def testRangeStepSansDirection(self):
"give step and no directon"
df = Datefile(datetime.strptime('2009-10-31', '%Y-%m-%d'),
persistence_file=self.persistence_file,
step=30)
start, end = df.get_date_range()
self.assertEquals(start.strftime('%Y-%m-%d'), '2009-10-02')
self.assertEquals(end.strftime('%Y-%m-%d'), '2009-10-31')
def testRangeSansFile(self):
"range w/o step"
df = Datefile(datetime.strptime('2009-10-31', '%Y-%m-%d'))
start, end = df.get_date_range()
self.assertEquals(start.strftime('%Y-%m-%d'),
'2009-10-31')
self.assertEquals(end.strftime('%Y-%m-%d'),
'2009-10-31')
def testRangeSansStep(self):
"range w/o step"
df = Datefile(datetime.strptime('2009-10-31', '%Y-%m-%d'),
persistence_file=self.persistence_file,
direction='forwards')
start, end = df.get_date_range()
self.assertEquals(start.strftime('%Y-%m-%d'), '2009-10-31')
self.assertEquals(end.strftime('%Y-%m-%d'), '2009-10-31')
def testStep(self):
df = Datefile(datetime.strptime('2009-01-01', '%Y-%m-%d'),
persistence_file=self.persistence_file,
direction='backwards', step=10)
self.assertEquals(df.get_date().strftime('%Y-%m-%d'),
'2009-01-01')
df.bump_date()
self.assertEquals(df.get_date().strftime('%Y-%m-%d'),
'2008-12-22')
def testStepDatesFwd(self):
"steping forward, end date should be step-1 days after start"
df = Datefile(datetime.strptime('2009-02-01', '%Y-%m-%d'),
persistence_file=self.persistence_file,
direction='forwards',
step=10)
start, end = df.get_date_range()
self.assertEquals(start.strftime('%Y-%m-%d'), '2009-02-01')
self.assertEquals(end.strftime('%Y-%m-%d'), '2009-02-10')
def testForwards(self):
df = Datefile(datetime.strptime('2009-01-01', '%Y-%m-%d'),
persistence_file=self.persistence_file,
direction='forwards')
self.assertEquals(df.get_date().strftime('%Y-%m-%d'),
'2009-01-01')
df.bump_date()
self.assertEquals(df.get_date().strftime('%Y-%m-%d'),
'2009-01-02')
def testStepDatesBack(self):
"steping backward, end date should be step-1 days before start"
df = Datefile(datetime.strptime('2009-01-31', '%Y-%m-%d'),
persistence_file=self.persistence_file,
direction='backwards',
step=10)
start, end = df.get_date_range()
self.assertEquals(start.strftime('%Y-%m-%d'), '2009-01-22')
self.assertEquals(end.strftime('%Y-%m-%d'), '2009-01-31')
def testRangeSansStep(self):
"range w/o step"
df = Datefile(datetime.strptime('2009-10-31', '%Y-%m-%d'),
persistence_file=self.persistence_file,
direction='forwards')
start, end = df.get_date_range()
self.assertEquals(start.strftime('%Y-%m-%d'),
'2009-10-31')
self.assertEquals(end.strftime('%Y-%m-%d'),
'2009-10-31')
def testStepDatesFwd(self):
"steping forward, end date should be step-1 days after start"
df = Datefile(datetime.strptime('2009-02-01', '%Y-%m-%d'),
persistence_file=self.persistence_file,
direction='forwards', step=10)
start, end = df.get_date_range()
self.assertEquals(start.strftime('%Y-%m-%d'),
'2009-02-01')
self.assertEquals(end.strftime('%Y-%m-%d'),
'2009-02-10')
def testRangeStepSansDirection(self):
"give step and no directon"
df = Datefile(datetime.strptime('2009-10-31', '%Y-%m-%d'),
persistence_file=self.persistence_file,
step=30)
start, end = df.get_date_range()
self.assertEquals(start.strftime('%Y-%m-%d'),
'2009-10-02')
self.assertEquals(end.strftime('%Y-%m-%d'),
'2009-10-31')
def testStepDatesBack(self):
"steping backward, end date should be step-1 days before start"
df = Datefile(datetime.strptime('2009-01-31', '%Y-%m-%d'),
persistence_file=self.persistence_file,
direction='backwards', step=10)
start, end = df.get_date_range()
self.assertEquals(start.strftime('%Y-%m-%d'),
'2009-01-22')
self.assertEquals(end.strftime('%Y-%m-%d'),
'2009-01-31')
def testInitialDate(self):
