lalinference_pipe_utils.py 112 KB
Newer Older
1

2
#flow DAG Class definitions for LALInference Pipeline
3
# (C) 2012 John Veitch, Vivien Raymond, Kiersten Ruisard, Kan Wang
4

John Douglas Veitch's avatar
John Douglas Veitch committed
5
import itertools
6
import glue
7
from glue import pipeline,segmentsUtils,segments
8
import os
9
import socket
10
from lalapps import inspiralutils
11 12
import uuid
import ast
13
import pdb
John Douglas Veitch's avatar
John Douglas Veitch committed
14
import string
15
from math import floor,ceil,log,pow
16
import sys
17
import random
18
from itertools import permutations
19
import shutil
20
import numpy as np
21 22 23 24 25

# We use the GLUE pipeline utilities to construct classes for each
# type of job. Each class has inputs and outputs, which are used to
# join together types of jobs into a DAG.

John Douglas Veitch's avatar
John Douglas Veitch committed
26 27 28 29
class Event():
  """
  Represents a unique event to run on
  """
John Douglas Veitch's avatar
John Douglas Veitch committed
30
  new_id=itertools.count().next
31
  def __init__(self,trig_time=None,SimInspiral=None,SimBurst=None,SnglInspiral=None,CoincInspiral=None,event_id=None,timeslide_dict=None,GID=None,ifos=None, duration=None,srate=None,trigSNR=None,fhigh=None,horizon_distance=None):
John Douglas Veitch's avatar
John Douglas Veitch committed
32
    self.trig_time=trig_time
John Douglas Veitch's avatar
John Douglas Veitch committed
33
    self.injection=SimInspiral
34
    self.burstinjection=SimBurst
John Douglas Veitch's avatar
John Douglas Veitch committed
35
    self.sngltrigger=SnglInspiral
36 37 38 39
    if timeslide_dict is None:
      self.timeslides={}
    else:
      self.timeslides=timeslide_dict
40 41
    self.GID=GID
    self.coinctrigger=CoincInspiral
42 43 44 45
    if ifos is None:
      self.ifos = []
    else:
      self.ifos = ifos
46 47
    self.duration = duration
    self.srate = srate
48
    self.trigSNR = trigSNR
49
    self.fhigh = fhigh
50
    self.horizon_distance = horizon_distance
John Douglas Veitch's avatar
John Douglas Veitch committed
51 52 53
    if event_id is not None:
        self.event_id=event_id
    else:
John Douglas Veitch's avatar
John Douglas Veitch committed
54 55 56
        self.event_id=Event.new_id()
    if self.injection is not None:
        self.trig_time=self.injection.get_end()
57
        if event_id is None: self.event_id=int(str(self.injection.simulation_id).split(':')[2])
58 59 60
    if self.burstinjection is not None:
        self.trig_time=self.burstinjection.get_end()
        if event_id is None: self.event_id=int(str(self.burstinjection.simulation_id).split(':')[2])
John Douglas Veitch's avatar
John Douglas Veitch committed
61
    if self.sngltrigger is not None:
62
        self.trig_time=self.sngltrigger.get_end()
John Douglas Veitch's avatar
John Douglas Veitch committed
63
        self.event_id=int(str(self.sngltrigger.event_id).split(':')[2])
64 65 66 67
    if self.coinctrigger is not None:
        self.trig_time=self.coinctrigger.end_time + 1.0e-9 * self.coinctrigger.end_time_ns
    if self.GID is not None:
        self.event_id=int(''.join(i for i in self.GID if i.isdigit()))
68 69 70 71 72 73 74
    self.engine_opts={}
  def set_engine_option(self,opt,val):
    """
    Can set event-specific options for the engine nodes
    using this option, e.g. ev.set_engine_option('time-min','1083759273')
    """
    self.engine_opts[opt]=val
John Douglas Veitch's avatar
John Douglas Veitch committed
75 76

dummyCacheNames=['LALLIGO','LALVirgo','LALAdLIGO','LALAdVirgo']
John Douglas Veitch's avatar
John Douglas Veitch committed
77

