latency_output = options.latency_output  Patrick Godwin committed Feb 14, 2017 426 427 428 429 430 431 432 433 434 435 436 437 438 439 440 441 442 443 444 445  if options.gps_start_time is not None: if options.gps_end_time is None: raise ValueError("must provide both --gps-start-time and --gps-end-time") try: start = LIGOTimeGPS(options.gps_start_time) except ValueError: raise ValueError("invalid --gps-start-time '%s'" % options.gps_start_time) try: end = LIGOTimeGPS(options.gps_end_time) except ValueError: raise ValueError("invalid --gps-end-time '%s'" % options.gps_end_time) if start >= end: raise ValueError("--gps-start-time must be < --gps-end-time: %s < %s" % (options.gps_start_time, options.gps_end_time)) ## Segment from gps start and stop time if given self.seg = segments.segment(LIGOTimeGPS(options.gps_start_time), LIGOTimeGPS(options.gps_end_time)) elif options.gps_end_time is not None: raise ValueError("must provide both --gps-start-time and --gps-end-time") elif options.data_source not in self.live_sources: raise ValueError("--gps-start-time and --gps-end-time must be specified when --data-source not one of %s" % ", ".join(sorted(self.live_sources)))  Patrick Godwin committed Feb 27, 2017 446 447 448 449 450 451 452 453 454 455 456  if options.frame_segments_file is not None: ## Frame segments from a user defined file self.frame_segments = ligolw_segments.segmenttable_get_by_name(ligolw_utils.load_filename(options.frame_segments_file, contenthandler=ligolw_segments.LIGOLWContentHandler), options.frame_segments_name).coalesce() if self.seg is not None: # Clip frame segments to seek segment if it # exists (not required, just saves some # memory and I/O overhead) self.frame_segments = segments.segmentlistdict((instrument, seglist & segments.segmentlist([self.seg])) for instrument, seglist in self.frame_segments.items()) else: ## if no frame segments provided, set them to an empty segment list dictionary  Patrick Godwin committed Mar 29, 2017 457  self.frame_segments = segments.segmentlistdict({self.instrument: None})  Patrick Godwin committed Feb 27, 2017 458 459  ## frame cache file  Patrick Godwin committed Mar 16, 2017 460  self.frame_cache = options.frame_cache  Patrick Godwin committed Feb 14, 2017 461 462 463 464 465  ## block size in bytes to read data from disk self.block_size = options.block_size ## Data source, one of python.datasource.DataSourceInfo.data_sources self.data_source = options.data_source  466 467 468  # FIXME: this is ugly, but we have to protect against busted shared memory partitions if self.data_source == "lvshm": import subprocess  Patrick Godwin committed Jun 13, 2019 469  subprocess.call(["smrepair", "--bufmode", "5", self.shm_part_dict[self.instrument]])  470   Patrick Godwin committed Feb 14, 2017 471 472 473 474 475 476 477 def append_options(parser): """! Append generic data source options to an OptionParser object in order to have consistent an unified command lines and parsing throughout the project for applications that read GW data. - --data-source [string]  478  Set the data source from [framexmit|lvshm|frames|silence|white|white_live].  Patrick Godwin committed Feb 14, 2017 479 480 481 482 483 484 485 486 487 488 489 490 491 492 493 494 495 496 497  - --block-size [int] (bytes) Data block size to read in bytes. Default 16384 * 8 * 512 which is 512 seconds of double precision data at 16384 Hz. This parameter is only used if --data-source is one of white, silence, AdvVirgo, LIGO, AdvLIGO, nds. - --gps-start-time [int] (seconds) Set the start time of the segment to analyze in GPS seconds. Required unless --data-source in lvshm,framexmit - --gps-end-time [int] (seconds) Set the end time of the segment to analyze in GPS seconds. Required unless --data-source in lvshm,framexmit - --channel-list [string] Set the list of the channels to process. File needs to be in format channel-name[spaces]sampling_rate with a new channel in each line. Command given as --channel-list=location/to/file.  Duncan Meacher committed Sep 21, 2017 498 499 500 501 - --channel-name [string] Set the name of the channels to process. Can be given multiple times as --channel-name=IFO:AUX-CHANNEL-NAME:RATE  502 503 504 505 - --max-streams [int] Set the maximum number of streams to process at a given time (num_channels * num_rates = num_streams). Used to split up channels given into roughly equal subsets to be processed in sequence.  