# write a date to the file, then call with a different initial
# date. should get the date from the file.
with open(self.persistence_file, 'w') as p_file:
p_file.write('2009-06-06')
df = Datefile(datetime.strptime('2009-01-01', '%Y-%m-%d'),
persistence_file=self.persistence_file,
direction='backwards')
self.assertEquals(df.get_date().strftime('%Y-%m-%d'),
'2009-06-06')
df.bump_date()
self.assertEquals(df.get_date().strftime('%Y-%m-%d'),
'2009-06-05')
def __init__(self, data_warehouse=None, data_mart=None,
reportDate=None, database_user=None,
database_password=None, verbosity=0):
self.data_warehouse = data_warehouse
self.warehouse_port = 5432 # postgres default
self.data_mart = data_mart
self.mart_port = 5432 # postgres default
self.reportDate = reportDate and parseDate(reportDate) or None
self.database_user = database_user
self.database_password = database_password
self.dir, thisFile = os.path.split(__file__)
self.verbosity = verbosity
self.queue = JoinableQueue()
self.datefile = "/tmp/longitudinal_datefile"
self.datePersistence = Datefile(initial_date=self.reportDate)
self.lock = FileLock(LOCKFILE)
self.skip_prep = False
def testNoDirection(self):
df = Datefile(datetime.now())
self.assert_(df.get_date(), "Even w/o direction, should get "
"initial date back")
datebefore = df.get_date()
df.bump_date()
dateafter = df.get_date()
self.assertEquals(datebefore, dateafter)
def processArgs(self):
""" Process any optional arguments and possitional parameters
"""
parser = OptionParser(usage=usage)
parser.add_option("-c", "--countdown", dest="countdown",
default=None,
help="count {down,up} date using date string in "\
"%s - set to 'forwards' or 'backwards' "\
"if desired" % self.datefile)
parser.add_option("-d", "--date", dest="date", default=None,
help="single admission date to dedup "\
"(by default, checks the entire database)")
parser.add_option("-s", "--skip-prep", dest="skip_prep",
default=False, action="store_true",
help="skip the expense of looking for new "\
"messages")
parser.add_option("-v", "--verbose", dest="verbosity",
action="count", default=self.verbosity,
help="increase output verbosity")
parser.add_option("-m", "--mart-port", dest="mart_port",
default=self.mart_port, type="int",
help="alternate port number for data mart")
parser.add_option("-w", "--warehouse-port", dest="warehouse_port",
default=self.warehouse_port, type="int",
help="alternate port number for data warehouse")
(options, args) = parser.parse_args()
if len(args) != 2:
parser.error("incorrect number of arguments")
self.data_warehouse = args[0]
self.data_mart = args[1]
self.warehouse_port = parser.values.warehouse_port
self.mart_port = parser.values.mart_port
self.verbosity = parser.values.verbosity
self.skip_prep = parser.values.skip_prep
initial_date = parser.values.date and \
parseDate(parser.values.date) or None
self.datePersistence = Datefile(initial_date=initial_date,
persistence_file=self.datefile,
direction=parser.values.countdown)
self.reportDate = self.datePersistence.get_date()
def testStepDatesBackThrice(self):
"steping backward three times, duplication avoided?"
df = Datefile(datetime.strptime('2011-04-30', '%Y-%m-%d'),
persistence_file=self.persistence_file,
direction='backwards',
step=10)
start, end = df.get_date_range()
self.assertEquals(start.strftime('%Y-%m-%d'), '2011-04-21')
self.assertEquals(end.strftime('%Y-%m-%d'), '2011-04-30')
df.bump_date()
start, end = df.get_date_range()
self.assertEquals(start.strftime('%Y-%m-%d'), '2011-04-11')
self.assertEquals(end.strftime('%Y-%m-%d'), '2011-04-20')
df.bump_date()
start, end = df.get_date_range()
self.assertEquals(start.strftime('%Y-%m-%d'), '2011-04-01')
self.assertEquals(end.strftime('%Y-%m-%d'), '2011-04-10')
def testBackwards(self):
df = Datefile(datetime.strptime('2009-01-01', '%Y-%m-%d'),
persistence_file=self.persistence_file,
direction='backwards')
self.assertEquals(df.get_date().strftime('%Y-%m-%d'), '2009-01-01')
df.bump_date()
self.assertEquals(df.get_date().strftime('%Y-%m-%d'), '2008-12-31')
def testNoDirection(self):
df = Datefile(datetime.now())
self.assert_(df.get_date(), "Even w/o direction, should get "
"initial date back")
datebefore = df.get_date()
df.bump_date()
dateafter = df.get_date()
self.assertEquals(datebefore, dateafter)
def testInitialDate(self):
# write a date to the file, then call with a different initial
# date. should get the date from the file.
with open(self.persistence_file, 'w') as p_file:
p_file.write('2009-06-06')
df = Datefile(datetime.strptime('2009-01-01', '%Y-%m-%d'),
persistence_file=self.persistence_file,
direction='backwards')
self.assertEquals(df.get_date().strftime('%Y-%m-%d'), '2009-06-06')
df.bump_date()
self.assertEquals(df.get_date().strftime('%Y-%m-%d'), '2009-06-05')
def testStepDatesBackThrice(self):
"steping backward three times, duplication avoided?"