78
def readLValert(threshold_snr=None,gid=None,flow=40.0,gracedb="gracedb",basepath="./",downloadpsd=True,roq=False):
79
  """
80
  Parse LV alert file, containing coinc, sngl, coinc_event_map.
81 82
  and create a list of Events as input for pipeline
  Based on Chris Pankow's script
83
  """
84
  output=[]
85 86
  from glue.ligolw import utils
  from glue.ligolw import lsctables
87
  from glue.ligolw import ligolw
Rory Smith's avatar
Rory Smith committed
88
  class PSDContentHandler(ligolw.LIGOLWContentHandler):
89
    pass
Rory Smith's avatar
Rory Smith committed
90
  lsctables.use_in(PSDContentHandler)
91 92
  from glue.ligolw import param
  from glue.ligolw import array
93
  import subprocess
Rory Smith's avatar
Rory Smith committed
94
  from lal import series as lalseries
95
  from subprocess import Popen, PIPE
96 97
  cwd=os.getcwd()
  os.chdir(basepath)
98
  print "%s download %s coinc.xml"%(gracedb,gid)
Vivien Raymond's avatar
Vivien Raymond committed
99
  subprocess.call([gracedb,"download", gid ,"coinc.xml"])
Rory Smith's avatar
Rory Smith committed
100
  xmldoc=utils.load_filename("coinc.xml",contenthandler = PSDContentHandler)
101
  coinctable = lsctables.CoincInspiralTable.get_table(xmldoc)
102
  coinc_events = [event for event in coinctable]
103
  sngltable = lsctables.SnglInspiralTable.get_table(xmldoc)
104
  sngl_events = [event for event in sngltable]
105 106
  ifos = coinctable[0].ifos.split(",")
  trigSNR = coinctable[0].snr
107
  # Parse PSD
108
  srate_psdfile=16384
109
  fhigh=None
110 111 112
  if downloadpsd:
    print "gracedb download %s psd.xml.gz" % gid
    subprocess.call([gracedb,"download", gid ,"psd.xml.gz"])
Rory Smith's avatar
Rory Smith committed
113
    xmlpsd = lalseries.read_psd_xmldoc(utils.load_filename('psd.xml.gz',contenthandler = lalseries.PSDContentHandler))
114 115 116 117 118 119
    if os.path.exists("psd.xml.gz"):
      psdasciidic=get_xml_psds(os.path.realpath("./psd.xml.gz"),ifos,os.path.realpath('./PSDs'),end_time=None)
      combine=np.loadtxt(psdasciidic[psdasciidic.keys()[0]])
      srate_psdfile = pow(2.0, ceil( log(float(combine[-1][0]), 2) ) ) * 2
    else:
      print "Failed to gracedb download %s psd.xml.gz. lalinference will estimate the psd itself." % gid
120
  coinc_map = lsctables.CoincMapTable.get_table(xmldoc)
121 122
  for coinc in coinc_events:
    these_sngls = [e for e in sngl_events if e.event_id in [c.event_id for c in coinc_map if c.coinc_event_id == coinc.coinc_event_id] ]
123 124
    dur=[]
    srate=[]
125
    horizon_distance=[]
126
    for e in these_sngls:
127
      # Review: Replace this with a call to LALSimulation function at some point
128 129 130 131 132
      if roq==False:
        p=Popen(["lalapps_chirplen","--flow",str(flow),"-m1",str(e.mass1),"-m2",str(e.mass2)],stdout=PIPE, stderr=PIPE, stdin=PIPE)
        strlen = p.stdout.read()
        dur.append(pow(2.0, ceil( log(max(8.0,float(strlen.splitlines()[2].split()[5]) + 2.0), 2) ) ) )
        srate.append(pow(2.0, ceil( log(float(strlen.splitlines()[1].split()[5]), 2) ) ) * 2 )
133 134 135 136 137 138 139
      snr = e.snr
      eff_dist = e.eff_distance
      if threshold_snr is not None:
          if snr > threshold_snr:
              horizon_distance.append(eff_dist * snr/threshold_snr)
          else:
              horizon_distance.append(2 * eff_dist)
140 141 142 143 144 145 146 147 148 149 150
    if srate:
      if max(srate)<srate_psdfile:
        srate = max(srate)
      else:
        srate = srate_psdfile
        if downloadpsd:
          fhigh = srate_psdfile/2.0 * 0.95 # Because of the drop-off near Nyquist of the PSD from gstlal
    else:
      srate = None
    if dur:
      duration = max(dur)
151
    else:
152
      duration = None
153
    horizon_distance = max(horizon_distance) if len(horizon_distance) > 0 else None
154
    ev=Event(CoincInspiral=coinc, GID=gid, ifos = ifos, duration = duration, srate = srate,
155 156
             trigSNR = trigSNR, fhigh = fhigh, horizon_distance=horizon_distance)
    output.append(ev)
157

158
  print "Found %d coinc events in table." % len(coinc_events)
159
  os.chdir(cwd)
160 161
  return output

162 163 164 165 166 167 168
def open_pipedown_database(database_filename,tmp_space):
    """
    Open the connection to the pipedown database
    """
    if not os.access(database_filename,os.R_OK):
	raise Exception('Unable to open input file: %s'%(database_filename))
    from glue.ligolw import dbtables
169
    import sqlite3
170 171 172 173
    working_filename=dbtables.get_connection_filename(database_filename,tmp_path=tmp_space)
    connection = sqlite3.connect(working_filename)
    if tmp_space:
	dbtables.set_temp_store_directory(connection,tmp_space)
174
    #dbtables.DBTable_set_connection(connection)
175
    return (connection,working_filename)
176

177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219
def get_zerolag_lloid(database_connection, dumpfile=None, gpsstart=None, gpsend=None, max_cfar=-1, min_cfar=-1):
	"""
	Returns a list of Event objects
	from pipedown data base. Can dump some stats to dumpfile if given,
	and filter by gpsstart and gpsend to reduce the nunmber or specify
	max_cfar to select by combined FAR
	"""
	output={}
	if gpsstart is not None: gpsstart=float(gpsstart)
	if gpsend is not None: gpsend=float(gpsend)
	# Get coincs
	get_coincs = "SELECT sngl_inspiral.end_time+sngl_inspiral.end_time_ns*1e-9,sngl_inspiral.ifo,coinc_event.coinc_event_id,sngl_inspiral.snr,sngl_inspiral.chisq,coinc_inspiral.combined_far \
		FROM sngl_inspiral join coinc_event_map on (coinc_event_map.table_name=='sngl_inspiral' and coinc_event_map.event_id ==\
		sngl_inspiral.event_id) join coinc_event on (coinc_event.coinc_event_id==coinc_event_map.coinc_event_id) \
		join coinc_inspiral on (coinc_event.coinc_event_id==coinc_inspiral.coinc_event_id) \
        WHERE coinc_event.time_slide_id=='time_slide:time_slide_id:1'\
		"
	if gpsstart is not None:
		get_coincs=get_coincs+' and coinc_inspiral.end_time+coinc_inspiral.end_time_ns*1.0e-9 > %f'%(gpsstart)
	if gpsend is not None:
		get_coincs=get_coincs+' and coinc_inspiral.end_time+coinc_inspiral.end_time_ns*1.0e-9 < %f'%(gpsend)
	if max_cfar !=-1:
		get_coincs=get_coincs+' and coinc_inspiral.combined_far < %f'%(max_cfar)
	if min_cfar != -1:
		get_coincs=get_coincs+' and coinc_inspiral.combined_far > %f'%(min_cfar)
	db_out=database_connection.cursor().execute(get_coincs)
    	extra={}
	for (sngl_time, ifo, coinc_id, snr, chisq, cfar) in db_out:
      		coinc_id=int(coinc_id.split(":")[-1])
	  	if not coinc_id in output.keys():
			output[coinc_id]=Event(trig_time=sngl_time,timeslide_dict={},event_id=int(coinc_id))
			extra[coinc_id]={}
		output[coinc_id].timeslides[ifo]=0
		output[coinc_id].ifos.append(ifo)
		extra[coinc_id][ifo]={'snr':snr,'chisq':chisq,'cfar':cfar}
	if dumpfile is not None:
		fh=open(dumpfile,'w')
		for co in output.keys():
			for ifo in output[co].ifos:
				fh.write('%s %s %s %s %s %s %s\n'%(str(co),ifo,str(output[co].trig_time),str(output[co].timeslides[ifo]),str(extra[co][ifo]['snr']),str(extra[co][ifo]['chisq']),str(extra[co][ifo]['cfar'])))
		fh.close()
	return output.values()