506 507 508 - --equal-subsets If set, forces an equal number of channels processed per channel subset.  509 510 511 512 513 514 515 516 - --max-sampling-rate [int] Maximum sampling rate for a given channel. If a given channel has a higher native sampling rate, it will be downsampled to this target rate. - --min-sampling-rate [int] Minimum sampling rate for a given channel when splitting a given channel into multiple frequency bands. If a channel has a lower sampling rate than this minimum, however, it will not be upsampled to this sampling rate.  Patrick Godwin committed Feb 14, 2017 517 518 519 520 521 522 523 524 525 526 527 - --framexmit-addr [string] Set the address of the framexmit service. Can be given multiple times as --framexmit-addr=IFO=xxx.xxx.xxx.xxx:port - --framexmit-iface [string] Set the address of the framexmit interface. - --shared-memory-partition [string] Set the name of the shared memory partition for a given instrument. Can be given multiple times as --shared-memory-partition=IFO=PARTITION-NAME  Patrick Godwin committed Jul 02, 2018 528 529 530 531 532 533 - --shared-memory-assumed-duration [int] Set the assumed span of files in seconds. Default = 4 seconds. - --shared-memory-block-size [int] Set the byte size to read per buffer. Default = 4096 bytes.  Patrick Godwin committed Feb 27, 2017 534 535 536 537 538 539 540 541 542 543 544 - --frame-type [string] Set the frame type required by the channels being used. - --frame-segments-file [filename] Set the name of the LIGO light-weight XML file from which to load frame segments. Optional iff --data-source is frames - --frame-segments-name [string] Set the name of the segments to extract from the segment tables. Required iff --frame-segments-file is given  545 - --section-include [string]  546  Set the channel sections to be included from the INI file. Can be given multiple times. Pass in spaces as underscores instead. If not specified, assumed to include all sections.  Patrick Godwin committed Feb 27, 2017 547   548 549 - --safety-include [string] Set the safety values for channels to be included from the INI file. Can be given multiple times. Default = "safe".  Patrick Godwin committed Feb 27, 2017 550 551  - --fidelity-exclude [string]  552  Set the fidelity values to be excluded from the INI file. Can be given multiple times.  Patrick Godwin committed Feb 27, 2017 553   554 - --safe-channel-include [string]  555  Set the channel names to be included from the INI file. Can be given multiple times. If not specified, assumed to include all channels.  Patrick Godwin committed Feb 27, 2017 556   557 558 559 - --unsafe-channel-include [string] Include this channel when reading the INI file, disregarding safety information (requires exact match). Can be repeated.  Patrick Godwin committed May 16, 2017 560 - --latency-output  Patrick Godwin committed Jul 02, 2018 561  Set whether to print out latency (in seconds) at various stages of the pipeline.  Patrick Godwin committed Feb 27, 2017 562   Patrick Godwin committed Feb 14, 2017 563 564  #### Typical usage case examples  Duncan Meacher committed Sep 21, 2017 565 566 567 568  -# Reading data from frames --data-source=frames --gps-start-time=999999000 --gps-end-time=999999999 --channel-name=H1:AUX-CHANNEL-NAME:RATE  Patrick Godwin committed Feb 14, 2017 569 570 571 572 573 574  -# Reading online data via framexmit --data-source=framexmit --channel-list=H1=location/to/file -# Many other combinations possible, please add some! """  Patrick Godwin committed Sep 10, 2018 575  group = optparse.OptionGroup(parser, "Data Source Options", "Use these options to set up the appropriate data source")  576  group.add_option("--data-source", metavar = "source", help = "Set the data source from [framexmit|lvshm|frames|silence|white|white_live]. Required.")  Patrick Godwin committed Feb 14, 2017 577  group.add_option("--block-size", type="int", metavar = "bytes", default = 16384 * 8 * 512, help = "Data block size to read in bytes. Default 16384 * 8 * 512 (512 seconds of double precision data at 16384 Hz. This parameter is only used if --data-source is one of white, silence.")  