df = Datefile(datetime.strptime('2011-04-30', '%Y-%m-%d'),
persistence_file=self.persistence_file,
direction='backwards', step=10)
start, end = df.get_date_range()
self.assertEquals(start.strftime('%Y-%m-%d'),
'2011-04-21')
self.assertEquals(end.strftime('%Y-%m-%d'),
'2011-04-30')
df.bump_date()
start, end = df.get_date_range()
self.assertEquals(start.strftime('%Y-%m-%d'),
'2011-04-11')
self.assertEquals(end.strftime('%Y-%m-%d'),
'2011-04-20')
df.bump_date()
start, end = df.get_date_range()
self.assertEquals(start.strftime('%Y-%m-%d'),
'2011-04-01')
self.assertEquals(end.strftime('%Y-%m-%d'),
'2011-04-10')
def __init__(
self,
data_warehouse=None,
data_mart=None,
reportDate=None,
database_user=None,
database_password=None,
verbosity=0,
):
self.data_warehouse = data_warehouse
self.warehouse_port = 5432 # postgres default
self.data_mart = data_mart
self.mart_port = 5432 # postgres default
self.reportDate = reportDate and parseDate(reportDate) or None
self.database_user = database_user
self.database_password = database_password
self.dir, thisFile = os.path.split(__file__)
self.verbosity = verbosity
self.queue = JoinableQueue()
self.datefile = "/tmp/longitudinal_datefile"
self.datePersistence = Datefile(initial_date=self.reportDate)
self.lock = FileLock(LOCKFILE)
self.skip_prep = False
class LongitudinalManager(object):
""" Abstraction to handle which db, user, etc. the deduplication
process should be run on. Handles runtime arguments and
execution.
Manages the process by farming out the individual visit
deduplication to a number of worker processes (necessary to take
advatage of multi-core processor and database as the limiting
factor).
"""
# The gating issue is the number of postgres connections that are
# allowed to run concurrently. Setting this to N-1 (where N is
# the number of cores) has proven the fastest and most reliable.
NUM_PROCS = 5
def __init__(
self,
data_warehouse=None,
data_mart=None,
reportDate=None,
database_user=None,
database_password=None,
verbosity=0,
):
self.data_warehouse = data_warehouse
self.warehouse_port = 5432 # postgres default
self.data_mart = data_mart
self.mart_port = 5432 # postgres default
self.reportDate = reportDate and parseDate(reportDate) or None
self.database_user = database_user
self.database_password = database_password
self.dir, thisFile = os.path.split(__file__)
self.verbosity = verbosity
self.queue = JoinableQueue()
self.datefile = "/tmp/longitudinal_datefile"
self.datePersistence = Datefile(initial_date=self.reportDate)
self.lock = FileLock(LOCKFILE)
self.skip_prep = False
def __call__(self):
return self.execute()
def processArgs(self):
""" Process any optional arguments and possitional parameters
"""
parser = OptionParser(usage=usage)
parser.add_option(
"-c",
"--countdown",
dest="countdown",
default=None,
help="count {down,up} date using date string in "
"%s - set to 'forwards' or 'backwards' "
"if desired" % self.datefile,
)
parser.add_option(
"-d",
"--date",
dest="date",
default=None,
help="single admission date to dedup " "(by default, checks the entire database)",
)
parser.add_option(
"-s",
"--skip-prep",
dest="skip_prep",
default=False,
action="store_true",
help="skip the expense of looking for new " "messages",
)
parser.add_option(
"-v",
"--verbose",
dest="verbosity",
action="count",
default=self.verbosity,
help="increase output verbosity",
)
parser.add_option(
"-m",
"--mart-port",
dest="mart_port",
default=self.mart_port,
type="int",
help="alternate port number for data mart",
)
parser.add_option(
"-w",
"--warehouse-port",
dest="warehouse_port",
default=self.warehouse_port,
type="int",
help="alternate port number for data warehouse",
)
(options, args) = parser.parse_args()
if len(args) != 2:
parser.error("incorrect number of arguments")
self.data_warehouse = args[0]
self.data_mart = args[1]
self.warehouse_port = parser.values.warehouse_port
self.mart_port = parser.values.mart_port
self.verbosity = parser.values.verbosity
self.skip_prep = parser.values.skip_prep
initial_date = parser.values.date and parseDate(parser.values.date) or None
self.datePersistence = Datefile(
initial_date=initial_date, persistence_file=self.datefile, direction=parser.values.countdown
)
self.reportDate = self.datePersistence.get_date()
def _prepDeduplicateTables(self):
""" Add any missing rows to the MessageProcessed table
This is the bridge between the data warehouse and the data
mart. In an effort to make the data mart independent of the
warehouse, the processed message data is kept in the mart. As
we're dealing with two distinct databases, there's no
referential integrity available at the database level, so care
should be taken.
"""
startTime = time.time()
logging.info("Starting INSERT INTO internal_message_processed " "at %s", startTime)
# We can take advantage of an "add only" data_warehouse,
# knowing the hl7_msh_id is a sequence moving in the positive
# direction. Simply add any values greater than the previous
# max.
stmt = "SELECT max(hl7_msh_id) from internal_message_processed"
max_id = self.data_mart_access.engine.execute(stmt).first()[0]
if not max_id:
max_id = 0
new_msgs = list()
stmt = (
"""SELECT hl7_msh_id, message_datetime, visit_id
FROM hl7_msh JOIN hl7_visit USING (hl7_msh_id) WHERE
hl7_msh_id > %d """
% max_id
)
rs = self.data_warehouse_access.engine.execute(stmt)
many = 500
while True:
results = rs.fetchmany(many)
if not results:
break
for r in results:
new_msgs.append(MessageProcessed(hl7_msh_id=r[0], message_datetime=r[1], visit_id=r[2]))
self.data_mart_access.session.add_all(new_msgs)
self.data_mart_access.session.commit()
logging.debug("added %d new messages" % len(new_msgs))
new_msgs = list()
logging.info("Added new rows to internal_message_processed in %s", time.time() - startTime)
def _visitsToProcess(self):
""" Look up all distinct visit ids needing attention
Obtain unique list of visit_ids that have messages that
haven't previously been processed. If the user requested just
one days worth (i.e. -d) only that days visits will be
returned.
