def calc_risk(gmfs, param, monitor):
"""
:param gmfs: an array of GMFs with fields sid, eid, gmv
:param param: a dictionary of parameters coming from the job.ini
:param monitor: a Monitor instance
:returns: a dictionary of arrays with keys elt, alt, losses_by_A, ...
"""
mon_risk = monitor('computing risk', measuremem=False)
mon_agg = monitor('aggregating losses', measuremem=False)
mon_avg = monitor('averaging losses', measuremem=False)
dstore = datastore.read(param['hdf5path'])
with monitor('getting assets'):
assets_df = dstore.read_df('assetcol/array', 'ordinal')
with monitor('getting crmodel'):
crmodel = monitor.read('crmodel')
weights = dstore['weights'][()]
L = len(param['elt'].loss_names)
aggkey = param['aggkey']
elt_dt = [('event_id', U32), ('loss', (F32, (L, )))]
acc = dict(events_per_sid=0)
elt = param['elt']
tempname = param['tempname']
aggby = param['aggregate_by']
haz_by_sid = general.group_array(gmfs, 'sid')
losses_by_A = numpy.zeros((len(assets_df), len(elt.loss_names)), F32)
for sid, asset_df in assets_df.groupby('site_id'):
try:
haz = haz_by_sid[sid]
except KeyError: # no hazard here
continue
with mon_risk:
assets = asset_df.to_records() # fast
acc['events_per_sid'] += len(haz)
assets_by_taxo = get_assets_by_taxo(assets, tempname) # fast
out = get_output(crmodel, assets_by_taxo, haz) # slow
with mon_agg:
elt.aggregate(out, param['minimum_asset_loss'], aggby)
# NB: after the aggregation out contains losses, not loss_ratios
if param['avg_losses']:
with mon_avg:
ws = weights[haz['rlz']]
for lni, ln in enumerate(elt.loss_names):
losses_by_A[assets['ordinal'], lni] += out[ln] @ ws
if len(gmfs):
acc['events_per_sid'] /= len(gmfs)
acc['alt'] = alt = {}
for key, k in aggkey.items():
s = ','.join(map(str, key)) + ','
alt[s] = numpy.array([(eid, arr[k])
for eid, arr in elt.items() if arr[k].sum()],
elt_dt)
# in the demo there are 264/1694 nonzero events, i.e. arr[k].sum()
# is zero most of the time
if param['avg_losses']:
acc['losses_by_A'] = losses_by_A * param['ses_ratio']
return acc
def calc_risk(df, param, monitor):
"""
:param df: a DataFrame of GMFs with fields sid, eid, gmv_...
:param param: a dictionary of parameters coming from the job.ini
:param monitor: a Monitor instance
:returns: a dictionary of arrays with keys alt, losses_by_A
"""
mon_risk = monitor('computing risk', measuremem=False)
mon_agg = monitor('aggregating losses', measuremem=False)
mon_avg = monitor('averaging losses', measuremem=False)
dstore = datastore.read(param['hdf5path'])
with monitor('getting assets'):
assets_df = dstore.read_df('assetcol/array', 'ordinal')
with monitor('getting crmodel'):
crmodel = monitor.read('crmodel')
weights = dstore['weights'][()]
acc = dict(events_per_sid=0)
alt = copy.copy(param['alt']) # avoid issues with OQ_DISTRIBUTE=no
tempname = param['tempname']
aggby = param['aggregate_by']
mal = param['minimum_asset_loss']
haz_by_sid = {s: d for s, d in df.groupby('sid')}
losses_by_A = numpy.zeros((len(assets_df), len(alt.loss_names)), F32)
acc['avg_gmf'] = avg_gmf = {}
for col in df.columns:
if col not in 'sid eid rlz':
avg_gmf[col] = numpy.zeros(param['N'], F32)
for sid, asset_df in assets_df.groupby('site_id'):
try:
haz = haz_by_sid[sid]
except KeyError: # no hazard here
continue
with mon_risk:
assets = asset_df.to_records() # fast
acc['events_per_sid'] += len(haz)
assets_by_taxo = get_assets_by_taxo(assets, tempname) # fast
out = get_output(crmodel, assets_by_taxo, haz) # slow
with mon_agg:
alt.aggregate(out, mal, aggby)
