def runImport(doc_handler=None):
""" Run the actual linking.
You can provide a :class:`DocHandler` subclass which will handle the JSON
documents, for example store them to MongoDB for the MongoDocHandler. These
classes are defined in `rxnorm_link_run.py` for now.
"""
# install keyboard interrupt handler
def signal_handler(signal, frame):
print("\nx> Aborted")
sys.exit(0)
signal.signal(signal.SIGINT, signal_handler)
# prepare RxNorm databases
try:
RxNorm.check_database()
rxhandle = RxNormLookup()
rxhandle.prepare_to_cache_classes()
except Exception as e:
logging.error(e)
sys.exit(1)
# fetch rxcui's for drug-type concepts (i.e. restrict by TTY)
drug_types = ('SCD', 'SCDC', 'SBDG', 'SBD', 'SBDC', 'BN', 'SBDF', 'SCDG',
'SCDF', 'IN', 'MIN', 'PIN', 'BPCK', 'GPCK')
param = ', '.join(['?' for d in drug_types])
all_sql = "SELECT RXCUI, TTY from RXNCONSO where SAB='RXNORM' and TTY in ({})".format(
param)
all_drugs = rxhandle.fetchAll(all_sql, drug_types)
num_drugs = len(all_drugs)
# traverse VA classes; starts the VA drug class caching process if needed,
# which runs a minute or two
if rxhandle.can_cache():
initVA(rxhandle)
traverseVA(rxhandle, rounds=5, expect=num_drugs)
# loop all concepts
i = 0
w_ti = 0
w_va = 0
w_either = 0
last_report = datetime.now()
print('-> Indexing {} items'.format(num_drugs))
for res in all_drugs:
params = [res[0]]
params.extend(drug_types)
label = rxhandle.lookup_rxcui_name(res[0]) # fast (indexed column)
ndc = rxhandle.ndc_for_rxcui(res[0]) # fast (indexed column)
ndc = RxNorm.ndc_normalize_list(ndc) # fast (string permutation)
# find ingredients, drug classes and more
ingr = toIngredients(rxhandle, [res[0]], res[1]) # rather slow
ti = toTreatmentIntents(rxhandle, ingr, 'IN') # requires "ingr"
va = toDrugClasses(rxhandle,
res[0]) # fast, loads from our cached table
gen = toBrandAndGeneric(rxhandle, [res[0]], res[1]) # fast
comp = toComponents(rxhandle, [res[0]], res[1]) # fast
mech = toMechanism(rxhandle, ingr, 'IN') # fast
# create JSON-ready dictionary (save space by not adding empty properties)
d = {
'rxcui': res[0],
'tty': res[1],
'label': label,
}
if len(ndc) > 0:
d['ndc'] = list(ndc)
if len(ingr) > 0:
d['ingredients'] = list(ingr)
if len(ti) > 0:
d['treatmentIntents'] = list(ti)
if len(va) > 0:
d['drugClasses'] = list(va)
if len(gen) > 0:
d['generics'] = list(gen)
if len(comp) > 0:
d['components'] = list(comp)
if len(mech) > 0:
d['mechanisms'] = list(mech)
# count
i += 1
if len(ti) > 0:
w_ti += 1
if len(va) > 0:
w_va += 1
if len(ti) > 0 or len(va) > 0:
w_either += 1
# The dictionary "d" at this point contains all the drug's precomputed
# properties, to debug print this:
#print(json.dumps(d, sort_keys=True, indent=2))
if doc_handler:
doc_handler.addDocument(d)
# log progress every 2 seconds or so
if (datetime.now() - last_report).seconds > 2:
last_report = datetime.now()
print('--> {:.1%} n: {}, ti: {}, va: {}, either: {}'.format(
i / num_drugs, i, w_ti, w_va, w_either),
end="\r")
# loop done, finalize
if doc_handler:
doc_handler.finalize()
print('--> {:.1%} n: {}, ti: {}, va: {}, either: {}'.format(
i / num_drugs, i, w_ti, w_va, w_either))
print('-> Done')
def runImport(doc_handler=None):
""" Run the actual linking.
You can provide a :class:`DocHandler` subclass which will handle the JSON
documents, for example store them to MongoDB for the MongoDocHandler. These
classes are defined in `rxnorm_link_run.py` for now.
