# -*- coding: utf-8 -*-
# vim: tabstop=4 shiftwidth=4 softtabstop=4
#
# Copyright (C) 2014-2018 GEM Foundation
#
# OpenQuake is free software: you can redistribute it and/or modify it
# under the terms of the GNU Affero General Public License as published
# by the Free Software Foundation, either version 3 of the License, or
# (at your option) any later version.
#
# OpenQuake 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 Affero General Public License for more details.
#
# You should have received a copy of the GNU Affero General Public License
# along with OpenQuake. If not, see <http://www.gnu.org/licenses/>.
from __future__ import division
import math
import logging
import operator
import numpy
from openquake.baselib import parallel, hdf5
from openquake.baselib.python3compat import encode
from openquake.baselib.general import AccumDict, block_splitter
from openquake.hazardlib.calc.hazard_curve import classical, ProbabilityMap
from openquake.hazardlib.stats import compute_pmap_stats
from openquake.hazardlib import source
from openquake.hazardlib.calc.filters import SourceFilter
from openquake.calculators import getters
from openquake.calculators import base
U16 = numpy.uint16
U32 = numpy.uint32
F32 = numpy.float32
F64 = numpy.float64
weight = operator.attrgetter('weight')
grp_source_dt = numpy.dtype([('grp_id', U16), ('source_id', hdf5.vstr),
('source_name', hdf5.vstr)])
source_data_dt = numpy.dtype(
[('taskno', U16), ('nsites', U32), ('nruptures', U32), ('weight', F32)])
[docs]def get_src_ids(sources):
"""
:returns:
a string with the source IDs of the given sources, stripping the
extension after the colon, if any
"""
src_ids = []
for src in sources:
long_src_id = src.source_id
try:
src_id, ext = long_src_id.rsplit(':', 1)
except ValueError:
src_id = long_src_id
src_ids.append(src_id)
return ' '.join(set(src_ids))
[docs]def saving_sources_by_task(iterargs, dstore):
"""
Yield the iterargs again by populating 'task_info/source_data'
"""
source_ids = []
data = []
for i, args in enumerate(iterargs, 1):
source_ids.append(get_src_ids(args[0]))
for src in args[0]: # collect source data
data.append((i, src.nsites, src.num_ruptures, src.weight))
yield args
dstore['task_info/task_sources'] = encode(source_ids)
dstore.extend('task_info/source_data', numpy.array(data, source_data_dt))
[docs]@base.calculators.add('psha')
class PSHACalculator(base.HazardCalculator):
"""
Classical PSHA calculator
"""
core_task = classical
[docs] def agg_dicts(self, acc, pmap_by_grp):
"""
Aggregate dictionaries of hazard curves by updating the accumulator.
:param acc: accumulator dictionary
:param pmap_by_grp: dictionary grp_id -> ProbabilityMap
"""
with self.monitor('aggregate curves', autoflush=True):
acc.eff_ruptures += pmap_by_grp.eff_ruptures
for grp_id in pmap_by_grp:
if pmap_by_grp[grp_id]:
acc[grp_id] |= pmap_by_grp[grp_id]
self.nsites.append(len(pmap_by_grp[grp_id]))
for srcid, (srcweight, nsites, calc_time, split) in \
pmap_by_grp.calc_times.items():
info = self.csm.infos[srcid]
info.num_sites += nsites
info.calc_time += calc_time
info.num_split += split
return acc
[docs] def zerodict(self):
"""
Initial accumulator, a dict grp_id -> ProbabilityMap(L, G)
"""
csm_info = self.csm.info
zd = AccumDict()
num_levels = len(self.oqparam.imtls.array)
for grp in self.csm.src_groups:
num_gsims = len(csm_info.gsim_lt.get_gsims(grp.trt))
zd[grp.id] = ProbabilityMap(num_levels, num_gsims)
zd.eff_ruptures = AccumDict() # grp_id -> eff_ruptures
return zd
[docs] def execute(self):
"""
Run in parallel `core_task(sources, sitecol, monitor)`, by
parallelizing on the sources according to their weight and
tectonic region type.