# NB: after the aggregation out contains losses, not loss_ratios
ws = weights[haz['rlz']]
for col in df.columns:
if col not in 'sid eid rlz':
avg_gmf[col][sid] = haz[col] @ ws
if param['avg_losses']:
with mon_avg:
for lni, ln in enumerate(alt.loss_names):
losses_by_A[assets['ordinal'], lni] += out[ln] @ ws
if len(df):
acc['events_per_sid'] /= len(df)
acc['alt'] = alt.to_dframe()
if param['avg_losses']:
acc['losses_by_A'] = losses_by_A * param['ses_ratio']
return acc
def _calc_risk(hazard, param, monitor):
gmfs = numpy.concatenate(hazard['gmfs'])
events = numpy.concatenate(hazard['events'])
mon_risk = monitor('computing risk', measuremem=False)
mon_agg = monitor('aggregating losses', measuremem=False)
dstore = datastore.read(param['hdf5path'])
with monitor('getting assets'):
assetcol = dstore['assetcol']
assets_by_site = assetcol.assets_by_site()
with monitor('getting crmodel'):
crmodel = riskmodels.CompositeRiskModel.read(dstore)
weights = dstore['weights'][()]
E = len(events)
L = len(param['lba'].loss_names)
A = sum(len(assets) for assets in assets_by_site)
shape = assetcol.tagcol.agg_shape((E, L), param['aggregate_by'])
elt_dt = [('event_id', U32), ('rlzi', U16), ('loss', (F32, shape[1:]))]
acc = dict(
elt=numpy.zeros(shape, F32), # shape (E, L, T...)
alt=numpy.zeros((A, E, L), F32) if param['asset_loss_table'] else None,
gmf_info=[],
events_per_sid=0,
lossbytes=0)
arr = acc['elt']
alt = acc['alt']
lba = param['lba']
tempname = param['tempname']
tagnames = param['aggregate_by']
eid2rlz = dict(events[['id', 'rlz_id']])
eid2idx = {eid: idx for idx, eid in enumerate(eid2rlz)}
for sid, haz in general.group_array(gmfs, 'sid').items():
assets_on_sid = assets_by_site[sid]
if len(assets_on_sid) == 0:
continue
acc['events_per_sid'] += len(haz)
if param['avg_losses']:
ws = weights[[eid2rlz[eid] for eid in haz['eid']]]
assets_by_taxo = get_assets_by_taxo(assets_on_sid, tempname)
eidx = [eid2idx[eid] for eid in haz['eid']]
with mon_risk:
out = get_output(crmodel, assets_by_taxo, haz)
with mon_agg:
for a, asset in enumerate(assets_on_sid):
aid = asset['ordinal']
tagi = asset[tagnames] if tagnames else ()
tagidxs = tuple(idx - 1 for idx in tagi)
losses_by_lt = {}
for lti, lt in enumerate(crmodel.loss_types):
lratios = out[lt][a]
if lt == 'occupants':
losses = lratios * asset['occupants_None']
else:
losses = lratios * asset['value-' + lt]
if param['asset_loss_table']:
alt[aid, eidx, lti] = losses
losses_by_lt[lt] = losses
for loss_idx, losses in lba.compute(asset, losses_by_lt):
arr[(eidx, loss_idx) + tagidxs] += losses
if param['avg_losses']:
lba.losses_by_A[aid, loss_idx] += (losses @ ws *
param['ses_ratio'])
acc['lossbytes'] += losses.nbytes
if len(gmfs):
acc['events_per_sid'] /= len(gmfs)
acc['gmf_info'] = numpy.array(hazard['gmf_info'], gmf_info_dt)
acc['elt'] = numpy.fromiter( # this is ultra-fast
(
(event['id'], event['rlz_id'], losses) # losses (L, T...)
for event, losses in zip(events, arr) if losses.sum()),
elt_dt)
if param['avg_losses']:
acc['losses_by_A'] = param['lba'].losses_by_A
# without resetting the cache the sequential avg_losses would be wrong!
del param['lba'].__dict__['losses_by_A']
if param['asset_loss_table']:
acc['alt'] = alt, events['id']
return acc
def ebrisk(rupgetter, srcfilter, param, monitor):
"""
:param rupgetter:
a RuptureGetter instance
:param srcfilter:
a SourceFilter instance
:param param:
a dictionary of parameters
:param monitor:
:class:`openquake.baselib.performance.Monitor` instance
:returns:
an ArrayWrapper with shape (E, L, T, ...)