Patrick Godwin committed Mar 26, 2018 578 579  group.add_option("--gps-start-time", type="int", metavar = "seconds", help = "Set the start time of the segment to analyze in GPS seconds. Required unless --data-source=lvshm") group.add_option("--gps-end-time", type="int", metavar = "seconds", help = "Set the end time of the segment to analyze in GPS seconds. Required unless --data-source=lvshm")  Patrick Godwin committed Mar 16, 2017 580  group.add_option("--frame-cache", metavar = "filename", help = "Set the name of the LAL cache listing the LIGO-Virgo .gwf frame files (optional). This is required iff --data-source=frames")  581  group.add_option("--max-streams", type = "int", default = 50, help = "Maximum number of streams to process for a given pipeline at once. Used to split up channel lists into subsets that can then be processed in serial. Default = 50.")  582  group.add_option("--equal-subsets", action = "store_true", help = "If set, forces an equal number of channels processed per channel subset.")  Patrick Godwin committed Sep 10, 2018 583  group.add_option("--max-sample-rate", type = "int", default = 4096, help = "Maximum sampling rate for a given channel. If a given channel has a higher native sampling rate, it will be downsampled to this target rate. Default = 4096.")  584  group.add_option("--min-sample-rate", type = "int", default = 32, help = "Minimum sampling rate for a given channel when splitting a given channel into multiple frequency bands. If a channel has a lower sampling rate than this minimum, however, it will not be upsampled to this sampling rate. Default = 32.")  Patrick Godwin committed Feb 14, 2017 585 586 587  group.add_option("--framexmit-addr", metavar = "name", action = "append", help = "Set the address of the framexmit service. Can be given multiple times as --framexmit-addr=IFO=xxx.xxx.xxx.xxx:port") group.add_option("--framexmit-iface", metavar = "name", help = "Set the multicast interface address of the framexmit service.") group.add_option("--shared-memory-partition", metavar = "name", action = "append", help = "Set the name of the shared memory partition for a given instrument. Can be given multiple times as --shared-memory-partition=IFO=PARTITION-NAME")  Patrick Godwin committed Jul 02, 2018 588 589  group.add_option("--shared-memory-assumed-duration", type = "int", default = 4, help = "Set the assumed span of files in seconds. Default = 4.") group.add_option("--shared-memory-block-size", type = "int", default = 4096, help = "Set the byte size to read per buffer. Default = 4096.")  Patrick Godwin committed Feb 27, 2017 590 591 592  group.add_option("--frame-type", type="string", metavar = "name", help = "Include only those channels with the frame type given.") group.add_option("--frame-segments-file", metavar = "filename", help = "Set the name of the LIGO light-weight XML file from which to load frame segments. Optional iff --data-source=frames") group.add_option("--frame-segments-name", metavar = "name", help = "Set the name of the segments to extract from the segment tables. Required iff --frame-segments-file is given")  Patrick Godwin committed Sep 10, 2018 593 594 595 596 597 598  group.add_option("--latency-output", action = "store_true", help = "Print out latency output (s) at different stages of the pipeline (measured as current time - buffer time).") parser.add_option_group(group) group = optparse.OptionGroup(parser, "Channel Options", "Settings used for deciding which auxiliary channels to process.") group.add_option("--channel-list", type="string", metavar = "name", help = "Set the list of the channels to process. Command given as --channel-list=location/to/file") group.add_option("--channel-name", metavar = "name", action = "append", help = "Set the name of the channels to process. Can be given multiple times as --channel-name=IFO:AUX-CHANNEL-NAME:RATE")  599  group.add_option("--section-include", default=[], type="string", action="append", help="Set the channel sections to be included from the INI file. Can be given multiple times. Pass in spaces as underscores instead. If not specified, assumed to include all sections")  600 601 602 603  group.add_option("--safety-include", default=["safe"], type="string", action="append", help="Set the safety values for channels to be included from the INI file. Can be given multiple times. Default = 'safe'.") group.add_option("--fidelity-exclude", default=[], type="string", action="append", help="Set the fidelity values for channels to be excluded from the INI file. Can supply multiple values by repeating this argument. Each must be on of (add here)") group.add_option("--safe-channel-include", default=[], action="append", type="string", help="Include this channel when reading the INI file (requires exact match). Can be repeated. If not specified, assume to include all channels.") group.add_option("--unsafe-channel-include", default=[], action="append", type="string", help="Include this channel when reading the INI file, disregarding safety information (requires exact match). Can be repeated.")  