def run(self):
output_dir = self.output().path
common.shell_cmd('mkdir -p %s', output_dir)
for i in range(len(self.input())):
input_dir = self.input()[i].path
download_util.extract_and_clean(input_dir, 'ISO-8859-1//TRANSLIT',
'UTF-8', 'txt')
def run(self):
input_dir = self.input().path
output_dir = self.output().path
common.shell_cmd("mkdir -p %s", dirname(output_dir))
NEEDS_HEADERS = {"estabtypes.txt": ["establishment_type_id", "description"]}
inputs = []
for input_file in glob.glob(input_dir + "/*.txt"):
if basename(input_file) in REMAPPED_FILES:
continue
header_key = basename(input_file)
fieldnames = NEEDS_HEADERS.get(header_key, None)
inputs.append(
parallel.Collection.from_glob(
input_file,
parallel.CSVDictLineInput(
delimiter="|", fieldnames=fieldnames, quoting=csv.QUOTE_NONE, escapechar="\\"
),
)
)
parallel.mapreduce(
inputs=inputs, mapper=TXT2JSONMapper(), reducer=parallel.IdentityReducer(), output_prefix=self.output().path
)
def run(self):
input_dir = self.input().path
output_dir = self.output().path
common.shell_cmd('mkdir -p %s', output_dir)
# TODO(hansnelsen): change to the openfda.parallel version of multiprocess
pool = multiprocessing.Pool(processes=3)
for i in range(PARTITIONS):
partition_dict = {}
output_filename = join(output_dir, str(i) + '.maude.json')
# Get all of the files for the current partition
for filename in glob.glob(input_dir + '/' + str(i) + '.*.txt'):
for file_type in CATEGORIES:
if file_type in filename:
logging.info('Using file %s for joining', filename)
partition_dict[file_type] = filename
logging.info('Starting Partition %d', i)
master_file = partition_dict['mdrfoi']
patient_file = partition_dict['patient']
device_file = partition_dict['foidev']
text_file = partition_dict['foitext']
pool.apply_async(join_maude.join_maude,
(master_file, patient_file, device_file,
text_file, output_filename))
pool.close()
pool.join()
def _run(self):
output_dir = self.output().path
common.shell_cmd('mkdir -p %s', output_dir)
change_log = csv.reader(open(self.change_log_file, 'r'))
batches = collections.defaultdict(list)
for row in change_log:
spl_id, spl_type, spl_date = row
# Only grab the human and ceullar therapy labels for this index
valid_types = ['cellular therapy', 'human']
is_valid = False
for valid_type in valid_types:
if spl_type.lower().find(valid_type) != -1:
is_valid = True
break
# only process valid document types
if is_valid:
# All blank dates to be treated as the week of June 1, 2009
if not spl_date:
spl_date = '20090601120000'
date = arrow.get(spl_date, 'YYYYMMDDHHmmss')
batches[date.ceil('week')].append(spl_id)
for batch_date, ids in batches.items():
batch_file = '%s.ids' % batch_date.format('YYYYMMDD')
batch_out = open(join(output_dir, batch_file), 'w')
unique_ids = list(set(ids))
batch_out.write('\n'.join(unique_ids))
def run(self):
output_dir = self.output().path
common.shell_cmd('mkdir -p %s', output_dir)
input_dir = self.local_dir
for zip_filename in glob.glob(input_dir + '/*.zip'):
common.shell_cmd_quiet('unzip -ou %s -d %s', zip_filename,
output_dir)
def run(self):
crawl_dir = dirname(dirname(self.output().path))
common.shell_cmd('mkdir -p %s', dirname(self.output().path))
manifests = walk_glob('manifest.json', crawl_dir)
records = []
for file_name in manifests:
records.append(json.load(open(file_name)))