"""
riskmodel = param['riskmodel']
E = rupgetter.num_events
L = len(riskmodel.lti)
N = len(srcfilter.sitecol.complete)
e1 = rupgetter.first_event
with monitor('getting assets', measuremem=False):
with datastore.read(srcfilter.filename) as dstore:
assetcol = dstore['assetcol']
assets_by_site = assetcol.assets_by_site()
A = len(assetcol)
getter = getters.GmfGetter(rupgetter, srcfilter, param['oqparam'])
with monitor('getting hazard'):
getter.init() # instantiate the computers
hazard = getter.get_hazard() # sid -> (sid, eid, gmv)
mon_risk = monitor('computing risk', measuremem=False)
mon_agg = monitor('aggregating losses', measuremem=False)
events = rupgetter.get_eid_rlz()
# numpy.testing.assert_equal(events['eid'], sorted(events['eid']))
eid2idx = dict(zip(events['eid'], range(e1, e1 + E)))
tagnames = param['aggregate_by']
shape = assetcol.tagcol.agg_shape((E, L), tagnames)
elt_dt = [('eid', U64), ('rlzi', U16), ('loss', (F32, shape[1:]))]
if param['asset_loss_table']:
alt = numpy.zeros((A, E, L), F32)
acc = numpy.zeros(shape, F32) # shape (E, L, T...)
if param['avg_losses']:
losses_by_A = numpy.zeros((A, L), F32)
else:
losses_by_A = 0
# NB: IMT-dependent weights are not supported in ebrisk
times = numpy.zeros(N) # risk time per site_id
num_events_per_sid = 0
epspath = param['epspath']
gmf_nbytes = 0
for sid, haz in hazard.items():
gmf_nbytes += haz.nbytes
t0 = time.time()
assets_on_sid = assets_by_site[sid]
if len(assets_on_sid) == 0:
continue
num_events_per_sid += len(haz)
if param['avg_losses']:
weights = getter.weights[
[getter.eid2rlz[eid] for eid in haz['eid']]]
assets_by_taxo = get_assets_by_taxo(assets_on_sid, epspath)
eidx = numpy.array([eid2idx[eid] for eid in haz['eid']]) - e1
haz['eid'] = eidx + e1
with mon_risk:
out = riskmodel.get_output(assets_by_taxo, haz)
with mon_agg:
for a, asset in enumerate(assets_on_sid):
aid = asset['ordinal']
tagi = asset[tagnames] if tagnames else ()
tagidxs = tuple(idx - 1 for idx in tagi)
for lti, lt in enumerate(riskmodel.loss_types):
lratios = out[lt][a]
if lt == 'occupants':
losses = lratios * asset['occupants_None']
else:
losses = lratios * asset['value-' + lt]
if param['asset_loss_table']:
alt[aid, eidx, lti] = losses
acc[(eidx, lti) + tagidxs] += losses
if param['avg_losses']:
losses_by_A[aid, lti] += losses @ weights
times[sid] = time.time() - t0
if hazard:
num_events_per_sid /= len(hazard)
with monitor('building event loss table'):
elt = numpy.fromiter(
((event['eid'], event['rlz'], losses)
for event, losses in zip(events, acc) if losses.sum()), elt_dt)
agg = general.AccumDict(accum=numpy.zeros(shape[1:], F32)) # rlz->agg
for rec in elt:
agg[rec['rlzi']] += rec['loss'] * param['ses_ratio']
res = {'elt': elt, 'agg_losses': agg, 'times': times,
'events_per_sid': num_events_per_sid, 'gmf_nbytes': gmf_nbytes}
if param['avg_losses']:
res['losses_by_A'] = losses_by_A * param['ses_ratio']
if param['asset_loss_table']:
eidx = numpy.array([eid2idx[eid] for eid in events['eid']])
res['alt_eidx'] = alt, eidx
return res
def ebrisk(rupgetter, srcfilter, param, monitor):
"""
:param rupgetter:
a RuptureGetter instance
:param srcfilter:
a SourceFilter instance
:param param:
a dictionary of parameters
:param monitor:
:class:`openquake.baselib.performance.Monitor` instance
:returns:
an ArrayWrapper with shape (E, L, T, ...)
