def get_inventory(segment, session=None, **kwargs):
"""raises tons of exceptions (see main). FIXME: write doc
:param session: if **not** None but a valid sqlalchemy session object, then
the inventory, if downloaded because not present, will be saveed to the db (compressed)
"""
data = segment.channel.station.inventory_xml
if not data:
query_url = get_inventory_query(segment.channel.station)
data = url_read(query_url, **kwargs)
if session and data:
segment.channel.station.inventory_xml = dumps_inv(data)
session.commit()
elif not data:
raise ValueError("No data from server")
return loads_inv(data)
def process_all(session, segments_model_instances, run_id,
notify_progress_func=lambda *a, **v: None, **processing_args):
"""
Processes all segments_model_instances. FIXME: write detailed doc
"""
# redirect stndard error to devnull. FIXME if we can capture it segment-wise (that
# would be great but.. how much effort and how much performances decreasing?)
# redirect_external_out(2)
# set after how many processed segments we want to commit. Setting it higher might speed up
# calculations at expense of loosing max_session_new segment if just one is wrong
max_session_new = 10
# commit for safety:
commit(session, on_exc=lambda exc: logger.error(str(exc)))
calculated = 0
saved = 0
logger.info("Processing %d segments", len(segments_model_instances))
ret = []
sta2segs = defdict(lambda: [])
for seg in segments_model_instances:
sta2segs[seg.channel.station_id].append(seg)
# process segments station-like, so that we load only one inventory at a time
# and hopefully it will garbage collected (inventory object is big)
for sta_id, segments in sta2segs.iteritems():
inventory = None
try:
inventory = get_inventory(segments[0], session, timeout=30)
except SQLAlchemyError as exc:
logger.warning("Error while saving inventory (station id=%s), "
"%d segment will not be processed: %s",
str(sta_id), len(segments), str(exc))
session.rollback()
except (urllib2.HTTPError, urllib2.URLError, httplib.HTTPException, socket.error) as _:
logger.warning("Error while downloading inventory (station id=%s), "
"%d segment will not be processed: %s URL: %s",
str(sta_id), len(segments), str(_), get_inventory_query(segments[0]))
except Exception as exc: # pylint:disable=broad-except
logger.warning("Error while creating inventory (station id=%s), "
"%d segment will not be processed: %s",
str(sta_id), len(segments), str(exc))
if inventory is None:
notify_progress_func(len(segments))
continue
# pass
# THIS IS THE METHOD WITHOUT MULTIPROCESS: 28, 24.7 secs on 30 segments
for seg in segments:
notify_progress_func(1)
pro = models.Processing(run_id=run_id)
# pro.segment = seg
# session.flush()
try:
pro = process(pro, seg, seg.channel, seg.channel.station, seg.event,
seg.datacenter, inventory, **processing_args)
pro.id = None
pro.segment = seg
calculated += 1
ret.append(pro)
# flush(session, on_exc=lambda exc: logger.error(str(exc)))
if len(ret) >= max_session_new:
added = len(ret)
session.add_all(ret)
ret = []
if commit(session,
on_exc=lambda exc: logger.warning(msgs.db.dropped_seg(added,
None,
exc))):
saved += added
except Exception as exc: # pylint:disable=broad-except
logger.warning(msgs.calc.dropped_seg(seg, "segments processing", exc))
added = len(ret)
if added and commit(session, on_exc=lambda exc: logger.warning(msgs.db.dropped_seg(added,
None,
exc))):
saved += added
logger.info("")
logger.info("%d segments successfully processed, %d succesfully saved", calculated, saved)
return ret