220
def get_zerolag_pipedown(database_connection, dumpfile=None, gpsstart=None, gpsend=None, max_cfar=-1, min_cfar=-1):
221 222 223 224 225 226 227 228 229 230 231 232 233 234
	"""
	Returns a list of Event objects
	from pipedown data base. Can dump some stats to dumpfile if given,
	and filter by gpsstart and gpsend to reduce the nunmber or specify
	max_cfar to select by combined FAR
	"""
	output={}
	if gpsstart is not None: gpsstart=float(gpsstart)
	if gpsend is not None: gpsend=float(gpsend)
	# Get coincs
	get_coincs = "SELECT sngl_inspiral.end_time+sngl_inspiral.end_time_ns*1e-9,sngl_inspiral.ifo,coinc_event.coinc_event_id,sngl_inspiral.snr,sngl_inspiral.chisq,coinc_inspiral.combined_far \
		FROM sngl_inspiral join coinc_event_map on (coinc_event_map.table_name=='sngl_inspiral' and coinc_event_map.event_id ==\
		sngl_inspiral.event_id) join coinc_event on (coinc_event.coinc_event_id==coinc_event_map.coinc_event_id) \
		join coinc_inspiral on (coinc_event.coinc_event_id==coinc_inspiral.coinc_event_id) \
235
		WHERE coinc_event.time_slide_id=='time_slide:time_slide_id:10049'\
236 237
		"
	if gpsstart is not None:
238
		get_coincs=get_coincs+' and coinc_inspiral.end_time+coinc_inspiral.end_time_ns*1.0e-9 > %f'%(gpsstart)
239
	if gpsend is not None:
240
		get_coincs=get_coincs+' and coinc_inspiral.end_time+coinc_inspiral.end_time_ns*1.0e-9 < %f'%(gpsend)
241 242
	if max_cfar !=-1:
		get_coincs=get_coincs+' and coinc_inspiral.combined_far < %f'%(max_cfar)
243 244
	if min_cfar != -1:
		get_coincs=get_coincs+' and coinc_inspiral.combined_far > %f'%(min_cfar)
245
	db_out=database_connection.cursor().execute(get_coincs)
John Douglas Veitch's avatar
John Douglas Veitch committed
246
    	extra={}
247
	for (sngl_time, ifo, coinc_id, snr, chisq, cfar) in db_out:
John Douglas Veitch's avatar
John Douglas Veitch committed
248 249
      		coinc_id=int(coinc_id.split(":")[-1])
	  	if not coinc_id in output.keys():
250
			output[coinc_id]=Event(trig_time=sngl_time,timeslide_dict={},event_id=int(coinc_id))
John Douglas Veitch's avatar
John Douglas Veitch committed
251 252 253 254
			extra[coinc_id]={}
		output[coinc_id].timeslides[ifo]=0
		output[coinc_id].ifos.append(ifo)
		extra[coinc_id][ifo]={'snr':snr,'chisq':chisq,'cfar':cfar}
255 256 257 258 259 260
	if dumpfile is not None:
		fh=open(dumpfile,'w')
		for co in output.keys():
			for ifo in output[co].ifos:
				fh.write('%s %s %s %s %s %s %s\n'%(str(co),ifo,str(output[co].trig_time),str(output[co].timeslides[ifo]),str(extra[co][ifo]['snr']),str(extra[co][ifo]['chisq']),str(extra[co][ifo]['cfar'])))
		fh.close()
261
	return output.values()
262