"""
riskmodel = param['riskmodel']
E = rupgetter.num_events
L = len(riskmodel.lti)
N = len(srcfilter.sitecol.complete)
e1 = rupgetter.first_event
with monitor('getting assets', measuremem=False):
with datastore.read(srcfilter.filename) as dstore:
assetcol = dstore['assetcol']
assets_by_site = assetcol.assets_by_site()
A = len(assetcol)
getter = getters.GmfGetter(rupgetter, srcfilter, param['oqparam'])
with monitor('getting hazard'):
getter.init() # instantiate the computers
hazard = getter.get_hazard_by_sid() # sid -> (sid, eid, gmv)
mon_risk = monitor('computing risk', measuremem=False)
mon_agg = monitor('aggregating losses', measuremem=False)
events = rupgetter.get_eid_rlz()
# numpy.testing.assert_equal(events['eid'], sorted(events['eid']))
eid2idx = dict(zip(events['eid'], range(e1, e1 + E)))
tagnames = param['aggregate_by']
shape = assetcol.tagcol.agg_shape((E, L), tagnames)
elt_dt = [('eid', U64), ('rlzi', U16), ('loss', (F32, shape[1:]))]
if param['asset_loss_table']:
alt = numpy.zeros((A, E, L), F32)
acc = numpy.zeros(shape, F32) # shape (E, L, T...)
if param['avg_losses']:
losses_by_A = numpy.zeros((A, L), F32)
else:
losses_by_A = 0
# NB: IMT-dependent weights are not supported in ebrisk
times = numpy.zeros(N) # risk time per site_id
num_events_per_sid = 0
epspath = param['epspath']
gmf_nbytes = 0
for sid, haz in hazard.items():
gmf_nbytes += haz.nbytes
t0 = time.time()
assets_on_sid = assets_by_site[sid]
if len(assets_on_sid) == 0:
continue
num_events_per_sid += len(haz)
if param['avg_losses']:
weights = getter.weights[[
getter.eid2rlz[eid] for eid in haz['eid']
]]
assets_by_taxo = get_assets_by_taxo(assets_on_sid, epspath)
eidx = numpy.array([eid2idx[eid] for eid in haz['eid']]) - e1
haz['eid'] = eidx + e1
with mon_risk:
out = riskmodel.get_output(assets_by_taxo, haz)
with mon_agg:
for a, asset in enumerate(assets_on_sid):
aid = asset['ordinal']
tagi = asset[tagnames] if tagnames else ()
tagidxs = tuple(idx - 1 for idx in tagi)
for lti, lt in enumerate(riskmodel.loss_types):
lratios = out[lt][a]
if lt == 'occupants':
losses = lratios * asset['occupants_None']
else:
losses = lratios * asset['value-' + lt]
if param['asset_loss_table']:
alt[aid, eidx, lti] = losses
acc[(eidx, lti) + tagidxs] += losses
if param['avg_losses']:
losses_by_A[aid, lti] += losses @ weights
times[sid] = time.time() - t0
if hazard:
num_events_per_sid /= len(hazard)
with monitor('building event loss table'):
elt = numpy.fromiter(
((event['eid'], event['rlz'], losses)
for event, losses in zip(events, acc) if losses.sum()), elt_dt)
agg = general.AccumDict(accum=numpy.zeros(shape[1:], F32)) # rlz->agg
for rec in elt:
agg[rec['rlzi']] += rec['loss'] * param['ses_ratio']
res = {
'elt': elt,
'agg_losses': agg,
'times': times,
'events_per_sid': num_events_per_sid,
'gmf_nbytes': gmf_nbytes
}
if param['avg_losses']:
res['losses_by_A'] = losses_by_A * param['ses_ratio']
if param['asset_loss_table']:
eidx = numpy.array([eid2idx[eid] for eid in events['eid']])
res['alt_eidx'] = alt, eidx
return res
def calc_risk(gmfs, param, monitor):
"""
:param gmfs: an array of GMFs with fields sid, eid, gmv
:param param: a dictionary of parameters coming from the job.ini
:param monitor: a Monitor instance
:returns: a dictionary of arrays with keys elt, alt, losses_by_A, ...