# Default data structure that creates the appropriate structure on the
# first put so that we can blindly use `+=` when appropriate.
combined = collections.defaultdict(
lambda: collections.defaultdict(
lambda: {
'export_date': None,
'partitions': [],
'total_records': 0
}
)
)
# Walk over all of the manifests and create a single dictionary
for row in records:
for domain, value in row.items():
for sub, val in value.items():
combined[domain][sub]['export_date'] = val.get('export_date', '')
combined[domain][sub]['partitions'] += val.get('partitions', [])
combined[domain][sub]['total_records'] += val.get('total_records', 0)
with open(join(self.output().path), 'w') as json_out:
json.dump(combined, json_out, indent=2)
def run(self):
input_dir = self.input().path
output_dir = self.output().path
common.shell_cmd('mkdir -p %s', dirname(output_dir))
NEEDS_HEADERS = {
'estabtypes.txt': ['establishment_type_id', 'description']
}
inputs = []
for input_file in glob.glob(input_dir + '/*.txt'):
if basename(input_file) in REMAPPED_FILES:
continue
header_key = basename(input_file)
fieldnames = NEEDS_HEADERS.get(header_key, None)
inputs.append(
parallel.Collection.from_glob(
input_file,
parallel.CSVDictLineInput(delimiter='|',
fieldnames=fieldnames,
quoting=csv.QUOTE_NONE,
escapechar='\\')))
parallel.mapreduce(inputs=inputs,
mapper=TXT2JSONMapper(),
reducer=parallel.IdentityReducer(),
output_prefix=self.output().path)
def run(self):
input_dir = self.input().path
output_dir = self.output().path
common.shell_cmd('mkdir -p %s', output_dir)
# TODO(hansnelsen): change to the openfda.parallel version of multiprocess
pool = multiprocessing.Pool(processes=6)
for i in range(PARTITIONS):
partition_dict = {}
output_filename = join(output_dir, str(i) + '.maude.json')
# Get all of the files for the current partition
for filename in glob.glob(input_dir + '/' + str(i) + '.*.txt'):
for file_type in CATEGORIES:
if file_type in filename:
logging.info('Using file %s for joining', filename)
partition_dict[file_type] = filename
logging.info('Starting Partition %d', i)
master_file = partition_dict['mdrfoi']
patient_file = partition_dict['patient']
device_file = partition_dict['foidev']
text_file = partition_dict['foitext']
pool.apply_async(join_maude.join_maude, (master_file,
patient_file,
device_file,
text_file,
output_filename))
pool.close()
pool.join()
def run(self):
common.shell_cmd('mkdir -p %s', self.local_dir)
soup = BeautifulSoup(urlopen(CAERS_DOWNLOAD_PAGE_URL).read(), 'lxml')
for a in soup.find_all(title=re.compile('CAERS ASCII.*')):
if 'Download CAERS ASCII' in re.sub(r'\s', ' ', a.text):
fileURL = urljoin('https://www.fda.gov', a['href'])
common.download(fileURL, join(self.output().path, a.attrs['title']+'.csv'))
def run(self):
# Since we only iterate over dates in the umbrella process, we need to
# skip batch files that do not exist
output_file = self.output().path
if not os.path.exists(self.batch):
common.shell_cmd('touch %s', output_file)
return
input_file = self.input()[1].path
es = elasticsearch.Elasticsearch(self.es_host)
index_util.start_index_transaction(es, 'druglabel', self.epoch)
parallel.mapreduce(
input_collection=parallel.Collection.from_sharded(input_file),
mapper=index_util.LoadJSONMapper(self.es_host,
'druglabel',
'spl',
self.epoch,
docid_key='set_id',
version_key='version'),
reducer=parallel.NullReducer(),
output_prefix='/tmp/loadjson.druglabel',
num_shards=1,
map_workers=1)
index_util.commit_index_transaction(es, 'druglabel')
common.shell_cmd('touch %s', output_file)
def run(self):
crawl_dir = dirname(dirname(self.output().path))
common.shell_cmd('mkdir -p %s', dirname(self.output().path))
manifests = walk_glob('manifest.json', crawl_dir)
records = []
for file_name in manifests:
records.append(json.load(open(file_name)))
