# -*- coding: utf-8 -*-
# vim: tabstop=4 shiftwidth=4 softtabstop=4
#
# Copyright (C) 2014-2019 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/>.
import itertools
import collections
import logging
import numpy
from openquake.baselib.python3compat import encode
from openquake.baselib import hdf5
from openquake.baselib.general import group_array, deprecated
from openquake.hazardlib import nrml
from openquake.hazardlib.stats import compute_stats2
from openquake.risklib import scientific
from openquake.calculators.extract import (
extract, build_damage_dt, build_damage_array)
from openquake.calculators.export import export, loss_curves
from openquake.calculators.export.hazard import savez, get_mesh
from openquake.calculators import getters
from openquake.commonlib import writers, hazard_writers
from openquake.commonlib.util import get_assets, compose_arrays
Output = collections.namedtuple('Output', 'ltype path array')
F32 = numpy.float32
F64 = numpy.float64
U16 = numpy.uint16
U32 = numpy.uint32
U64 = numpy.uint64
TWO32 = 2 ** 32
stat_dt = numpy.dtype([('mean', F32), ('stddev', F32)])
[docs]def add_quotes(values):
# used to escape tags in CSV files
return numpy.array([encode('"%s"' % val) for val in values], (bytes, 100))
[docs]def get_rup_data(ebruptures):
dic = {}
for ebr in ebruptures:
point = ebr.rupture.surface.get_middle_point()
dic[ebr.serial] = (ebr.rupture.mag, point.x, point.y, point.z)
return dic
# ############################### exporters ############################## #
# this is used by event_based_risk and ebrisk
[docs]@export.add(('agg_curves-rlzs', 'csv'), ('agg_curves-stats', 'csv'))
def export_agg_curve_rlzs(ekey, dstore):
oq = dstore['oqparam']
R = len(dstore['weights'])
agg_curve = dstore[ekey[0]]
tags = (['rlz-%03d' % r for r in range(R)] if ekey[0].endswith('-rlzs')
else ['mean'] + ['quantile-%s' % q for q in oq.quantiles])
periods = agg_curve.attrs['return_periods']
loss_types = tuple(agg_curve.attrs['loss_types'].split())
if any(lt.endswith('_ins') for lt in loss_types):
L = len(loss_types) // 2
else:
L = len(loss_types)
tagnames = tuple(dstore['oqparam'].aggregate_by)
tagcol = dstore['assetcol/tagcol']
writer = writers.CsvWriter(fmt=writers.FIVEDIGITS)
header = ('annual_frequency_of_exceedence', 'return_period',
'loss_type') + tagnames + ('loss_value', 'loss_ratio')
expvalue = dstore['exposed_value'].value # shape (T1, T2, ..., L)
for r, tag in enumerate(tags):
rows = []
for multi_idx, loss in numpy.ndenumerate(agg_curve[:, r]):
p, l, *tagidxs = multi_idx
evalue = expvalue[tuple(tagidxs) + (l % L,)]
row = tagcol.get_tagvalues(tagnames, tagidxs) + (
loss, loss / evalue)
rows.append((1 / periods[p], periods[p], loss_types[l]) + row)
dest = dstore.build_fname('agg_loss', tag, 'csv')
writer.save(rows, dest, header)
return writer.getsaved()
def _get_data(dstore, dskey, stats):
name, kind = dskey.split('-') # i.e. ('avg_losses', 'stats')
if kind == 'stats':
weights = dstore['weights'][dstore['weights'].dtype.names[0]]
tags, stats = zip(*stats)
if dskey in set(dstore): # precomputed
value = dstore[dskey].value # shape (A, S, LI)
else: # computed on the fly
value = compute_stats2(
dstore[name + '-rlzs'].value, stats, weights)
else: # rlzs
value = dstore[dskey].value # shape (A, R, LI)
R = value.shape[1]
tags = ['rlz-%03d' % r for r in range(R)]
return name, value, tags
# used by ebrisk
[docs]@export.add(('agg_maps-rlzs', 'csv'), ('agg_maps-stats', 'csv'))
def export_agg_maps_csv(ekey, dstore):
name, kind = ekey[0].split('-')
oq = dstore['oqparam']
tagcol = dstore['assetcol/tagcol']
agg_maps = dstore[ekey[0]].value # shape (C, R, L, T...)