Patrick Godwin committed Feb 14, 2017 604 605 606 607 608 609 610 611 612 613 614 615 616 617 618 619 620 621 622 623 624 625 626 627 628 629 630 631 632 633 634 635 636 637 638 639 640 641 642 643 644 645 646 647 648 649 650 651  parser.add_option_group(group) ## # _Gstreamer graph describing this function:_ # # @dot # digraph mkbasicsrc { # compound=true; # node [shape=record fontsize=10 fontname="Verdana"]; # subgraph clusterfakesrc { # fake_0 [label="fakesrc: white, silence", URL="\ref pipeparts.mkfakesrc()"]; # color=black; # label="Possible path #1"; # } # subgraph clusteronline { # color=black; # online_0 [label="lvshmsrc|framexmit", URL="\ref pipeparts.mklvshmsrc()"]; # online_1 [label ="framecppchanneldemux", URL="\ref pipeparts.mkframecppchanneldemux()"]; # online_2a [label ="channel 1 queue", URL="\ref pipeparts.mkqueue()"]; # online_2b [label ="channel 2 queue", URL="\ref pipeparts.mkqueue()"]; # online_2c [label ="channel 3 queue", URL="\ref pipeparts.mkqueue()"]; # online_3a [label ="audiorate 1", URL="\ref pipeparts.mkaudiorate()"]; # online_3b [label ="audiorate 2", URL="\ref pipeparts.mkaudiorate()"]; # online_3c [label ="audiorate 3", URL="\ref pipeparts.mkaudiorate()"]; # online_0 -> online_1; # online_1 -> online_2a; # online_1 -> online_2b; # online_1 -> online_2c; # online_2a -> online_3a; # online_2b -> online_3b; # online_2c -> online_3c; # label="Possible path #2"; # } # audioconv [label="audioconvert", URL="\ref pipeparts.mkaudioconvert()"]; # progress [label="progressreport (if verbose)", style=filled, color=lightgrey, URL="\ref pipeparts.mkprogressreport()"]; # sim [label="lalsimulation (if injections requested)", style=filled, color=lightgrey, URL="\ref pipeparts.mkinjections()"]; # queue [label="queue (if injections requested)", style=filled, color=lightgrey, URL="\ref pipeparts.mkqueue()"]; # # // The connections # fake_0 -> audioconv [ltail=clusterfakesrc]; # frames_4 -> audioconv [ltail=clusterframes]; # online_6 -> audioconv [ltail=clusteronline]; # nds_0 -> audioconv [ltail=clusternds]; # audioconv -> progress -> sim -> queue -> "?" # } # @enddot # #  652 def mkbasicmultisrc(pipeline, data_source_info, channels, verbose = False):  Patrick Godwin committed Feb 14, 2017 653 654 655 656 657 658 659  """! All the things for reading real or simulated channel data in one place. Consult the append_options() function and the DataSourceInfo class This src in general supports only one instrument although DataSourceInfo contains dictionaries of multi-instrument things. By  660 661 662  specifying the channels when calling this function you will only process the channels specified. A code wishing to have multiple basicmultisrcs will need to call this multiple times with different sets of channels specified.  Patrick Godwin committed Feb 14, 2017 663 664 665  """ if data_source_info.data_source == "white":  666  head = {channel : pipeparts.mkfakesrc(pipeline, instrument = data_source_info.instrument, channel_name = channel, volume = 1.0, rate = data_source_info.channel_dict[channel]['fsamp'], timestamp_offset = int(data_source_info.seg[0]) * Gst.SECOND) for channel in channels}  Patrick Godwin committed Feb 14, 2017 667  elif data_source_info.data_source == "silence":  668 669 670  head = {channel : pipeparts.mkfakesrc(pipeline, instrument = data_source_info.instrument, channel_name = channel, rate = data_source_info.channel_dict[channel]['fsamp'], timestamp_offset = int(data_source_info.seg[0]) * Gst.SECOND) for channel in channels} elif data_source_info.data_source == "white_live": head = {channel : pipeparts.mkfakesrc(pipeline, instrument = data_source_info.instrument, channel_name = channel, volume = 1.0, is_live = True, rate = data_source_info.channel_dict[channel]['fsamp'], timestamp_offset = int(data_source_info.seg[0]) * Gst.SECOND) for channel in channels}  Patrick Godwin committed Feb 27, 2017 671  elif data_source_info.data_source == "frames":  672  src = pipeparts.mklalcachesrc(pipeline, location = data_source_info.frame_cache, cache_src_regex = data_source_info.instrument[0], cache_dsc_regex = data_source_info.instrument, blocksize = 1048576)  673  demux = pipeparts.mkframecppchanneldemux(pipeline, src, do_file_checksum = False, skip_bad_files = True, channel_list = channels)  Patrick Godwin committed Feb 27, 2017 674 675  # allow frame reading and decoding to occur in a different # thread  676  head = dict.fromkeys(channels, None)  Patrick Godwin committed Mar 23, 2017 677  for channel in head:  Patrick Godwin committed Apr 10, 2018 678  head[channel] = pipeparts.mkreblock(pipeline, None, block_duration = 4 * Gst.SECOND)  Patrick Godwin committed Mar 29, 2017 679  pipeparts.src_deferred_link(demux, channel, head[channel].get_static_pad("sink"))  Patrick Godwin committed Apr 10, 2018 680  head[channel] = pipeparts.mkqueue(pipeline, head[channel], max_size_buffers = 0, max_size_bytes = 0, max_size_time = 8 * Gst.SECOND)  681  if data_source_info.frame_segments[data_source_info.instrument]:  Patrick Godwin committed Feb 27, 2017 682 683  # FIXME: make segmentsrc generate segment samples at the channel sample rate? # FIXME: make gate leaky when I'm certain that will work.  684 685  head[channel] = pipeparts.mkgate(pipeline, head[channel], threshold = 1, control = pipeparts.mksegmentsrc(pipeline, data_source_info.frame_segments[data_source_info.instrument]), name = "%s_frame_segments_gate" % channel) pipeparts.framecpp_channeldemux_check_segments.set_probe(head[channel].get_static_pad("src"), data_source_info.frame_segments[data_source_info.instrument])  Patrick Godwin committed Feb 27, 2017 686   Patrick Godwin committed Mar 23, 2017 687 688  # fill in holes, skip duplicate data head[channel] = pipeparts.mkaudiorate(pipeline, head[channel], skip_to_first = True, silent = False)  Patrick Godwin committed Feb 27, 2017 689   690  elif data_source_info.data_source in ("framexmit", "lvshm", "white_live"):  Patrick Godwin committed Feb 14, 2017 691 692  if data_source_info.data_source == "lvshm": # FIXME make wait_time adjustable through web interface or command line or both  Patrick Godwin committed Jul 02, 2018 693  src = pipeparts.mklvshmsrc(pipeline, shm_name = data_source_info.shm_part_dict[data_source_info.instrument], assumed_duration = data_source_info.shm_assumed_duration, blocksize = data_source_info.shm_block_size, wait_time = 120)  Patrick Godwin committed Feb 14, 2017 694  elif data_source_info.data_source == "framexmit":  695  src = pipeparts.mkframexmitsrc(pipeline, multicast_iface = data_source_info.framexmit_iface, multicast_group = data_source_info.framexmit_addr[data_source_info.instrument][0], port = data_source_info.framexmit_addr[data_source_info.instrument][1], wait_time = 120)  Patrick Godwin committed Feb 14, 2017 696 697 698 699  else: # impossible code path raise ValueError(data_source_info.data_source)  700  demux = pipeparts.mkframecppchanneldemux(pipeline, src, do_file_checksum = False, skip_bad_files = True, channel_list = channels)  Patrick Godwin committed Feb 14, 2017 701 702  # channels  703  head = dict.fromkeys(channels, None)  Patrick Godwin committed Feb 14, 2017 704  for channel in head:  705  head[channel] = pipeparts.mkqueue(pipeline, None, max_size_buffers = 0, max_size_bytes = 0, max_size_time = Gst.SECOND* 60 * 1) # 1 minute of buffering  Patrick Godwin committed Mar 29, 2017 706  pipeparts.src_deferred_link(demux, channel, head[channel].get_static_pad("sink"))  Patrick Godwin committed May 16, 2017 707 708 709  if data_source_info.latency_output: head[channel] = pipeparts.mklatency(pipeline, head[channel], name = 'stage1_afterFrameXmit_%s' % channel)  Patrick Godwin committed Feb 14, 2017 710 711 712 713 714 715 716 717 718  # fill in holes, skip duplicate data head[channel] = pipeparts.mkaudiorate(pipeline, head[channel], skip_to_first = True, silent = False) # 10 minutes of buffering head[channel] = pipeparts.mkqueue(pipeline, head[channel], max_size_buffers = 0, max_size_bytes = 0, max_size_time = Gst.SECOND * 60 * 10) else: raise ValueError("invalid data_source: %s" % data_source_info.data_source)  719 720 721 722  for channel in head: head[channel] = pipeparts.mkaudioconvert(pipeline, head[channel]) # progress report if verbose:  Patrick Godwin committed Mar 19, 2018 723  head[channel] = pipeparts.mkprogressreport(pipeline, head[channel], "datasource_progress_for_%s" % channel)  724   Patrick Godwin committed Feb 14, 2017 725 726 727 728 729 730  return head # Unit tests if __name__ == "__main__": import doctest doctest.testmod()