"""
visit_ids = list()
if not self.reportDate:
logging.info("Launch deduplication for entire database")
# Do the whole batch, that is, all that haven't been
# processed before.
stmt = """SELECT DISTINCT(visit_id) FROM
internal_message_processed
WHERE processed_datetime IS NULL"""
rs = self.data_mart_access.engine.execute(stmt)
many = 10000
while True:
results = rs.fetchmany(many)
if not results:
break
for r in results:
visit_ids.append(r[0])
else:
logging.info("Launch deduplication for %s", self.reportDate)
# Process the requested day only - as we can't join across
# db boundaries - first acquire the full list of visits
# for the requested day from the data_warehouse to use in
# a massive 'in' clause
stmt = """SELECT DISTINCT(visit_id) FROM hl7_visit WHERE
admit_datetime BETWEEN '%s' AND '%s';""" % (
self.reportDate,
self.reportDate + timedelta(days=1),
)
self.access.raw_query(stmt)
rs = self.data_warehouse_access.engine.execute(stmt)
many = 1000
potential_visit_ids = list()
while True:
results = rs.fetchmany(many)
if not results:
break
for r in results:
# tmp_table.insert(r[0])
potential_visit_ids.append(r[0])
if potential_visit_ids:
query = (
self.data_mart_access.session.query(MessageProcessed.visit_id)
.distinct()
.filter(
and_(
MessageProcessed.processed_datetime == None,
MessageProcessed.visit_id.in_(potential_visit_ids),
)
)
)
for r in query:
visit_ids.append(r[0])
logging.info("Found %d visits needing attention", len(visit_ids))
return visit_ids
def tearDown(self):
""" Clean up any open handles/connections """
# now done in execute when we're done with teh connections
def execute(self):
""" Start the process """
# Initialize logging now (verbosity is now set regardless of
# invocation method)
configure_logging(verbosity=self.verbosity, logfile="longitudinal-manager.log")
logging.info("Initiate deduplication for %s", (self.reportDate and self.reportDate or "whole database"))
# Only allow one instance of the manager to run at a time.
if self.lock.is_locked():
logging.warn("Can't continue, %s is locked ", LOCKFILE)
return
if systemUnderLoad():
logging.warn("system under load - continue anyhow")
try:
self.lock.acquire()
self.access = DirectAccess(
database=self.data_warehouse,
port=self.warehouse_port,
user=self.database_user,
password=self.database_password,
)
self.data_warehouse_access = AlchemyAccess(
database=self.data_warehouse,
port=self.warehouse_port,
user=self.database_user,
password=self.database_password,
)
self.data_mart_access = AlchemyAccess(
database=self.data_mart, port=self.mart_port, user=self.database_user, password=self.database_password
)
startTime = time.time()
if not self.skip_prep:
self._prepDeduplicateTables()
visits_to_process = self._visitsToProcess()
# Now done with db access needs at the manager level
# free up resources:
self.data_mart_access.disconnect()
self.data_warehouse_access.disconnect()
self.access.close()