R = agg_maps.shape[1]
kinds = (['rlz-%03d' % r for r in range(R)] if ekey[0].endswith('-rlzs')
else ['mean'] + ['quantile-%s' % q for q in oq.quantiles])
clp = [str(p) for p in oq.conditional_loss_poes]
dic = dict(tagnames=['clp', 'kind', 'loss_type'] + oq.aggregate_by,
clp=['?'] + clp, kind=['?'] + kinds,
loss_type=('?',) + oq.loss_dt().names)
for tagname in oq.aggregate_by:
dic[tagname] = getattr(tagcol, tagname)
aw = hdf5.ArrayWrapper(agg_maps, dic)
writer = writers.CsvWriter(fmt=writers.FIVEDIGITS)
fname = dstore.export_path('%s.%s' % ekey)
writer.save(aw.to_table(), fname)
return [fname]
# this is used by event_based_risk and classical_risk
[docs]@export.add(('avg_losses-rlzs', 'csv'), ('avg_losses-stats', 'csv'))
def export_avg_losses(ekey, dstore):
"""
:param ekey: export key, i.e. a pair (datastore key, fmt)
:param dstore: datastore object
"""
dskey = ekey[0]
oq = dstore['oqparam']
dt = oq.loss_dt()
name, value, tags = _get_data(dstore, dskey, oq.hazard_stats().items())
writer = writers.CsvWriter(fmt=writers.FIVEDIGITS)
assets = get_assets(dstore)
for tag, values in zip(tags, value.transpose(1, 0, 2)):
dest = dstore.build_fname(name, tag, 'csv')
array = numpy.zeros(len(values), dt)
for l, lt in enumerate(dt.names):
array[lt] = values[:, l]
writer.save(compose_arrays(assets, array), dest)
return writer.getsaved()
# this is used by ebrisk
[docs]@export.add(('agg_losses-rlzs', 'csv'), ('agg_losses-stats', 'csv'))
def export_agg_losses(ekey, dstore):
"""
:param ekey: export key, i.e. a pair (datastore key, fmt)
:param dstore: datastore object
"""
dskey = ekey[0]
oq = dstore['oqparam']
dt = oq.loss_dt()
name, value, tags = _get_data(dstore, dskey, oq.hazard_stats().items())
writer = writers.CsvWriter(fmt=writers.FIVEDIGITS)
expvalue = dstore['exposed_value'].value # shape (T1, T2, ..., L)
tagcol = dstore['assetcol/tagcol']
tagnames = tuple(dstore['oqparam'].aggregate_by)
header = ('loss_type',) + tagnames + (
'loss_value', 'exposed_value', 'loss_ratio')
for r, tag in enumerate(tags):
rows = []
for multi_idx, loss in numpy.ndenumerate(value[:, r]):
l, *tagidxs = multi_idx
evalue = expvalue[tuple(tagidxs) + (l,)]
row = tagcol.get_tagvalues(tagnames, tagidxs) + (
loss, evalue, loss / evalue)
rows.append((dt.names[l],) + row)
dest = dstore.build_fname(name, tag, 'csv')
writer.save(rows, dest, header)
return writer.getsaved()
# this is used by ebrisk
[docs]@export.add(('avg_losses', 'csv'))
def export_avg_losses_ebrisk(ekey, dstore):
"""
:param ekey: export key, i.e. a pair (datastore key, fmt)
:param dstore: datastore object
"""
name = ekey[0]
oq = dstore['oqparam']
dt = oq.loss_dt()
value = dstore[name].value # shape (A, L)
writer = writers.CsvWriter(fmt=writers.FIVEDIGITS)
assets = get_assets(dstore)
dest = dstore.build_fname(name, 'mean', 'csv')
array = numpy.zeros(len(value), dt)
for l, lt in enumerate(dt.names):
array[lt] = value[:, l]
writer.save(compose_arrays(assets, array), dest)
return writer.getsaved()
# this is used by scenario_risk
[docs]@export.add(('losses_by_asset', 'csv'))
def export_losses_by_asset(ekey, dstore):
"""
:param ekey: export key, i.e. a pair (datastore key, fmt)
:param dstore: datastore object
"""
loss_dt = dstore['oqparam'].loss_dt(stat_dt)
losses_by_asset = dstore[ekey[0]].value