263
def get_timeslides_pipedown(database_connection, dumpfile=None, gpsstart=None, gpsend=None, max_cfar=-1):
264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279
	"""
	Returns a list of Event objects
	with times and timeslide offsets
	"""
	output={}
	if gpsstart is not None: gpsstart=float(gpsstart)
	if gpsend is not None: gpsend=float(gpsend)
	db_segments=[]
	sql_seg_query="SELECT search_summary.out_start_time, search_summary.out_end_time from search_summary join process on process.process_id==search_summary.process_id where process.program=='thinca'"
	db_out = database_connection.cursor().execute(sql_seg_query)
	for d in db_out:
		if d not in db_segments:
			db_segments.append(d)
	seglist=segments.segmentlist([segments.segment(d[0],d[1]) for d in db_segments])
	db_out_saved=[]
	# Get coincidences
280
	get_coincs="SELECT sngl_inspiral.end_time+sngl_inspiral.end_time_ns*1e-9,time_slide.offset,sngl_inspiral.ifo,coinc_event.coinc_event_id,sngl_inspiral.snr,sngl_inspiral.chisq,coinc_inspiral.combined_far \
281
		    FROM sngl_inspiral join coinc_event_map on (coinc_event_map.table_name == 'sngl_inspiral' and coinc_event_map.event_id \
282 283
		    == sngl_inspiral.event_id) join coinc_event on (coinc_event.coinc_event_id==coinc_event_map.coinc_event_id) join time_slide\
		    on (time_slide.time_slide_id == coinc_event.time_slide_id and time_slide.instrument==sngl_inspiral.ifo)\
284
		    join coinc_inspiral on (coinc_inspiral.coinc_event_id==coinc_event.coinc_event_id) where coinc_event.time_slide_id!='time_slide:time_slide_id:10049'"
John Douglas Veitch's avatar
John Douglas Veitch committed
285
	joinstr = ' and '
286
	if gpsstart is not None:
287
		get_coincs=get_coincs+ joinstr + ' coinc_inspiral.end_time+coinc_inspiral.end_time_ns*1e-9 > %f'%(gpsstart)
288
	if gpsend is not None:
289
		get_coincs=get_coincs+ joinstr+' coinc_inspiral.end_time+coinc_inspiral.end_time_ns*1e-9 <%f'%(gpsend)
290 291
	if max_cfar!=-1:
		get_coincs=get_coincs+joinstr+' coinc_inspiral.combined_far < %f'%(max_cfar)
292
	db_out=database_connection.cursor().execute(get_coincs)
293 294
	from pylal import SnglInspiralUtils
	extra={}
295
	for (sngl_time, slide, ifo, coinc_id, snr, chisq, cfar) in db_out:
296 297 298 299 300 301 302 303 304
		coinc_id=int(coinc_id.split(":")[-1])
		seg=filter(lambda seg:sngl_time in seg,seglist)[0]
		slid_time = SnglInspiralUtils.slideTimeOnRing(sngl_time,slide,seg)
		if not coinc_id in output.keys():
			output[coinc_id]=Event(trig_time=slid_time,timeslide_dict={},event_id=int(coinc_id))
			extra[coinc_id]={}
		output[coinc_id].timeslides[ifo]=slid_time-sngl_time
		output[coinc_id].ifos.append(ifo)
		extra[coinc_id][ifo]={'snr':snr,'chisq':chisq,'cfar':cfar}
305 306 307 308 309 310
	if dumpfile is not None:
		fh=open(dumpfile,'w')
		for co in output.keys():
			for ifo in output[co].ifos:
				fh.write('%s %s %s %s %s %s %s\n'%(str(co),ifo,str(output[co].trig_time),str(output[co].timeslides[ifo]),str(extra[co][ifo]['snr']),str(extra[co][ifo]['chisq']),str(extra[co][ifo]['cfar'])))
		fh.close()
311
	return output.values()
312

313 314 315 316 317 318 319 320
def mkdirs(path):
  """
  Helper function. Make the given directory, creating intermediate
  dirs if necessary, and don't complain about it already existing.
  """
  if os.access(path,os.W_OK) and os.path.isdir(path): return
  else: os.makedirs(path)

John Douglas Veitch's avatar
John Douglas Veitch committed
321 322 323
def chooseEngineNode(name):
  if name=='lalinferencenest':
    return LALInferenceNestNode
324 325
  if name=='lalinferenceburst':
    return LALInferenceBurstNode
John Douglas Veitch's avatar
John Douglas Veitch committed
326 327
  if name=='lalinferencemcmc':
    return LALInferenceMCMCNode
328
  if name=='lalinferencebambi' or name=='lalinferencebambimpi':
329
    return LALInferenceBAMBINode
330 331
  if name=='lalinferencedatadump':
    return LALInferenceDataDumpNode
332 333
  if name=='bayeswavepsd':
    return BayesWavePSDNode
John Douglas Veitch's avatar
John Douglas Veitch committed
334 335
  return EngineNode

336 337 338
def get_engine_name(cp):
    name=cp.get('analysis','engine')
    if name=='random':
339
        engine_list=['lalinferencenest','lalinferencemcmc','lalinferencebambimpi']
340 341 342 343 344 345 346 347 348
        if cp.has_option('input','gid'):
            gid=cp.get('input','gid')
            engine_number=int(''.join(i for i in gid if i.isdigit())) % 2
        else:
            engine_number=random.randint(0,1)
        return engine_list[engine_number]
    else:
        return name

349
def scan_timefile(timefile):
350 351 352 353 354 355 356 357 358 359 360 361
    import re
    p=re.compile('[\d.]+')
    times=[]
    timefilehandle=open(timefile,'r')
    for time in timefilehandle:
      if not p.match(time):
	continue
      if float(time) in times:
	print 'Skipping duplicate time %s'%(time)
	continue
      print 'Read time %s'%(time)
      times.append(float(time))
362 363
    timefilehandle.close()
    return times
364

365 366 367 368 369 370 371 372 373 374 375 376 377 378 379
def get_xml_psds(psdxml,ifos,outpath,end_time=None):
  """
  Get a psd.xml.gz file and:
  1) Reads it
  2) Converts PSD (10e-44) -> ASD ( ~10e-22)
  3) Checks the psd file contains all the IFO we want to analyze
  4) Writes down the ASDs into an ascii file for each IFO in psd.xml.gz. The name of the file contains the trigtime (if given) and the ifo name.
  Input:
    psdxml: psd.xml.gz file
    ifos: list of ifos used for the analysis
    outpath: path where the ascii ASD will be written to
    (end_time): trigtime for this event. Will be used a part of the ASD file name
  """
  lal=1
  from glue.ligolw import utils
380
  try:
Rory Smith's avatar
Rory Smith committed
381 382
    #from pylal import series
    from lal import series as series
383
    lal=0
384 385 386
  except ImportError:
    print "ERROR, cannot import pylal.series in bppu/get_xml_psds()\n"
    exit(1)
387

388 389 390 391 392 393 394 395 396 397 398
  out={}
  if not os.path.isdir(outpath):
    os.makedirs(outpath)
  if end_time is not None:
    time=repr(float(end_time))
  else:
    time=''
  #check we don't already have ALL the psd files #
  got_all=1
  for ifo in ifos:
    path_to_ascii_psd=os.path.join(outpath,ifo+'_psd_'+time+'.txt')
399
    # Check we don't already have that ascii (e.g. because we are running parallel runs of the save event
400 401 402 403 404 405 406 407 408
    if os.path.isfile(path_to_ascii_psd):
      got_all*=1
    else:
      got_all*=0
  if got_all==1:
    #print "Already have PSD files. Nothing to do...\n"
    for ifo in ifos:
      out[ifo]=os.path.join(outpath,ifo+'_psd_'+time+'.txt')
    return out
409