# Set of locks used, one for each table needing protection
# from asynchronous inserts. Names should match table
# minus 'dim_' prefix, plus '_lock' suffix
# i.e. dim_location -> 'location_lock'
table_locks = {
"admission_source_lock": Lock(),
"admission_o2sat_lock": Lock(),
"admission_temp_lock": Lock(),
"assigned_location_lock": Lock(),
"admit_reason_lock": Lock(),
"chief_complaint_lock": Lock(),
"diagnosis_lock": Lock(),
"disposition_lock": Lock(),
"flu_vaccine_lock": Lock(),
"h1n1_vaccine_lock": Lock(),
"lab_flag_lock": Lock(),
"lab_result_lock": Lock(),
"location_lock": Lock(),
"note_lock": Lock(),
"order_number_lock": Lock(),
"performing_lab_lock": Lock(),
"pregnancy_lock": Lock(),
"race_lock": Lock(),
"reference_range_lock": Lock(),
"service_area_lock": Lock(),
"specimen_source_lock": Lock(),
}
# If we have visits to process, fire up the workers...
if len(visits_to_process) > 1:
for i in range(self.NUM_PROCS):
dw = Process(
target=LongitudinalWorker,
kwargs={
"queue": self.queue,
"procNumber": i,
"data_warehouse": self.data_warehouse,
"warehouse_port": self.warehouse_port,
"data_mart": self.data_mart,
"mart_port": self.mart_port,
"dbUser": self.database_user,
"dbPass": self.database_password,
"table_locks": table_locks,
"verbosity": self.verbosity,
},
)
dw.daemon = True
dw.start()
# Populate the queue
for v in visits_to_process:
self.queue.put(v)
# Wait on the queue until empty
self.queue.join()
# Common cleanup
self.tearDown()
self.datePersistence.bump_date()
logging.info("Queue is empty - done in %s", time.time() - startTime)
finally:
self.lock.release()
def testRangeSansFile(self):
"range w/o step"
df = Datefile(datetime.strptime('2009-10-31', '%Y-%m-%d'))
start, end = df.get_date_range()
self.assertEquals(start.strftime('%Y-%m-%d'), '2009-10-31')
self.assertEquals(end.strftime('%Y-%m-%d'), '2009-10-31')
def testDefaultAccess(self):
df = Datefile(initial_date=datetime.strptime('2009-01-01', '%Y-%m-%d'))
self.assertEquals(df.get_date().strftime('%Y-%m-%d'), '2009-01-01')
def process_args(self):
"""Process any optional arguments and possitional parameters
Using the values provided, assemble ReportCriteria and
Datefile instances to control report generation.
"""
parser = OptionParser(usage=usage)
# Provide the ReportCriteria instance an error callback so any
# command line errors provoke the standard graceful exit with
# warning text.
self.criteria.error_callback = parser.error
parser.add_option("-u", "--user", dest="user",
default=self.user, help="database user")
parser.add_option("-p", "--password", dest="password",
default=self.password,
help="database password, or file containing "\
"just the password")
parser.add_option("-c", "--countdown", dest="countdown",
default=None,
help="count {down,up} the start and end dates "\
"set to 'forwards' or 'backwards' "\
"if desired")
parser.add_option("-i", "--include-updates",
action='store_true', dest="includeUpdates",
default=False, help="include "\
"visits updated since last similar report")
parser.add_option("--include-vitals",
action='store_true', dest="includeVitals",
default=False, help="include "\
"vitals (measured temperature, O2 "\
"saturation, influenza and H1N1 vaccine "\
"data) as additional columns in the "\
"report")
parser.add_option("-k", "--patient-class",
dest="patient_class",
default=None, help="use "\
"to filter report on a specific patient "\
"class [E,I,O]")
parser.add_option("-r", "--region", dest="region",
default=None,
help="reportable region defining limited set "\
"of facilities to include, by default "\
"all facilities are included")
parser.add_option("-s", "--save-and-upload",
action='store_true', dest="save_upload",
default=False, help="save file and upload to "\
"DOH")
parser.add_option("-x", "--save-without-upload",
action='store_true', dest="save_only",
default=False, help="save file but don't upload")
parser.add_option("-d", "--upload-diff",
action='store_true', dest="upload_diff",
default=False, help="upload differences only "\
"(from yesterdays like report) to DOH")
parser.add_option("-t", "--thirty-days",
action='store_true', dest="thirty_days",
default=False, help="include 30 days up to "\
"requested date ")
parser.add_option("-v", "--verbose", dest="verbosity",
action="count", default=self.verbosity,
help="increase output verbosity")
(options, args) = parser.parse_args()
if len(args) != 2:
parser.error("incorrect number of arguments")
# Database to query
self.criteria.database = args[0]
self.user = options.user
self.password = options.password
self.criteria.credentials(user=self.user,
password=self.password)
# Potential region restriction
self.criteria.reportable_region = options.region
# Potential patient class restriction
self.criteria.patient_class = options.patient_class
# Potential to include vitals (not tied to gipse format)
self.criteria.include_vitals = options.includeVitals
# Potential inclusion of updates
self.criteria.include_updates = options.includeUpdates
# Report date(s) and potential step direction.
# NB - several options affect report_method and must be set
# first!
initial_date = parseDate(args[1])
config = Config()
ps_file = os.path.join(config.get('general', 'tmp_dir',
default='/tmp'),
self.criteria.report_method)
step = options.thirty_days and 30 or None
direction = options.countdown
self.datefile = Datefile(initial_date=initial_date,
persistence_file=ps_file,
direction=direction,
step=step)
self.criteria.start_date, self.criteria.end_date =\
self.datefile.get_date_range()