"""
mon_risk = monitor('computing risk', measuremem=False)
mon_agg = monitor('aggregating losses', measuremem=False)
dstore = datastore.read(param['hdf5path'])
with monitor('getting assets'):
assets_df = dstore.read_df('assetcol/array', 'ordinal')
with monitor('getting crmodel'):
crmodel = monitor.read('crmodel')
weights = dstore['weights'][()]
L = len(param['lba'].loss_names)
elt_dt = [('event_id', U32), ('loss', (F32, (L, )))]
# aggkey -> eid -> loss
acc = dict(events_per_sid=0, numlosses=numpy.zeros(2, int)) # (kept, tot)
lba = param['lba']
lba.alt = general.AccumDict( # idx -> eid -> loss
accum=general.AccumDict(accum=numpy.zeros(L, F32)))
lba.losses_by_E = general.AccumDict( # eid -> loss
accum=numpy.zeros(L, F32))
tempname = param['tempname']
aggby = param['aggregate_by']
minimum_loss = []
for lt, lti in crmodel.lti.items():
val = param['minimum_asset_loss'][lt]
minimum_loss.append(val)
if lt in lba.policy_dict: # same order as in lba.compute
minimum_loss.append(val)
haz_by_sid = general.group_array(gmfs, 'sid')
for sid, asset_df in assets_df.groupby('site_id'):
try:
haz = haz_by_sid[sid]
except KeyError: # no hazard here
continue
with mon_risk:
assets = asset_df.to_records() # fast
acc['events_per_sid'] += len(haz)
if param['avg_losses']:
ws = weights[haz['rlz']]
else:
ws = None
assets_by_taxo = get_assets_by_taxo(assets, tempname) # fast
out = get_output(crmodel, assets_by_taxo, haz) # slow
with mon_agg:
tagidxs = assets[aggby] if aggby else None
acc['numlosses'] += lba.aggregate(out, haz['eid'], minimum_loss,
tagidxs, ws)
if len(gmfs):
acc['events_per_sid'] /= len(gmfs)
acc['elt'] = numpy.fromiter( # this is ultra-fast
((eid, losses)
for eid, losses in lba.losses_by_E.items() if losses.sum()), elt_dt)
acc['alt'] = {
idx: numpy.fromiter( # already sorted by aid, ultra-fast
((eid, loss) for eid, loss in lba.alt[idx].items()), elt_dt)
for idx in lba.alt
}
if param['avg_losses']:
acc['losses_by_A'] = param['lba'].losses_by_A * param['ses_ratio']
# without resetting the cache the sequential avg_losses would be wrong!
del param['lba'].__dict__['losses_by_A']
return acc
def calc_risk(gmfs, param, monitor):
mon_risk = monitor('computing risk', measuremem=False)
mon_agg = monitor('aggregating losses', measuremem=False)
eids = numpy.unique(gmfs['eid'])
dstore = datastore.read(param['hdf5path'])
with monitor('getting assets'):
assetcol = dstore['assetcol']
assets_by_site = assetcol.assets_by_site()
exposed_values = dstore['exposed_values/agg'][()]
with monitor('getting crmodel'):
crmodel = riskmodels.CompositeRiskModel.read(dstore)
events = dstore['events'][list(eids)]
weights = dstore['weights'][()]
E = len(eids)
L = len(param['lba'].loss_names)
elt_dt = [('event_id', U32), ('rlzi', U16), ('loss', (F32, (L, )))]
alt = general.AccumDict(accum=numpy.zeros(L, F32)) # aid, eid -> loss
arr = numpy.zeros((E, L), F32)
acc = dict(events_per_sid=0, numlosses=numpy.zeros(2, int)) # (kept, tot)
lba = param['lba']
tempname = param['tempname']
eid2rlz = dict(events[['id', 'rlz_id']])
eid2idx = {eid: idx for idx, eid in enumerate(eids)}
minimum_loss = []
fraction = param['minimum_loss_fraction'] / len(assetcol)
for lt, lti in crmodel.lti.items():
val = exposed_values[lti] * fraction
minimum_loss.append(val)
if lt in lba.policy_dict: # same order as in lba.compute
minimum_loss.append(val)
for sid, haz in general.group_array(gmfs, 'sid').items():
assets_on_sid = assets_by_site[sid]
if len(assets_on_sid) == 0:
continue
acc['events_per_sid'] += len(haz)
if param['avg_losses']:
ws = weights[[eid2rlz[eid] for eid in haz['eid']]]
assets_by_taxo = get_assets_by_taxo(assets_on_sid, tempname)
eidx = numpy.array([eid2idx[eid] for eid in haz['eid']])
with mon_risk:
out = get_output(crmodel, assets_by_taxo, haz)