"""
try:
self.csm
except AttributeError:
raise RuntimeError('No CompositeSourceModel, did you forget to '
'run the hazard or the --hc option?')
with self.monitor('managing sources', autoflush=True):
allargs = self.gen_args(self.monitor('classical'))
iterargs = saving_sources_by_task(allargs, self.datastore)
if isinstance(allargs, list):
# there is a trick here: if the arguments are known
# (a list, not an iterator), keep them as a list
# then the Starmap will understand the case of a single
# argument tuple and it will run in core the task
iterargs = list(iterargs)
ires = parallel.Starmap(
self.core_task.__func__, iterargs).submit_all()
self.nsites = []
acc = ires.reduce(self.agg_dicts, self.zerodict())
if not self.nsites:
raise RuntimeError('All sources were filtered out!')
logging.info('Effective sites per task: %d', numpy.mean(self.nsites))
with self.monitor('store source_info', autoflush=True):
self.store_source_info(self.csm.infos, acc)
return acc
[docs] def gen_args(self, monitor):
"""
Used in the case of large source model logic trees.
:param monitor: a :class:`openquake.baselib.performance.Monitor`
:yields: (sources, sites, gsims, monitor) tuples
"""
oq = self.oqparam
opt = self.oqparam.optimize_same_id_sources
num_tiles = math.ceil(len(self.sitecol) / oq.sites_per_tile)
tasks_per_tile = math.ceil(oq.concurrent_tasks / math.sqrt(num_tiles))
if num_tiles > 1:
tiles = self.sitecol.split_in_tiles(num_tiles)
else:
tiles = [self.sitecol]
param = dict(truncation_level=oq.truncation_level, imtls=oq.imtls)
minweight = source.MINWEIGHT * math.sqrt(len(self.sitecol))
totweight = 0
for tile_i, tile in enumerate(tiles, 1):
num_tasks = 0
num_sources = 0
if num_tiles > 1:
logging.info('Prefiltering tile %d of %d', tile_i, len(tiles))
else:
logging.info('Prefiltering sources')
with self.monitor('prefiltering'):
src_filter = SourceFilter(tile, oq.maximum_distance)
csm = self.csm.filter(src_filter)
if tile_i == 1: # set it only on the first tile
maxweight = csm.get_maxweight(
weight, tasks_per_tile, minweight)
if maxweight == minweight:
logging.info('Using minweight=%d', minweight)
else:
logging.info('Using maxweight=%d', maxweight)
totweight += csm.info.tot_weight
else:
totweight += csm.get_weight(weight)
if csm.has_dupl_sources and not opt:
logging.warn('Found %d duplicated sources',
csm.has_dupl_sources)
for sg in csm.src_groups:
if sg.src_interdep == 'mutex':
gsims = self.csm.info.gsim_lt.get_gsims(sg.trt)
yield sg, csm.src_filter, gsims, param, monitor
num_tasks += 1
num_sources += len(sg.sources)
# NB: csm.get_sources_by_trt discards the mutex sources
for trt, sources in csm.get_sources_by_trt().items():
gsims = self.csm.info.gsim_lt.get_gsims(trt)
for block in block_splitter(sources, maxweight, weight):
yield block, src_filter, gsims, param, monitor
num_tasks += 1
num_sources += len(block)
logging.info('Sent %d sources in %d tasks', num_sources, num_tasks)
self.csm.info.tot_weight = totweight
[docs] def post_execute(self, pmap_by_grp_id):
"""
Collect the hazard curves by realization and export them.
:param pmap_by_grp_id:
a dictionary grp_id -> hazard curves
"""
oq = self.oqparam
grp_trt = self.csm.info.grp_by("trt")
grp_source = self.csm.info.grp_by("name")
if oq.disagg_by_src:
src_name = {src.src_group_id: src.name
for src in self.csm.get_sources()}
data = []
with self.monitor('saving probability maps', autoflush=True):
for grp_id, pmap in pmap_by_grp_id.items():
if pmap: # pmap can be missing if the group is filtered away
fix_ones(pmap) # avoid saving PoEs == 1
key = 'poes/grp-%02d' % grp_id
self.datastore[key] = pmap
self.datastore.set_attrs(key, trt=grp_trt[grp_id])
if oq.disagg_by_src:
data.append(
(grp_id, grp_source[grp_id], src_name[grp_id]))
if 'poes' in self.datastore:
self.datastore.set_nbytes('poes')
if oq.disagg_by_src and self.csm.info.get_num_rlzs() == 1:
# this is useful for disaggregation, which is implemented
# only for the case of a single realization
self.datastore['disagg_by_src/source_id'] = numpy.array(
sorted(data), grp_source_dt)
[docs]def fix_ones(pmap):
"""
Physically, an extremely small intensity measure level can have an
extremely large probability of exceedence, however that probability
cannot be exactly 1 unless the level is exactly 0. Numerically, the
PoE can be 1 and this give issues when calculating the damage (there
is a log(0) in
:class:`openquake.risklib.scientific.annual_frequency_of_exceedence`).