# What to do once report is completed. Complicated, protect
# user from themselves!
self.save_report = options.save_upload or \
options.save_only or options.upload_diff
self.transmit_report = options.save_upload
self.transmit_differences = options.upload_diff
if options.save_only and options.save_upload:
parser.error("save-without-upload and save-and-upload "\
"are mutually exclusive")
if options.save_only and options.upload_diff:
parser.error("save-without-upload and upload-diff "\
"are mutually exclusive")
if options.upload_diff and options.save_upload:
parser.error("upload-diff and save-and-upload"\
"are mutually exclusive")
# Can't transmit w/o saving
if options.save_upload or options.upload_diff:
assert(self.save_report)
# Sanity check
if options.save_only:
assert(self.save_report and not self.transmit_report and
not self.transmit_differences)
# How verbosely to log
self.verbosity = options.verbosity
def processArgs(self):
""" Process any optional arguments and possitional parameters
"""
parser = OptionParser(usage=usage)
parser.add_option(
"-c",
"--countdown",
dest="countdown",
default=None,
help="count {down,up} date using date string in "
"%s - set to 'forwards' or 'backwards' "
"if desired" % self.datefile,
)
parser.add_option(
"-d",
"--date",
dest="date",
default=None,
help="single admission date to dedup " "(by default, checks the entire database)",
)
parser.add_option(
"-s",
"--skip-prep",
dest="skip_prep",
default=False,
action="store_true",
help="skip the expense of looking for new " "messages",
)
parser.add_option(
"-v",
"--verbose",
dest="verbosity",
action="count",
default=self.verbosity,
help="increase output verbosity",
)
parser.add_option(
"-m",
"--mart-port",
dest="mart_port",
default=self.mart_port,
type="int",
help="alternate port number for data mart",
)
parser.add_option(
"-w",
"--warehouse-port",
dest="warehouse_port",
default=self.warehouse_port,
type="int",
help="alternate port number for data warehouse",
)
(options, args) = parser.parse_args()
if len(args) != 2:
parser.error("incorrect number of arguments")
self.data_warehouse = args[0]
self.data_mart = args[1]
self.warehouse_port = parser.values.warehouse_port
self.mart_port = parser.values.mart_port
self.verbosity = parser.values.verbosity
self.skip_prep = parser.values.skip_prep
initial_date = parser.values.date and parseDate(parser.values.date) or None
self.datePersistence = Datefile(
initial_date=initial_date, persistence_file=self.datefile, direction=parser.values.countdown
)
self.reportDate = self.datePersistence.get_date()
def testDefaultAccess(self):
df = Datefile(initial_date=
datetime.strptime('2009-01-01', '%Y-%m-%d'))
self.assertEquals(df.get_date().strftime('%Y-%m-%d'),
'2009-01-01')
def process_args(self):
"""Process any optional arguments and possitional parameters
Using the values provided, assemble ReportCriteria and
Datefile instances to control report generation.
"""
parser = OptionParser(usage=usage)
# Provide the ReportCriteria instance an error callback so any
# command line errors provoke the standard graceful exit with
# warning text.
self.criteria.error_callback = parser.error
parser.add_option("-u",
"--user",
dest="user",
default=self.user,
help="database user")
parser.add_option("-p", "--password", dest="password",
default=self.password,
help="database password, or file containing "\
"just the password")
parser.add_option("-c", "--countdown", dest="countdown",
default=None,
help="count {down,up} the start and end dates "\
"set to 'forwards' or 'backwards' "\
"if desired")
parser.add_option("-i", "--include-updates",
action='store_true', dest="includeUpdates",
default=False, help="include "\
"visits updated since last similar report")
parser.add_option("--include-vitals",
action='store_true', dest="includeVitals",
default=False, help="include "\
"vitals (measured temperature, O2 "\
"saturation, influenza and H1N1 vaccine "\
"data) as additional columns in the "\
"report")
parser.add_option("-k", "--patient-class",
dest="patient_class",
default=None, help="use "\
"to filter report on a specific patient "\
"class [E,I,O]")
parser.add_option("-r", "--region", dest="region",
default=None,
help="reportable region defining limited set "\
"of facilities to include, by default "\
"all facilities are included")
parser.add_option("-s", "--save-and-upload",
action='store_true', dest="save_upload",
default=False, help="save file and upload to "\
"DOH")
parser.add_option("-x",
"--save-without-upload",
action='store_true',
dest="save_only",
default=False,
help="save file but don't upload")
parser.add_option("-d", "--upload-diff",
action='store_true', dest="upload_diff",
default=False, help="upload differences only "\
"(from yesterdays like report) to DOH")
parser.add_option("-t", "--thirty-days",
action='store_true', dest="thirty_days",
default=False, help="include 30 days up to "\
"requested date ")
parser.add_option("-v",
"--verbose",
dest="verbosity",
action="count",
default=self.verbosity,
help="increase output verbosity")
(options, args) = parser.parse_args()
if len(args) != 2:
parser.error("incorrect number of arguments")
# Database to query
self.criteria.database = args[0]
self.user = options.user
self.password = options.password
self.criteria.credentials(user=self.user, password=self.password)