for lti, lt in enumerate(crmodel.loss_types):
lratios = out[lt]
if lt == 'occupants':
field = 'occupants_None'
else:
field = 'value-' + lt
for a, asset in enumerate(assets_on_sid):
aid = asset['ordinal']
ls = asset[field] * lratios[a]
for loss_idx, losses in lba.compute(asset, ls, lt):
kept = 0
with mon_agg:
if param['aggregate_by']:
for loss, eid in zip(losses, out.eids):
if loss >= minimum_loss[loss_idx]:
alt[aid, eid][loss_idx] = loss
kept += 1
arr[eidx, loss_idx] += losses
if param['avg_losses']: # this is really fast
lba.losses_by_A[aid, loss_idx] += losses @ ws
acc['numlosses'] += numpy.array([kept, len(losses)])
if len(gmfs):
acc['events_per_sid'] /= len(gmfs)
acc['elt'] = numpy.fromiter( # this is ultra-fast
((event['id'], event['rlz_id'], losses)
for event, losses in zip(events, arr) if losses.sum()), elt_dt)
acc['alt'] = alt = numpy.fromiter( # already sorted by aid
((aid, eid, eid2rlz[eid], loss) for (aid, eid), loss in alt.items()),
param['ael_dt'])
alt.sort(order='rlzi')
acc['indices'] = general.get_indices(alt['rlzi'])
if param['avg_losses']:
acc['losses_by_A'] = param['lba'].losses_by_A * param['ses_ratio']
# without resetting the cache the sequential avg_losses would be wrong!
del param['lba'].__dict__['losses_by_A']
return acc
def calc_risk(gmfs, param, monitor):
"""
:param gmfs: an array of GMFs with fields sid, eid, gmv
:param param: a dictionary of parameters coming from the job.ini
:param monitor: a Monitor instance
:returns: a dictionary of arrays with keys alt, losses_by_A
"""
mon_risk = monitor('computing risk', measuremem=False)
mon_agg = monitor('aggregating losses', measuremem=False)
mon_avg = monitor('averaging losses', measuremem=False)
dstore = datastore.read(param['hdf5path'])
with monitor('getting assets'):
assets_df = dstore.read_df('assetcol/array', 'ordinal')
with monitor('getting crmodel'):
crmodel = monitor.read('crmodel')
weights = dstore['weights'][()]
acc = dict(events_per_sid=0)
alt = copy.copy(param['alt']) # avoid issues with OQ_DISTRIBUTE=no
alt_dt = param['oqparam'].alt_dt()
tempname = param['tempname']
aggby = param['aggregate_by']
haz_by_sid = general.group_array(gmfs, 'sid')
losses_by_A = numpy.zeros((len(assets_df), len(alt.loss_names)), F32)
acc['avg_gmf'] = avg_gmf = {}
for col in gmfs.dtype.names:
if col not in 'sid eid rlz':
avg_gmf[col] = numpy.zeros(param['N'], F32)
for sid, asset_df in assets_df.groupby('site_id'):
try:
haz = haz_by_sid[sid]
except KeyError: # no hazard here
continue
with mon_risk:
assets = asset_df.to_records() # fast
acc['events_per_sid'] += len(haz)
assets_by_taxo = get_assets_by_taxo(assets, tempname) # fast
out = get_output(crmodel, assets_by_taxo, haz) # slow
with mon_agg:
alt.aggregate(out, param['minimum_asset_loss'], aggby)
# NB: after the aggregation out contains losses, not loss_ratios
ws = weights[haz['rlz']]
for col in gmfs.dtype.names:
if col not in 'sid eid rlz':
avg_gmf[col][sid] = haz[col] @ ws
if param['avg_losses']:
with mon_avg:
for lni, ln in enumerate(alt.loss_names):
losses_by_A[assets['ordinal'], lni] += out[ln] @ ws
if len(gmfs):
acc['events_per_sid'] /= len(gmfs)
out = []
for eid, arr in alt.items():
for k, vals in enumerate(arr): # arr has shape K, L'
if vals.sum() > 0:
# in the demo there are 264/1694 nonzero events, i.e.
# vals.sum() is zero most of the time
out.append((eid, k) + tuple(vals))
acc['alt'] = numpy.array(out, alt_dt)
if param['avg_losses']:
acc['losses_by_A'] = losses_by_A * param['ses_ratio']
return acc
def calc_risk(gmfs, param, monitor):
"""
:param gmfs: an array of GMFs with fields sid, eid, gmv
:param param: a dictionary of parameters coming from the job.ini
:param monitor: a Monitor instance
:returns: a dictionary of arrays with keys elt, alt, losses_by_A, ...