"""
# install keyboard interrupt handler
def signal_handler(signal, frame):
print("\nx> Aborted")
sys.exit(0)
signal.signal(signal.SIGINT, signal_handler)
# prepare RxNorm databases
try:
RxNorm.check_database()
rxhandle = RxNormLookup()
rxhandle.prepare_to_cache_classes()
except Exception as e:
logging.error(e)
sys.exit(1)
# fetch rxcui's for drug-type concepts (i.e. restrict by TTY)
drug_types = ('SCD', 'SCDC', 'SBDG', 'SBD', 'SBDC', 'BN', 'SBDF', 'SCDG', 'SCDF', 'IN', 'MIN', 'PIN', 'BPCK', 'GPCK')
param = ', '.join(['?' for d in drug_types])
all_sql = "SELECT RXCUI, TTY from RXNCONSO where SAB='RXNORM' and TTY in ({})".format(param)
all_drugs = rxhandle.fetchAll(all_sql, drug_types)
num_drugs = len(all_drugs)
# traverse VA classes; starts the VA drug class caching process if needed,
# which runs a minute or two
if rxhandle.can_cache():
initVA(rxhandle)
traverseVA(rxhandle, rounds=5, expect=num_drugs)
# loop all concepts
i = 0
w_ti = 0
w_va = 0
w_either = 0
last_report = datetime.now()
print('-> Indexing {} items'.format(num_drugs))
for res in all_drugs:
params = [res[0]]
params.extend(drug_types)
label = rxhandle.lookup_rxcui_name(res[0]) # fast (indexed column)
ndc = rxhandle.ndc_for_rxcui(res[0]) # fast (indexed column)
ndc = RxNorm.ndc_normalize_list(ndc) # fast (string permutation)
# find ingredients, drug classes and more
ingr = toIngredients(rxhandle, [res[0]], res[1]) # rather slow
ti = toTreatmentIntents(rxhandle, ingr, 'IN') # requires "ingr"
va = toDrugClasses(rxhandle, res[0]) # fast, loads from our cached table
gen = toBrandAndGeneric(rxhandle, [res[0]], res[1]) # fast
comp = toComponents(rxhandle, [res[0]], res[1]) # fast
mech = toMechanism(rxhandle, ingr, 'IN') # fast
# create JSON-ready dictionary (save space by not adding empty properties)
d = {
'rxcui': res[0],
'tty': res[1],
'label': label,
}
if len(ndc) > 0:
d['ndc'] = list(ndc)
if len(ingr) > 0:
d['ingredients'] = list(ingr)
if len(ti) > 0:
d['treatmentIntents'] = list(ti)
if len(va) > 0:
d['drugClasses'] = list(va)
if len(gen) > 0:
d['generics'] = list(gen)
if len(comp) > 0:
d['components'] = list(comp)
if len(mech) > 0:
d['mechanisms'] = list(mech)
# count
i += 1
if len(ti) > 0:
w_ti += 1
if len(va) > 0:
w_va += 1
if len(ti) > 0 or len(va) > 0:
w_either += 1
# The dictionary "d" at this point contains all the drug's precomputed
# properties, to debug print this:
#print(json.dumps(d, sort_keys=True, indent=2))
if doc_handler:
doc_handler.addDocument(d)
# log progress every 5 seconds or so
if (datetime.now() - last_report).seconds > 5:
last_report = datetime.now()
print('-> {:.1%} n: {}, ti: {}, va: {}, either: {}'.format(i / num_drugs, i, w_ti, w_va, w_either), end="\r")
# loop done, finalize
if doc_handler:
doc_handler.finalize()
print('-> {:.1%} n: {}, ti: {}, va: {}, either: {}'.format(i / num_drugs, i, w_ti, w_va, w_either))
print('=> Done')
def test_ndc_normalization(self):
""" Test NDC normalization.
"""
# 6-4-2
self.assertEqual('00074148614', RxNorm.ndc_normalize('000074-1486-14'))
self.assertEqual('51227615900', RxNorm.ndc_normalize('051227-6159-**'))
self.assertEqual('58734000101', RxNorm.ndc_normalize('058734-0001-*1'))
# 6-4-1
self.assertEqual('00854684102', RxNorm.ndc_normalize('000854-6841-2'))
# 6-4: treat as 6-4-2 with two trailing zeroes
self.assertEqual('57982011000', RxNorm.ndc_normalize('057982-0110'))
self.assertEqual('12579005600', RxNorm.ndc_normalize('012579-*056'))
# 6-3-2
self.assertEqual('57982012312', RxNorm.ndc_normalize('057982-123-12'))
# 6-3-1
self.assertEqual('57982098709', RxNorm.ndc_normalize('057982-987-9'))
# 5-4-2
self.assertEqual('17317093201', RxNorm.ndc_normalize('17317-0932-01'))
# 5-4-1
self.assertEqual('36987315601', RxNorm.ndc_normalize('36987-3156-1'))
# 5-3-2
self.assertEqual('24730041205', RxNorm.ndc_normalize('24730-412-05'))
# 4-4-2
self.assertEqual('00268010310', RxNorm.ndc_normalize('0268-0103-10'))
# 12 digit VANDF
self.assertEqual('03475476541', RxNorm.ndc_normalize('003475476541'))
# normalized already
self.assertEqual('04458632698', RxNorm.ndc_normalize('04458632698'))
# invalid
self.assertIsNone(RxNorm.ndc_normalize('0054478962'))
self.assertIsNone(RxNorm.ndc_normalize('547668531244'))
self.assertIsNone(RxNorm.ndc_normalize('0054478962796'))
self.assertIsNone(RxNorm.ndc_normalize('0a79b2-c87-9'))
self.assertIsNone(RxNorm.ndc_normalize('si-lly-te-st'))
self.assertIsNone(RxNorm.ndc_normalize('just-a-rand-test-string'))