410 411 412 413
  # We need to convert the PSD for one or more IFOS. Open the file
  if not os.path.isfile(psdxml):
    print "ERROR: impossible to open the psd file %s. Exiting...\n"%psdxml
    sys.exit(1)
Rory Smith's avatar
Rory Smith committed
414
  xmlpsd =  series.read_psd_xmldoc(utils.load_filename(psdxml,contenthandler = series.PSDContentHandler))
415 416 417 418 419 420 421 422 423
  # Check the psd file contains all the IFOs we want to analize
  for ifo in ifos:
    if not ifo in [i.encode('ascii') for i in xmlpsd.keys()]:
      print "ERROR. The PSD for the ifo %s does not seem to be contained in %s\n"%(ifo,psdxml)
      sys.exit(1)
  #loop over ifos in psd xml file
  for instrument in xmlpsd.keys():
    #name of the ascii file we are going to write the PSD into
    path_to_ascii_psd=os.path.join(outpath,instrument.encode('ascii')+'_psd_'+time+'.txt')
424
    # Check we don't already have that ascii (e.g. because we are running parallel runs of the save event
425
    if os.path.isfile(path_to_ascii_psd):
426
      continue
427 428 429 430 431 432 433 434 435 436 437 438 439 440 441
    # get data for the IFO
    ifodata=xmlpsd[instrument]
    #check data is not empty
    if ifodata is None:
      continue
    # we have data. Get psd array
    if lal==0:
      #pylal stores the series in ifodata.data
      data=ifodata
    else:
      # lal stores it in ifodata.data.data
      data=ifodata.data
    # Fill a two columns array of (freq, psd) and save it in the ascii file
    f0=ifodata.f0
    deltaF=ifodata.deltaF
442

443
    combine=[]
444

Rory Smith's avatar
Rory Smith committed
445 446
    for i in np.arange(len(data.data.data)) :
      combine.append([f0+i*deltaF,np.sqrt(data.data.data[i])])
447
    np.savetxt(path_to_ascii_psd,combine)
448
    ifo=instrument.encode('ascii')
449 450 451
    # set node.psds dictionary with the path to the ascii files
    out[ifo]=os.path.join(outpath,ifo+'_psd_'+time+'.txt')
  return out
452

453 454 455 456 457 458 459 460 461 462 463 464 465 466 467 468 469 470 471 472 473 474 475 476
def get_trigger_chirpmass(gid=None,gracedb="gracedb"):
  from glue.ligolw import lsctables
  from glue.ligolw import ligolw
  from glue.ligolw import utils
  class PSDContentHandler(ligolw.LIGOLWContentHandler):
    pass
  lsctables.use_in(PSDContentHandler)
  import subprocess

  cwd=os.getcwd()
  subprocess.call([gracedb,"download", gid ,"coinc.xml"])
  xmldoc=utils.load_filename("coinc.xml",contenthandler = PSDContentHandler)
  coinctable = lsctables.CoincInspiralTable.get_table(xmldoc)
  coinc_events = [event for event in coinctable]
  sngltable = lsctables.SnglInspiralTable.get_table(xmldoc)
  sngl_events = [event for event in sngltable]
  coinc_map = lsctables.CoincMapTable.get_table(xmldoc)
  mass1 = []
  mass2 = []
  for coinc in coinc_events:
    these_sngls = [e for e in sngl_events if e.event_id in [c.event_id for c in coinc_map if c.coinc_event_id == coinc.coinc_event_id] ]
    for e in these_sngls:
      mass1.append(e.mass1)
      mass2.append(e.mass2)
477
  # check that trigger masses are identical in each IFO
478 479 480
  assert len(set(mass1)) == 1
  assert len(set(mass2)) == 1

481
  mchirp = (mass1[0]*mass2[0])**(3./5.) / ( (mass1[0] + mass2[0])**(1./5.) )
482 483 484 485
  os.remove("coinc.xml")

  return mchirp

486
def get_roq_mchirp_priors(path, roq_paths, roq_params, key, gid=None,sim_inspiral=None):
487

488 489
  ## XML and GID cannot be given at the same time
  ## sim_inspiral must already point at the right row
490
  mc_priors = {}
Vivien Raymond's avatar
Vivien Raymond committed
491

492 493 494
  if gid is not None and sim_inspiral is not None:
    print "Error in get_roq_mchirp_priors, cannot use both gid and sim_inspiral\n"
    sys.exit(1)
495 496 497 498 499 500

  for roq in roq_paths:
    params=os.path.join(path,roq,'params.dat')
    roq_params[roq]=np.genfromtxt(params,names=True)
    mc_priors[roq]=[float(roq_params[roq]['chirpmassmin']),float(roq_params[roq]['chirpmassmax'])]
  ordered_roq_paths=[item[0] for item in sorted(roq_params.items(), key=key)][::-1]
501 502
  # below is to construct non-overlapping mc priors for multiple roq mass-bin runs
  '''i=0
503 504 505 506 507 508 509
  for roq in ordered_roq_paths:
    if i>0:
      # change min, just set to the max of the previous one since we have already aligned it in the previous iteration of this loop
      #mc_priors[roq][0]+= (mc_priors[roq_lengths[i-1]][1]-mc_priors[roq][0])/2.
      mc_priors[roq][0]=mc_priors[ordered_roq_paths[i-1]][1]
    if i<len(roq_paths)-1:
      mc_priors[roq][1]-= (mc_priors[roq][1]- mc_priors[ordered_roq_paths[i+1]][0])/2.
510
    i+=1'''
511 512
  if gid is not None:
  	trigger_mchirp = get_trigger_chirpmass(gid)
513 514
  elif sim_inspiral is not None:
        trigger_mchirp = sim_inspiral.mchirp
515 516
  else:
	trigger_mchirp = None
517 518 519

  return mc_priors, trigger_mchirp

Vivien Raymond's avatar
Vivien Raymond committed
520 521 522 523 524 525 526 527 528 529 530 531 532 533 534 535 536 537 538 539 540 541 542 543 544 545
def get_roq_component_mass_priors(path, roq_paths, roq_params, key, gid=None,sim_inspiral=None):