"""
mon_risk = monitor('computing risk', measuremem=False)
mon_agg = monitor('aggregating losses', measuremem=False)
eids = numpy.unique(gmfs['eid'])
dstore = datastore.read(param['hdf5path'])
with monitor('getting assets'):
assets_df = dstore.read_df('assetcol/array', 'ordinal')
exposed_values = dstore['exposed_values/agg'][()]
with monitor('getting crmodel'):
crmodel = riskmodels.CompositeRiskModel.read(dstore)
events = dstore['events'][list(eids)]
weights = dstore['weights'][()]
E = len(eids)
L = len(param['lba'].loss_names)
elt_dt = [('event_id', U32), ('rlzi', U16), ('loss', (F32, (L,)))]
alt = general.AccumDict(accum=general.AccumDict(accum=numpy.zeros(L, F32)))
# aggkey -> eid -> loss
arr = numpy.zeros((E, L), F32)
acc = dict(events_per_sid=0, numlosses=numpy.zeros(2, int)) # (kept, tot)
lba = param['lba']
tempname = param['tempname']
eid2rlz = dict(events[['id', 'rlz_id']])
eid2idx = {eid: idx for idx, eid in enumerate(eids)}
aggby = param['aggregate_by']
minimum_loss = []
fraction = param['minimum_loss_fraction'] / len(assets_df)
for lt, lti in crmodel.lti.items():
val = exposed_values[lti] * fraction
minimum_loss.append(val)
if lt in lba.policy_dict: # same order as in lba.compute
minimum_loss.append(val)
haz_by_sid = general.group_array(gmfs, 'sid')
for sid, asset_df in assets_df.groupby('site_id'):
try:
haz = haz_by_sid[sid]
except KeyError: # no hazard here
continue
assets = asset_df.to_records()
with mon_risk:
acc['events_per_sid'] += len(haz)
if param['avg_losses']:
ws = weights[[eid2rlz[eid] for eid in haz['eid']]]
assets_by_taxo = get_assets_by_taxo(assets, tempname) # fast
eidx = numpy.array([eid2idx[eid] for eid in haz['eid']])
out = get_output(crmodel, assets_by_taxo, haz)
with mon_agg:
for lti, lt in enumerate(crmodel.loss_types):
lratios = out[lt]
if lt == 'occupants':
field = 'occupants_None'
else:
field = 'value-' + lt
if aggby:
tagidxs = assets[aggby]
for a, asset in enumerate(assets):
if aggby:
idx = ','.join(map(str, tagidxs[a]))
aid = asset['ordinal']
ls = asset[field] * lratios[a]
for loss_idx, losses in lba.compute(asset, ls, lt):
kept = 0
if aggby:
for loss, eid in zip(losses, out.eids):
if loss >= minimum_loss[loss_idx]:
alt[idx][eid][loss_idx] += loss
kept += 1
arr[eidx, loss_idx] += losses
if param['avg_losses']: # this is really fast
lba.losses_by_A[aid, loss_idx] += losses @ ws
acc['numlosses'] += numpy.array([kept, len(losses)])
if len(gmfs):
acc['events_per_sid'] /= len(gmfs)
acc['elt'] = numpy.fromiter( # this is ultra-fast
((event['id'], event['rlz_id'], losses)
for event, losses in zip(events, arr) if losses.sum()), elt_dt)
acc['alt'] = {idx: numpy.fromiter( # already sorted by aid, ultra-fast
((eid, eid2rlz[eid], loss) for eid, loss in alt[idx].items()),
elt_dt) for idx in alt}
if param['avg_losses']:
acc['losses_by_A'] = param['lba'].losses_by_A * param['ses_ratio']
# without resetting the cache the sequential avg_losses would be wrong!
del param['lba'].__dict__['losses_by_A']
return acc