Here we solve the issue by replacing the unphysical probabilities 1
with .9999999999999999 (the float64 closest to 1).
"""
for sid in pmap:
array = pmap[sid].array
array[array == 1.] = .9999999999999999
[docs]def build_hcurves_and_stats(pgetter, hstats, monitor):
"""
:param pgetter: an :class:`openquake.commonlib.getters.PmapGetter`
:param hstats: a list of pairs (statname, statfunc)
:param monitor: instance of Monitor
:returns: a dictionary kind -> ProbabilityMap
The "kind" is a string of the form 'rlz-XXX' or 'mean' of 'quantile-XXX'
used to specify the kind of output.
"""
with monitor('combine pmaps'):
pgetter.init() # if not already initialized
try:
pmaps = pgetter.get_pmaps(pgetter.sids)
except IndexError: # no data
return {}
if sum(len(pmap) for pmap in pmaps) == 0: # no data
return {}
pmap_by_kind = {}
for kind, stat in hstats:
with monitor('compute ' + kind):
pmap = compute_pmap_stats(pmaps, [stat], pgetter.weights)
pmap_by_kind[kind] = pmap
return pmap_by_kind
[docs]@base.calculators.add('classical')
class ClassicalCalculator(PSHACalculator):
"""
Classical PSHA calculator
"""
pre_calculator = 'psha'
core_task = build_hcurves_and_stats
[docs] def execute(self):
"""
Build statistical hazard curves from the stored PoEs
"""
if 'poes' not in self.datastore: # for short report
return
oq = self.oqparam
num_rlzs = self.datastore['csm_info'].get_num_rlzs()
if num_rlzs == 1: # no stats to compute
return {}
elif not oq.hazard_stats():
if oq.hazard_maps or oq.uniform_hazard_spectra:
logging.warn('mean_hazard_curves was false in the job.ini, '
'so no outputs were generated.\nYou can compute '
'the statistics without repeating the calculation'
' with the --hc option')
return {}
# initialize datasets
N = len(self.sitecol)
L = len(oq.imtls.array)
attrs = dict(
__pyclass__='openquake.hazardlib.probability_map.ProbabilityMap',
sids=numpy.arange(N, dtype=numpy.uint32))
nbytes = N * L * 4 # bytes per realization (32 bit floats)
totbytes = 0
if num_rlzs > 1:
for name, stat in oq.hazard_stats():
self.datastore.create_dset(
'hcurves/' + name, F32, (N, L, 1), attrs=attrs)
totbytes += nbytes
if 'hcurves' in self.datastore:
self.datastore.set_attrs('hcurves', nbytes=totbytes)
self.datastore.flush()
with self.monitor('sending pmaps', autoflush=True, measuremem=True):
ires = parallel.Starmap(
self.core_task.__func__, self.gen_args()
).submit_all()
nbytes = ires.reduce(self.save_hcurves)
return nbytes
[docs] def gen_args(self):
"""
:yields: pgetter, hstats, monitor
"""
monitor = self.monitor('build_hcurves_and_stats')
hstats = self.oqparam.hazard_stats()
parent = self.can_read_parent()
if parent is None:
parent = self.datastore
for t in self.sitecol.split_in_tiles(self.oqparam.concurrent_tasks):
pgetter = getters.PmapGetter(parent, t.sids, self.rlzs_assoc)
if parent is self.datastore: # read now, not in the workers
logging.info('Reading PoEs on %d sites', len(t))
pgetter.init()
yield pgetter, hstats, monitor
[docs] def save_hcurves(self, acc, pmap_by_kind):
"""
Works by side effect by saving hcurves and statistics on the
datastore; the accumulator stores the number of bytes saved.
:param acc: dictionary kind -> nbytes
:param pmap_by_kind: a dictionary of ProbabilityMaps
"""
with self.monitor('saving statistical hcurves', autoflush=True):
for kind in pmap_by_kind:
pmap = pmap_by_kind[kind]
if pmap:
key = 'hcurves/' + kind
dset = self.datastore.getitem(key)
for sid in pmap:
dset[sid] = pmap[sid].array
# in the datastore we save 4 byte floats, thus we
# divide the memory consumption by 2: pmap.nbytes / 2
acc += {kind: pmap.nbytes // 2}
self.datastore.flush()
return acc
[docs] def post_execute(self, acc):
"""Save the number of bytes per each dataset"""
for kind, nbytes in acc.items():
self.datastore.getitem('hcurves/' + kind).attrs['nbytes'] = nbytes