# Potential region restriction
self.criteria.reportable_region = options.region
# Potential patient class restriction
self.criteria.patient_class = options.patient_class
# Potential to include vitals (not tied to gipse format)
self.criteria.include_vitals = options.includeVitals
# Potential inclusion of updates
self.criteria.include_updates = options.includeUpdates
# Report date(s) and potential step direction.
# NB - several options affect report_method and must be set
# first!
initial_date = parseDate(args[1])
config = Config()
ps_file = os.path.join(
config.get('general', 'tmp_dir', default='/tmp'),
self.criteria.report_method)
step = options.thirty_days and 30 or None
direction = options.countdown
self.datefile = Datefile(initial_date=initial_date,
persistence_file=ps_file,
direction=direction,
step=step)
self.criteria.start_date, self.criteria.end_date =\
self.datefile.get_date_range()
# What to do once report is completed. Complicated, protect
# user from themselves!
self.save_report = options.save_upload or \
options.save_only or options.upload_diff
self.transmit_report = options.save_upload
self.transmit_differences = options.upload_diff
if options.save_only and options.save_upload:
parser.error("save-without-upload and save-and-upload "\
"are mutually exclusive")
if options.save_only and options.upload_diff:
parser.error("save-without-upload and upload-diff "\
"are mutually exclusive")
if options.upload_diff and options.save_upload:
parser.error("upload-diff and save-and-upload"\
"are mutually exclusive")
# Can't transmit w/o saving
if options.save_upload or options.upload_diff:
assert (self.save_report)
# Sanity check
if options.save_only:
assert (self.save_report and not self.transmit_report
and not self.transmit_differences)
# How verbosely to log
self.verbosity = options.verbosity
class LongitudinalManager(object):
""" Abstraction to handle which db, user, etc. the deduplication
process should be run on. Handles runtime arguments and
execution.
Manages the process by farming out the individual visit
deduplication to a number of worker processes (necessary to take
advatage of multi-core processor and database as the limiting
factor).
"""
# The gating issue is the number of postgres connections that are
# allowed to run concurrently. Setting this to N-1 (where N is
# the number of cores) has proven the fastest and most reliable.
NUM_PROCS = 5
def __init__(self, data_warehouse=None, data_mart=None,
reportDate=None, database_user=None,
database_password=None, verbosity=0):
self.data_warehouse = data_warehouse
self.warehouse_port = 5432 # postgres default
self.data_mart = data_mart
self.mart_port = 5432 # postgres default
self.reportDate = reportDate and parseDate(reportDate) or None
self.database_user = database_user
self.database_password = database_password
self.dir, thisFile = os.path.split(__file__)
self.verbosity = verbosity
self.queue = JoinableQueue()
self.datefile = "/tmp/longitudinal_datefile"
self.datePersistence = Datefile(initial_date=self.reportDate)
self.lock = FileLock(LOCKFILE)
self.skip_prep = False
def __call__(self):
return self.execute()
def processArgs(self):
""" Process any optional arguments and possitional parameters
"""
parser = OptionParser(usage=usage)
parser.add_option("-c", "--countdown", dest="countdown",
default=None,
help="count {down,up} date using date string in "\
"%s - set to 'forwards' or 'backwards' "\
"if desired" % self.datefile)
parser.add_option("-d", "--date", dest="date", default=None,
help="single admission date to dedup "\
"(by default, checks the entire database)")
parser.add_option("-s", "--skip-prep", dest="skip_prep",
default=False, action="store_true",
help="skip the expense of looking for new "\
"messages")
parser.add_option("-v", "--verbose", dest="verbosity",
action="count", default=self.verbosity,
help="increase output verbosity")
parser.add_option("-m", "--mart-port", dest="mart_port",
default=self.mart_port, type="int",
help="alternate port number for data mart")
parser.add_option("-w", "--warehouse-port", dest="warehouse_port",
default=self.warehouse_port, type="int",
help="alternate port number for data warehouse")
(options, args) = parser.parse_args()
if len(args) != 2:
parser.error("incorrect number of arguments")
self.data_warehouse = args[0]
self.data_mart = args[1]
self.warehouse_port = parser.values.warehouse_port
self.mart_port = parser.values.mart_port
self.verbosity = parser.values.verbosity
self.skip_prep = parser.values.skip_prep
initial_date = parser.values.date and \
parseDate(parser.values.date) or None
self.datePersistence = Datefile(initial_date=initial_date,
persistence_file=self.datefile,
direction=parser.values.countdown)
self.reportDate = self.datePersistence.get_date()
def _prepDeduplicateTables(self):
""" Add any missing rows to the MessageProcessed table
This is the bridge between the data warehouse and the data
mart. In an effort to make the data mart independent of the
warehouse, the processed message data is kept in the mart. As
we're dealing with two distinct databases, there's no
referential integrity available at the database level, so care
should be taken.
"""
startTime = time.time()
logging.info("Starting INSERT INTO internal_message_processed "
"at %s", startTime)
# We can take advantage of an "add only" data_warehouse,
# knowing the hl7_msh_id is a sequence moving in the positive
# direction. Simply add any values greater than the previous
# max.
stmt = "SELECT max(hl7_msh_id) from internal_message_processed"
max_id = self.data_mart_access.engine.execute(stmt).first()[0]
if not max_id:
max_id = 0
new_msgs = list()
stmt = """SELECT hl7_msh_id, message_datetime, visit_id
FROM hl7_msh JOIN hl7_visit USING (hl7_msh_id) WHERE
hl7_msh_id > %d """ % max_id
rs = self.data_warehouse_access.engine.execute(stmt)
many = 500
while True:
results = rs.fetchmany(many)
if not results:
break
for r in results:
new_msgs.append(MessageProcessed(hl7_msh_id=r[0],
message_datetime=r[1],
visit_id=r[2]))
self.data_mart_access.session.add_all(new_msgs)
self.data_mart_access.session.commit()
logging.debug("added %d new messages" % len(new_msgs))
new_msgs = list()
logging.info("Added new rows to internal_message_processed in %s",
time.time() - startTime)
def _visitsToProcess(self):
""" Look up all distinct visit ids needing attention
Obtain unique list of visit_ids that have messages that
haven't previously been processed. If the user requested just
one days worth (i.e. -d) only that days visits will be
returned.