  ## XML and GID cannot be given at the same time
  ## sim_inspiral must already point at the right row
  m1_priors = {}
  m2_priors = {}

  if gid is not None and sim_inspiral is not None:
    print "Error in get_roq_mchirp_priors, cannot use both gid and sim_inspiral\n"
    sys.exit(1)

  for roq in roq_paths:
    params=os.path.join(path,roq,'params.dat')
    roq_params[roq]=np.genfromtxt(params,names=True)
    m1_priors[roq]=[float(roq_params[roq]['mass1min']),float(roq_params[roq]['mass1max'])]
    m2_priors[roq]=[float(roq_params[roq]['mass2min']),float(roq_params[roq]['mass2max'])]

  if gid is not None:
    trigger_mchirp = get_trigger_chirpmass(gid)
  elif sim_inspiral is not None:
    trigger_mchirp = sim_inspiral.mchirp
  else:
    trigger_mchirp = None

  return m1_priors, m2_priors, trigger_mchirp

546
def get_roq_mass_freq_scale_factor(mc_priors, trigger_mchirp, force_flow=None):
547 548 549 550 551
  mc_priors_keys_int = map(lambda k : int(k[:-1]), mc_priors.keys())
  roq_min = mc_priors.keys()[np.argmin(mc_priors_keys_int)]
  roq_max = mc_priors.keys()[np.argmax(mc_priors_keys_int)]
  mc_max = mc_priors[roq_min][1]
  mc_min = mc_priors[roq_max][0]
552
  scale_factor = 1.
553
  if force_flow == None and trigger_mchirp != None:
554 555 556
     if trigger_mchirp >= mc_max:
        scale_factor = 2.**(floor(trigger_mchirp/mc_max))
     if trigger_mchirp <= mc_min:
557
        scale_factor = (2./3.2)**(ceil(trigger_mchirp/mc_min))
558
  elif force_flow != None:
559
     scale_factor = 20./force_flow
560
  return scale_factor
561

562 563
def create_pfn_tuple(filename,protocol='file://',site='local'):
    return( (os.path.basename(filename),protocol+os.path.abspath(filename),site) )
564

Vivien Raymond's avatar
Vivien Raymond committed
565 566 567 568 569 570 571 572 573 574 575 576 577 578 579 580 581 582 583 584 585
def mchirp_from_components(m1, m2):
  return (m1*m2)**(3.0/5.0) / (m1+m2)**(1.0/5.0)

def Query_ROQ_Bounds_Type(path, roq_paths):
  # Assume that parametrization of ROQ bounds is independent of seglen; just look at first one
  import numpy as np
  roq = roq_paths[0]
  params = os.path.join(path,roq,'params.dat')
  roq_params0 = np.genfromtxt(params,names=True)
  roq_names_set = set(roq_params0.dtype.names)
  component_mass_bounds_set = set(['mass1min', 'mass1max', 'mass2min', 'mass2max'])
  chirp_mass_q_bounds_set = set(['chirpmassmin', 'chirpmassmax', 'qmin', 'qmax'])
  if roq_names_set.issuperset(component_mass_bounds_set):
    roq_bounds = 'component_mass'
  elif roq_names_set.issuperset(chirp_mass_q_bounds_set):
    roq_bounds = 'chirp_mass_q'
  else:
    print 'Invalid bounds for ROQ. Ether (m1,m2) or (mc,q) bounds are supported.'
    sys.exit(1)
  return roq_bounds