rlzs = dstore['csm_info'].get_rlzs_assoc().realizations
assets = get_assets(dstore)
writer = writers.CsvWriter(fmt=writers.FIVEDIGITS)
for rlz in rlzs:
losses = losses_by_asset[:, rlz.ordinal]
dest = dstore.build_fname('losses_by_asset', rlz, 'csv')
data = compose_arrays(assets, losses.copy().view(loss_dt)[:, 0])
writer.save(data, dest)
return writer.getsaved()
# this is used by scenario_risk, event_based_risk and ebrisk
[docs]@export.add(('losses_by_event', 'csv'))
def export_losses_by_event(ekey, dstore):
"""
:param ekey: export key, i.e. a pair (datastore key, fmt)
:param dstore: datastore object
"""
oq = dstore['oqparam']
writer = writers.CsvWriter(fmt=writers.FIVEDIGITS)
dest = dstore.build_fname('losses_by_event', '', 'csv')
if oq.calculation_mode.startswith('scenario'):
dtlist = [('eid', U64)] + oq.loss_dt_list()
arr = dstore['losses_by_event'].value[['eid', 'loss']]
writer.save(arr.copy().view(dtlist), dest)
else:
dtlist = [('event_id', U64), ('rup_id', U32), ('year', U32)] + \
oq.loss_dt_list()
eids = dstore['losses_by_event']['eid']
year_of = year_dict(dstore['events']['eid'],
oq.investigation_time, oq.ses_seed)
arr = numpy.zeros(len(dstore['losses_by_event']), dtlist)
arr['event_id'] = eids
arr['rup_id'] = arr['event_id'] / TWO32
arr['year'] = [year_of[eid] for eid in eids]
loss = dstore['losses_by_event']['loss'].T # shape (L, E)
for losses, loss_type in zip(loss, oq.loss_dt().names):
arr[loss_type] = losses
writer.save(arr, dest)
return writer.getsaved()
[docs]@export.add(('losses_by_asset', 'npz'))
def export_losses_by_asset_npz(ekey, dstore):
"""
:param ekey: export key, i.e. a pair (datastore key, fmt)
:param dstore: datastore object
"""
fname = dstore.export_path('%s.%s' % ekey)
savez(fname, **dict(extract(dstore, 'losses_by_asset')))
return [fname]
def _compact(array):
# convert an array of shape (a, e) into an array of shape (a,)
dt = array.dtype
a, e = array.shape
lst = []
for name in dt.names:
lst.append((name, (dt[name], e)))
return array.view(numpy.dtype(lst)).reshape(a)
[docs]@export.add(('rup_loss_table', 'xml'))
def export_maxloss_ruptures(ekey, dstore):
"""
:param ekey: export key, i.e. a pair (datastore key, fmt)
:param dstore: datastore object
"""
oq = dstore['oqparam']
mesh = get_mesh(dstore['sitecol'])
rlzs_by_gsim = dstore['csm_info'].get_rlzs_by_gsim_grp()
num_ses = oq.ses_per_logic_tree_path
fnames = []
for loss_type in oq.loss_dt().names:
ebr = getters.get_maxloss_rupture(dstore, loss_type)
root = hazard_writers.rupture_to_element(
ebr.export(mesh, rlzs_by_gsim[ebr.grp_id], num_ses))
dest = dstore.export_path('rupture-%s.xml' % loss_type)
with open(dest, 'wb') as fh:
nrml.write(list(root), fh)
fnames.append(dest)
return fnames
[docs]def year_dict(eids, investigation_time, ses_seed):
numpy.random.seed(ses_seed)
years = numpy.random.choice(int(investigation_time), len(eids)) + 1
return dict(zip(numpy.sort(eids), years)) # eid -> year
# this is used by event_based_risk
[docs]@export.add(('agg_loss_table', 'csv'))
@deprecated(msg='This exporter will be removed soon')
def export_agg_losses_ebr(ekey, dstore):
"""
:param ekey: export key, i.e. a pair (datastore key, fmt)
:param dstore: datastore object
"""
if 'ruptures' not in dstore:
logging.warning('There are no ruptures in the datastore')
return []
name, ext = export.keyfunc(ekey)
agg_losses = dstore['losses_by_event']
has_rup_data = 'ruptures' in dstore
extra_list = [('magnitude', F32),
('centroid_lon', F32),
('centroid_lat', F32),
('centroid_depth', F32)] if has_rup_data else []
oq = dstore['oqparam']
lti = oq.lti
dtlist = ([('event_id', U64), ('rup_id', U32), ('year', U32)]
+ extra_list + oq.loss_dt_list())
elt_dt = numpy.dtype(dtlist)
elt = numpy.zeros(len(agg_losses), elt_dt)
writer = writers.CsvWriter(fmt=writers.FIVEDIGITS)
events = dstore['events'].value
events_by_rupid = collections.defaultdict(list)
for event in events:
rupid = event['eid'] // TWO32
events_by_rupid[rupid].append(event)
year_of = year_dict(events['eid'], oq.investigation_time, oq.ses_seed)
rup_data = {}
event_by_eid = {} # eid -> event
# populate rup_data and event_by_eid
# TODO: avoid reading the events twice
for rgetter in getters.gen_rupture_getters(dstore):
ruptures = rgetter.get_ruptures()
for ebr in ruptures:
for event in events_by_rupid[ebr.serial]:
event_by_eid[event['eid']] = event
if has_rup_data:
rup_data.update(get_rup_data(ruptures))
for r, row in enumerate(agg_losses):
rec = elt[r]
event = event_by_eid[row['eid']]
rec['event_id'] = eid = event['eid']
rec['year'] = year_of[eid]
if rup_data:
rec['rup_id'] = rup_id = event['eid'] // TWO32
(rec['magnitude'], rec['centroid_lon'], rec['centroid_lat'],
rec['centroid_depth']) = rup_data[rup_id]
for lt, i in lti.items():
rec[lt] = row['loss'][i]
elt.sort(order=['year', 'event_id'])
dest = dstore.build_fname('elt', '', 'csv')
writer.save(elt, dest)
return writer.getsaved()
# this is used by classical_risk and event_based_risk
[docs]@export.add(('loss_curves-rlzs', 'csv'), ('loss_curves-stats', 'csv'),
('loss_curves', 'csv'))
def export_loss_curves(ekey, dstore):
if '/' in ekey[0]:
kind = ekey[0].split('/', 1)[1]
else:
kind = ekey[0].split('-', 1)[1] # rlzs or stats
return loss_curves.LossCurveExporter(dstore).export('csv', kind)
# used by classical_risk and event_based_risk
[docs]@export.add(('loss_maps-rlzs', 'csv'), ('loss_maps-stats', 'csv'))
def export_loss_maps_csv(ekey, dstore):
kind = ekey[0].split('-')[1] # rlzs or stats
assets = get_assets(dstore)
value = get_loss_maps(dstore, kind)
if kind == 'rlzs':
tags = dstore['csm_info'].get_rlzs_assoc().realizations
else:
oq = dstore['oqparam']
tags = ['mean'] + ['quantile-%s' % q for q in oq.quantiles]
writer = writers.CsvWriter(fmt=writers.FIVEDIGITS)
for i, tag in enumerate(tags):
fname = dstore.build_fname('loss_maps', tag, ekey[1])
writer.save(compose_arrays(assets, value[:, i]), fname)
return writer.getsaved()
# used by classical_risk and event_based_risk
[docs]@export.add(('loss_maps-rlzs', 'npz'), ('loss_maps-stats', 'npz'))
def export_loss_maps_npz(ekey, dstore):
kind = ekey[0].split('-')[1] # rlzs or stats
assets = get_assets(dstore)
value = get_loss_maps(dstore, kind)
R = dstore['csm_info'].get_num_rlzs()
if kind == 'rlzs':
tags = ['rlz-%03d' % r for r in range(R)]
else:
oq = dstore['oqparam']
tags = ['mean'] + ['quantile-%s' % q for q in oq.quantiles]
fname = dstore.export_path('%s.%s' % ekey)
dic = {}
for i, tag in enumerate(tags):
dic[tag] = compose_arrays(assets, value[:, i])
savez(fname, **dic)