"""
visit_ids = list()
if not self.reportDate:
logging.info("Launch deduplication for entire database")
# Do the whole batch, that is, all that haven't been
# processed before.
stmt = """SELECT DISTINCT(visit_id) FROM
internal_message_processed
WHERE processed_datetime IS NULL"""
rs = self.data_mart_access.engine.execute(stmt)
many = 10000
while True:
results = rs.fetchmany(many)
if not results:
break
for r in results:
visit_ids.append(r[0])
else:
logging.info("Launch deduplication for %s",
self.reportDate)
# Process the requested day only - as we can't join across
# db boundaries - first acquire the full list of visits
# for the requested day from the data_warehouse to use in
# a massive 'in' clause
stmt = """SELECT DISTINCT(visit_id) FROM hl7_visit WHERE
admit_datetime BETWEEN '%s' AND '%s';""" %\
(self.reportDate, self.reportDate + timedelta(days=1))
self.access.raw_query(stmt)
rs = self.data_warehouse_access.engine.execute(stmt)
many = 1000
potential_visit_ids = list()
while True:
results = rs.fetchmany(many)
if not results:
break
for r in results:
#tmp_table.insert(r[0])
potential_visit_ids.append(r[0])
if potential_visit_ids:
query = self.data_mart_access.session.query(\
MessageProcessed.visit_id).distinct().\
filter(and_(MessageProcessed.processed_datetime ==
None,
MessageProcessed.visit_id.\
in_(potential_visit_ids)))
for r in query:
visit_ids.append(r[0])
logging.info("Found %d visits needing attention",
len(visit_ids))
return visit_ids
def tearDown(self):
""" Clean up any open handles/connections """
# now done in execute when we're done with teh connections
def execute(self):
""" Start the process """
# Initialize logging now (verbosity is now set regardless of
# invocation method)
configure_logging(verbosity=self.verbosity,
logfile="longitudinal-manager.log")
logging.info("Initiate deduplication for %s",
(self.reportDate and self.reportDate or
"whole database"))
# Only allow one instance of the manager to run at a time.
if self.lock.is_locked():
logging.warn("Can't continue, %s is locked ", LOCKFILE)
return
if systemUnderLoad():
logging.warn("system under load - continue anyhow")
try:
self.lock.acquire()
self.access = DirectAccess(database=self.data_warehouse,
port=self.warehouse_port,
user=self.database_user,
password=self.database_password)
self.data_warehouse_access = AlchemyAccess(
database=self.data_warehouse,
port=self.warehouse_port,
user=self.database_user, password=self.database_password)
self.data_mart_access = AlchemyAccess(
database=self.data_mart, port=self.mart_port,
user=self.database_user, password=self.database_password)
startTime = time.time()
if not self.skip_prep:
self._prepDeduplicateTables()
visits_to_process = self._visitsToProcess()
# Now done with db access needs at the manager level
# free up resources:
self.data_mart_access.disconnect()
self.data_warehouse_access.disconnect()
self.access.close()
# Set of locks used, one for each table needing protection
# from asynchronous inserts. Names should match table
# minus 'dim_' prefix, plus '_lock' suffix
# i.e. dim_location -> 'location_lock'
table_locks = {'admission_source_lock': Lock(),
'admission_o2sat_lock': Lock(),
'admission_temp_lock': Lock(),
'assigned_location_lock': Lock(),
'admit_reason_lock': Lock(),
'chief_complaint_lock': Lock(),
'diagnosis_lock': Lock(),
'disposition_lock': Lock(),
'flu_vaccine_lock': Lock(),
'h1n1_vaccine_lock': Lock(),
'lab_flag_lock': Lock(),
'lab_result_lock': Lock(),
'location_lock': Lock(),
'note_lock': Lock(),
'order_number_lock': Lock(),
'performing_lab_lock': Lock(),
'pregnancy_lock': Lock(),
'race_lock': Lock(),
'reference_range_lock': Lock(),
'service_area_lock': Lock(),
'specimen_source_lock': Lock(),
}
# If we have visits to process, fire up the workers...
if len(visits_to_process) > 1:
for i in range(self.NUM_PROCS):
dw = Process(target=LongitudinalWorker,
kwargs={'queue': self.queue,
'procNumber': i,
'data_warehouse': self.data_warehouse,
'warehouse_port': self.warehouse_port,
'data_mart': self.data_mart,
'mart_port': self.mart_port,
'dbUser': self.database_user,
'dbPass': self.database_password,
'table_locks': table_locks,
'verbosity': self.verbosity})
dw.daemon = True
dw.start()
# Populate the queue
for v in visits_to_process:
self.queue.put(v)
# Wait on the queue until empty
self.queue.join()
# Common cleanup
self.tearDown()
self.datePersistence.bump_date()
logging.info("Queue is empty - done in %s", time.time() -
startTime)
finally:
self.lock.release()