586
class LALInferencePipelineDAG(pipeline.CondorDAG):
587
  def __init__(self,cp,dax=False,first_dag=True,previous_dag=None,site='local'):
588 589
    self.subfiles=[]
    self.config=cp
590
    self.engine=get_engine_name(cp)
John Douglas Veitch's avatar
John Douglas Veitch committed
591
    self.EngineNode=chooseEngineNode(self.engine)
592
    self.site=site
593
    if cp.has_option('paths','basedir'):
John Douglas Veitch's avatar
John Douglas Veitch committed
594
      self.basepath=cp.get('paths','basedir')
595 596 597
    else:
      self.basepath=os.getcwd()
      print 'No basepath specified, using current directory: %s'%(self.basepath)
598
    mkdirs(self.basepath)
599 600
    if dax:
        os.chdir(self.basepath)
601
    self.posteriorpath=os.path.join(self.basepath,'posterior_samples')
John Douglas Veitch's avatar
John Douglas Veitch committed
602
    mkdirs(self.posteriorpath)
603 604 605 606 607 608 609 610 611 612 613 614
    if first_dag:
      daglogdir=cp.get('paths','daglogdir')
      mkdirs(daglogdir)
      self.daglogfile=os.path.join(daglogdir,'lalinference_pipeline-'+str(uuid.uuid1())+'.log')
      pipeline.CondorDAG.__init__(self,self.daglogfile,dax=dax)
    elif not first_dag and previous_dag is not None:
      daglogdir=cp.get('paths','daglogdir')
      mkdirs(daglogdir)
      self.daglogfile=os.path.join(daglogdir,'lalinference_pipeline-'+str(uuid.uuid1())+'.log')
      pipeline.CondorDAG.__init__(self,self.daglogfile,dax=dax)
      for node in previous_dag.get_nodes():
        self.add_node(node)
615
    if cp.has_option('paths','cachedir'):
John Douglas Veitch's avatar
John Douglas Veitch committed
616
      self.cachepath=cp.get('paths','cachedir')
617 618
    else:
      self.cachepath=os.path.join(self.basepath,'caches')
619
    mkdirs(self.cachepath)
620
    if cp.has_option('paths','logdir'):
John Douglas Veitch's avatar
John Douglas Veitch committed
621
      self.logpath=cp.get('paths','logdir')
622 623
    else:
      self.logpath=os.path.join(self.basepath,'log')
624
    mkdirs(self.logpath)
625
    if cp.has_option('analysis','ifos'):
626
      self.ifos=ast.literal_eval(cp.get('analysis','ifos'))
627 628 629
    else:
      self.ifos=['H1','L1','V1']
    self.segments={}
630 631 632 633
    if cp.has_option('datafind','veto-categories'):
      self.veto_categories=cp.get('datafind','veto-categories')
    else: self.veto_categories=[]
    for ifo in self.ifos:
634
      self.segments[ifo]=[]
635 636 637
    self.computeroqweightsnode={}
    self.bayeslinenode={}
    self.bayeswavepsdnode={}
638
    self.dq={}
639
    self.frtypes=ast.literal_eval(cp.get('datafind','types'))
640
    self.channels=ast.literal_eval(cp.get('data','channels'))
John Douglas Veitch's avatar
John Douglas Veitch committed
641
    self.use_available_data=False
642
    self.webdir=cp.get('paths','webdir')
643 644 645 646
    if cp.has_option('analysis','dataseed'):
      self.dataseed=cp.getint('analysis','dataseed')
    else:
      self.dataseed=None
647
    # Set up necessary job files.
648
    self.prenodes={}
649
    self.datafind_job = pipeline.LSCDataFindJob(self.cachepath,self.logpath,self.config,dax=self.is_dax())
650
    self.datafind_job.add_opt('url-type','file')
651 652
    if cp.has_option('analysis','accounting_group'):
      self.datafind_job.add_condor_cmd('accounting_group',cp.get('analysis','accounting_group'))
653 654
    if cp.has_option('analysis','accounting_group_user'):
      self.datafind_job.add_condor_cmd('accounting_group_user',cp.get('analysis','accounting_group_user'))
655
    self.datafind_job.set_sub_file(os.path.abspath(os.path.join(self.basepath,'datafind.sub')))
656
    self.preengine_job = EngineJob(self.config, os.path.join(self.basepath,'prelalinference.sub'),self.logpath,engine='lalinferencedatadump',ispreengine=True,dax=self.is_dax())
657 658
    self.preengine_job.set_grid_site('local')
    self.preengine_job.set_universe('vanilla')
659 660 661 662 663 664
    if self.config.has_option('condor','computeroqweights'):
      self.computeroqweights_job = ROMJob(self.config,os.path.join(self.basepath,'computeroqweights.sub'),self.logpath,dax=self.is_dax())
      self.computeroqweights_job.set_grid_site('local')
    if self.config.has_option('condor','bayesline'):
      self.bayesline_job = BayesLineJob(self.config,os.path.join(self.basepath,'bayesline.sub'),self.logpath,dax=self.is_dax())
      self.bayesline_job.set_grid_site('local')
665 666 667 668 669
    self.bayeswavepsd_job={}
    if self.config.has_option('condor','bayeswave'):
      for ifo in self.ifos:
         self.bayeswavepsd_job[ifo] = BayesWavePSDJob(self.config,os.path.join(self.basepath,'bayeswavepsd_%s.sub'%(ifo)),self.logpath,dax=self.is_dax())
         self.bayeswavepsd_job[ifo].set_grid_site('local')
670 671 672 673
    # Need to create a job file for each IFO combination
    self.engine_jobs={}
    ifocombos=[]
    for N in range(1,len(self.ifos)+1):
674
        for a in permutations(self.ifos,N):
675 676
            ifocombos.append(a)
    for ifos in ifocombos:
677
        self.engine_jobs[ifos] = EngineJob(self.config, os.path.join(self.basepath,'engine_%s.sub'%(reduce(lambda x,y:x+y, map(str,ifos)))),self.logpath,engine=self.engine,dax=self.is_dax(), site=site)
678
    self.results_page_job = ResultsPageJob(self.config,os.path.join(self.basepath,'resultspage.sub'),self.logpath,dax=self.is_dax())
679
    self.results_page_job.set_grid_site('local')
680
    self.cotest_results_page_job = ResultsPageJob(self.config,os.path.join(self.basepath,'resultspagecoherent.sub'),self.logpath,dax=self.is_dax())
681
    self.cotest_results_page_job.set_grid_site('local')
682
    self.merge_job = MergeNSJob(self.config,os.path.join(self.basepath,'merge_runs.sub'),self.logpath,dax=self.is_dax())
683
    self.merge_job.set_grid_site('local')
684
    self.coherence_test_job = CoherenceTestJob(self.config,os.path.join(self.basepath,'coherence_test.sub'),self.logpath,dax=self.is_dax())
685
    self.coherence_test_job.set_grid_site('local')
686
    self.gracedbjob = GraceDBJob(self.config,os.path.join(self.basepath,'gracedb.sub'),self.logpath,dax=self.is_dax())
687
    self.gracedbjob.set_grid_site('local')
688
    # Process the input to build list of analyses to do
John Douglas Veitch's avatar
Working  
John Douglas Veitch committed
689
    self.events=self.setup_from_inputs()
690

691
    # Sanity checking
John Douglas Veitch's avatar
Working  
John Douglas Veitch committed
692
    if len(self.events)==0:
693 694
      print 'No input events found, please check your config if you expect some events'
    self.times=[e.trig_time for e in self.events]
695

696
    # Set up the segments
697
    if not (self.config.has_option('input','gps-start-time') and self.config.has_option('input','gps-end-time')) and len(self.times)>0:
698
      (mintime,maxtime)=self.get_required_data(self.times)
699 700 701 702
      if not self.config.has_option('input','gps-start-time'):
        self.config.set('input','gps-start-time',str(int(floor(mintime))))
      if not self.config.has_option('input','gps-end-time'):
        self.config.set('input','gps-end-time',str(int(ceil(maxtime))))
703
    self.add_science_segments()
704