return [fname]
[docs]@export.add(('damages-rlzs', 'csv'), ('damages-stats', 'csv'))
def export_damages_csv(ekey, dstore):
rlzs = dstore['csm_info'].get_rlzs_assoc().realizations
oq = dstore['oqparam']
loss_types = oq.loss_dt().names
assets = get_assets(dstore)
value = dstore[ekey[0]].value # matrix N x R x LI or T x R x LI
writer = writers.CsvWriter(fmt=writers.FIVEDIGITS)
if ekey[0].endswith('stats'):
tags = ['mean'] + ['quantile-%s' % q for q in oq.quantiles]
else:
tags = ['rlz-%03d' % r for r in range(len(rlzs))]
for lti, lt in enumerate(loss_types):
for tag, values in zip(tags, value[:, :, lti].T):
fname = dstore.build_fname('damages-%s' % lt, tag, ekey[1])
writer.save(compose_arrays(assets, values), fname)
return writer.getsaved()
[docs]@export.add(('dmg_by_asset', 'csv'))
def export_dmg_by_asset_csv(ekey, dstore):
damage_dt = build_damage_dt(dstore)
rlzs = dstore['csm_info'].get_rlzs_assoc().realizations
data = dstore[ekey[0]]
writer = writers.CsvWriter(fmt='%.6E')
assets = get_assets(dstore)
for rlz in rlzs:
dmg_by_asset = build_damage_array(data[:, rlz.ordinal], damage_dt)
fname = dstore.build_fname(ekey[0], rlz, ekey[1])
writer.save(compose_arrays(assets, dmg_by_asset), fname)
return writer.getsaved()
[docs]@export.add(('dmg_by_asset', 'npz'))
def export_dmg_by_asset_npz(ekey, dstore):
fname = dstore.export_path('%s.%s' % ekey)
savez(fname, **dict(extract(dstore, 'dmg_by_asset')))
return [fname]
[docs]@export.add(('dmg_by_event', 'csv'))
def export_dmg_by_event(ekey, dstore):
"""
:param ekey: export key, i.e. a pair (datastore key, fmt)
:param dstore: datastore object
"""
damage_dt = build_damage_dt(dstore, mean_std=False)
dt_list = [('event_id', numpy.uint64), ('rlzi', numpy.uint16)] + [
(f, damage_dt.fields[f][0]) for f in damage_dt.names]
all_losses = dstore[ekey[0]].value # shape (E, R, LI)
events_by_rlz = group_array(dstore['events'], 'rlz')
rlzs = dstore['csm_info'].get_rlzs_assoc().realizations
writer = writers.CsvWriter(fmt=writers.FIVEDIGITS)
fname = dstore.build_fname('dmg_by_event', '', 'csv')
writer.save(numpy.zeros(0, dt_list), fname)
with open(fname, 'ab') as dest:
for rlz in rlzs:
data = all_losses[:, rlz.ordinal].copy().view(damage_dt) # shape E
arr = numpy.zeros(len(data), dt_list)
arr['event_id'] = events_by_rlz[rlz.ordinal]['eid']
arr['rlzi'] = rlz.ordinal
for field in damage_dt.names:
arr[field] = data[field].squeeze()
writer.save_block(arr, dest)
return [fname]
# emulate a Django point
[docs]class Location(object):
def __init__(self, x, y):
self.x, self.y = x, y
self.wkt = 'POINT(%s %s)' % (x, y)
[docs]def indices(*sizes):
return itertools.product(*map(range, sizes))
def _to_loss_maps(array, loss_maps_dt):
# convert a 4D array into a 2D array of dtype loss_maps_dt
A, R, C, LI = array.shape
lm = numpy.zeros((A, R), loss_maps_dt)
for li, name in enumerate(loss_maps_dt.names):
for p, poe in enumerate(loss_maps_dt[name].names):
lm[name][poe] = array[:, :, p, li]
return lm
[docs]def get_loss_maps(dstore, kind):
"""
:param dstore: a DataStore instance
:param kind: 'rlzs' or 'stats'
"""
oq = dstore['oqparam']
name = 'loss_maps-%s' % kind
if name in dstore: # event_based risk
return _to_loss_maps(dstore[name].value, oq.loss_maps_dt())
name = 'loss_curves-%s' % kind