705
    # Save the final configuration that is being used
706
    # first to the run dir
707 708
    conffilename=os.path.join(self.basepath,'config.ini')
    with open(conffilename,'wb') as conffile:
John Douglas Veitch's avatar
John Douglas Veitch committed
709
      self.config.write(conffile)
710 711 712 713
    if self.config.has_option('paths','webdir'):
      mkdirs(self.config.get('paths','webdir'))
      with open(os.path.join(self.config.get('paths','webdir'),'config.ini'),'wb') as conffile:
        self.config.write(conffile)
714

715
    # Generate the DAG according to the config given
716
    for event in self.events: self.add_full_analysis(event)
717
    self.add_skyarea_followup()
718 719 720
    if self.config.has_option('analysis','upload-to-gracedb'):
      if self.config.getboolean('analysis','upload-to-gracedb'):
        self.add_gracedb_FITSskymap_upload(self.events[0],engine=self.engine)
721 722 723 724 725 726 727 728 729
    self.dagfilename="lalinference_%s-%s"%(self.config.get('input','gps-start-time'),self.config.get('input','gps-end-time'))
    self.set_dag_file(self.dagfilename)
    if self.is_dax():
      self.set_dax_file(self.dagfilename)

  def add_skyarea_followup(self):
    # Add skyarea jobs if the executable is given
    # Do one for each results page for now
    if self.config.has_option('condor','skyarea'):
730 731
      self.skyareajob=SkyAreaJob(self.config,os.path.join(self.basepath,'skyarea.sub'),self.logpath,dax=self.is_dax())
      respagenodes=filter(lambda x: isinstance(x,ResultsPageNode) ,self.get_nodes())
732 733 734 735
      if self.engine=='lalinferenceburst':
          prefix='LIB_'
      else:
          prefix='LALInference_'
736
      for p in respagenodes:
737
          skyareanode=SkyAreaNode(self.skyareajob,prefix=prefix)
738
          skyareanode.add_resultspage_parent(p)
739
          skyareanode.set_ifos(p.ifos)
740
          self.add_node(skyareanode)
741

742
  def add_full_analysis(self,event):
743
    if self.engine=='lalinferencenest' or  self.engine=='lalinferenceburst':
744
      result=self.add_full_analysis_lalinferencenest(event)
745
    elif self.engine=='lalinferencemcmc':
746
      result=self.add_full_analysis_lalinferencemcmc(event)
747
    elif self.engine=='lalinferencebambi' or self.engine=='lalinferencebambimpi':
748 749 750
      result=self.add_full_analysis_lalinferencebambi(event)
    return result

751 752 753 754 755 756 757 758 759
  def create_frame_pfn_file(self):
    """
    Create a pegasus cache file name, uses inspiralutils
    """
    import inspiralutils as iu
    gpsstart=self.config.get('input','gps-start-time')
    gpsend=self.config.get('input','gps-end-time')
    pfnfile=iu.create_frame_pfn_file(self.frtypes,gpsstart,gpsend)
    return pfnfile
760

John Douglas Veitch's avatar
John Douglas Veitch committed
761 762
  def get_required_data(self,times):
    """
John Douglas Veitch's avatar
John Douglas Veitch committed
763
    Calculate the data that will be needed to process all events
John Douglas Veitch's avatar
John Douglas Veitch committed
764
    """
765
    #psdlength = self.config.getint('input','max-psd-length')
766
    padding=self.config.getint('input','padding')
767 768 769 770 771 772
    if self.config.has_option('engine','seglen') or self.config.has_option('lalinference','seglen'):
      if self.config.has_option('engine','seglen'):
        seglen = self.config.getint('engine','seglen')
      if self.config.has_option('lalinference','seglen'):
        seglen = self.config.getint('lalinference','seglen')

773
      if os.path.isfile(os.path.join(self.basepath,'psd.xml.gz')) or self.config.has_option('condor','bayesline') or self.config.has_option('condor','bayeswave'):
774
        psdlength = 0
775 776
        padding = 0
        self.config.set('input','padding',str(padding))
777 778 779
        if self.config.has_option('condor','bayeswave'):
          if (np.log2(seglen)%1):
            seglen = np.power(2., np.ceil(np.log2(seglen)))
780

781 782
      else:
        psdlength = 32*seglen
783 784
    else:
      seglen = max(e.duration for e in self.events)
785
      if os.path.isfile(os.path.join(self.basepath,'psd.xml.gz')) or self.config.has_option('condor','bayesline') or self.config.has_option('condor','bayeswave'):
786
        psdlength = 0
787 788
        padding = 0
        self.config.set('input','padding',str(padding))
789 790 791
        if self.config.has_option('condor','bayeswave'):
          if (np.log2(seglen)%1):
            seglen = np.power(2., np.ceil(np.log2(seglen)))
792 793 794
      else:
        psdlength = 32*seglen
    # Assume that the data interval is (end_time - seglen -padding , end_time + psdlength +padding )
795
    # -> change to (trig_time - seglen - padding - psdlength + 2 , trig_time + padding + 2) to estimate the psd before the trigger for online follow-up.
796
    # Also require padding before start time
797
    return (min(times)-padding-seglen-psdlength+2,max(times)+padding+2)
John Douglas Veitch's avatar
John Douglas Veitch committed
798

799 800 801 802 803
  def setup_from_times(self,times):
    """
    Generate a DAG from a list of times
    """
    for time in self.times:
804
      self.add_full_analysis(Event(trig_time=time))
805

806 807 808 809 810 811 812 813 814 815 816 817 818 819 820 821 822 823 824 825 826 827 828 829
  def select_events(self):
    """
    Read events from the config parser. Understands both ranges and comma separated events, or combinations
    eg. events=[0,1,5:10,21] adds to the analysis the events: 0,1,5,6,7,8,9,10 and 21
    """
    events=[]
    times=[]
    raw_events=self.config.get('input','events').replace('[','').replace(']','').split(',')
    for raw_event in raw_events:
        if ':' in raw_event:
            limits=raw_event.split</