if name in dstore: # classical_risk
# the loss maps are built on the fly from the loss curves
loss_curves = dstore[name]
loss_maps = scientific.broadcast(
scientific.loss_maps, loss_curves, oq.conditional_loss_poes)
return loss_maps
raise KeyError('loss_maps/loss_curves missing in %s' % dstore)
agg_dt = numpy.dtype([('unit', (bytes, 6)), ('mean', F32), ('stddev', F32)])
# this is used by scenario_risk
[docs]@export.add(('agglosses', 'csv'))
def export_agglosses(ekey, dstore):
oq = dstore['oqparam']
loss_dt = oq.loss_dt()
cc = dstore['assetcol/cost_calculator']
unit_by_lt = cc.units
unit_by_lt['occupants'] = 'people'
agglosses = dstore[ekey[0]]
losses = []
header = ['loss_type', 'unit', 'mean', 'stddev']
for l, lt in enumerate(loss_dt.names):
unit = unit_by_lt[lt.replace('_ins', '')]
mean = agglosses[l]['mean']
stddev = agglosses[l]['stddev']
losses.append((lt, unit, mean, stddev))
dest = dstore.build_fname('agglosses', '', 'csv')
writers.write_csv(dest, losses, header=header)
return [dest]
AggCurve = collections.namedtuple(
'AggCurve', ['losses', 'poes', 'average_loss', 'stddev_loss'])
[docs]def get_paths(rlz):
"""
:param rlz:
a logic tree realization (composite or simple)
:returns:
a dict {'source_model_tree_path': string, 'gsim_tree_path': string}
"""
dic = {}
if hasattr(rlz, 'sm_lt_path'): # composite realization
dic['source_model_tree_path'] = '_'.join(rlz.sm_lt_path)
dic['gsim_tree_path'] = '_'.join(rlz.gsim_lt_path)
else: # simple GSIM realization
dic['source_model_tree_path'] = ''
dic['gsim_tree_path'] = '_'.join(rlz.lt_path)
return dic
[docs]@export.add(('bcr-rlzs', 'csv'), ('bcr-stats', 'csv'))
def export_bcr_map(ekey, dstore):
oq = dstore['oqparam']
assets = get_assets(dstore)
bcr_data = dstore[ekey[0]]
N, R = bcr_data.shape
if ekey[0].endswith('stats'):
tags = ['mean'] + ['quantile-%s' % q for q in oq.quantiles]
else:
tags = ['rlz-%03d' % r for r in range(R)]
fnames = []
writer = writers.CsvWriter(fmt=writers.FIVEDIGITS)
for t, tag in enumerate(tags):
path = dstore.build_fname('bcr', tag, 'csv')
writer.save(compose_arrays(assets, bcr_data[:, t]), path)
fnames.append(path)
return writer.getsaved()
[docs]@export.add(('losses_by_tag', 'csv'), ('curves_by_tag', 'csv'))
@deprecated(msg='This exporter will be removed soon')
def export_by_tag_csv(ekey, dstore):
"""
:param ekey: export key, i.e. a pair (datastore key, fmt)
:param dstore: datastore object
"""
token, tag = ekey[0].split('/')
data = extract(dstore, token + '/' + tag)
fnames = []
writer = writers.CsvWriter(fmt=writers.FIVEDIGITS)
for stat, arr in data:
tup = (ekey[0].replace('/', '-'), stat, ekey[1])
path = '%s-%s.%s' % tup
fname = dstore.export_path(path)
writer.save(arr, fname)
fnames.append(fname)
return fnames
[docs]@export.add(('aggregate_by', 'csv'))
def export_aggregate_by_csv(ekey, dstore):
"""
:param ekey: export key, i.e. a pair (datastore key, fmt)
:param dstore: datastore object
"""
token, what = ekey[0].split('/', 1)
data = extract(dstore, token + '/' + what)
fnames = []
writer = writers.CsvWriter(fmt=writers.FIVEDIGITS)
for stat, aw in data:
tup = (ekey[0].replace('/', '-').replace('-stats', ''), stat, ekey[1])
path = '%s-%s.%s' % tup
fname = dstore.export_path(path)
writer.save(aw.to_table(), fname)
fnames.append(fname)
return fnames