# Default data structure that creates the appropriate structure on the
# first put so that we can blindly use `+=` when appropriate.
combined = collections.defaultdict(
lambda: collections.defaultdict(lambda: {
'export_date': None,
'partitions': [],
'total_records': 0
}))
# Walk over all of the manifests and create a single dictionary
for row in records:
for domain, value in row.items():
for sub, val in value.items():
combined[domain][sub]['export_date'] = val.get(
'export_date', '')
combined[domain][sub]['partitions'] += val.get(
'partitions', [])
combined[domain][sub]['total_records'] += val.get(
'total_records', 0)
with open(join(self.output().path), 'w') as json_out:
json.dump(combined, json_out, indent=2)
def run(self):
output_dir = dirname(self.output().path)
common.shell_cmd('mkdir -p %s', output_dir)
dt = arrow.get(self.batch)
url = DOWNLOAD_URL.replace('--Y--', str(dt.year))
url = url.replace('--M--', str(dt.month))
url = url.replace('--D--', str(dt.day))
download_to_file_with_retry(url, self.output().path)
def run(self):
common.shell_cmd('mkdir -p %s', dirname(self.output().path))
input_files = glob.glob(self.input().path + '/*.txt')
parallel.mapreduce(
parallel.Collection.from_glob(
input_files, parallel.CSVDictLineInput(delimiter='|', strip_str='\0')),
mapper=PMAMapper(),
reducer=parallel.IdentityReducer(),
output_prefix=self.output().path)
def run(self):
common.shell_cmd('mkdir -p %s', self.local_dir)
soup = BeautifulSoup(urllib2.urlopen(DAILY_MED_DOWNLOADS_PAGE).read(), 'lxml')
for a in soup.find_all(href=re.compile('.*.zip')):
if '_human_' in a.text:
try:
common.download(a['href'], join(self.local_dir, a['href'].split('/')[-1]))
except ProcessException as e:
logging.error("Could not download a DailyMed SPL archive: {0}: {1}".format(a['href'], e))
def run(self):
output_dir = self.output().path
common.shell_cmd('mkdir -p %s', output_dir)
for i in range(len(self.input())):
input_dir = self.input()[i].path
download_util.extract_and_clean(input_dir,
'ISO-8859-1//TRANSLIT',
'UTF-8',
'txt')
def run(self):
common.shell_cmd('mkdir -p %s', dirname(self.output().path))
input_files = glob.glob(self.input().path + '/*.txt')
parallel.mapreduce(parallel.Collection.from_glob(
input_files,
parallel.CSVDictLineInput(delimiter='|', strip_str='\0')),
mapper=PMAMapper(),
reducer=parallel.IdentityReducer(),
output_prefix=self.output().path)
def run(self):
common.shell_cmd('mkdir -p %s', join(BASE_DIR, 'tmp'))
files = glob.glob(self.input().path + '/*/*.json')
parallel.mapreduce(
parallel.Collection.from_glob(files, parallel.JSONLineInput()),
mapper=ParallelExportMapper(output_dir=self.output().path),
reducer=parallel.NullReducer(),
output_prefix=join(BASE_DIR, 'tmp'),
output_format=parallel.NullOutput(),
map_workers=10)
def run(self):
common.shell_cmd('mkdir -p %s', join(BASE_DIR, 'tmp'))
files = glob.glob(self.input().path + '/*/*.json')
parallel.mapreduce(
parallel.Collection.from_glob(files, parallel.JSONLineInput()),
mapper=ParallelExportMapper(output_dir=self.output().path),
reducer=parallel.NullReducer(),
output_prefix=join(BASE_DIR, 'tmp'),
output_format=parallel.NullOutput(),
map_workers=10)
def run(self):
common.shell_cmd('mkdir -p %s', dirname(self.output().path))
input_files = glob.glob(self.input().path + '/*.txt')
parallel.mapreduce(parallel.Collection.from_glob(
input_files,
parallel.CSVDictLineInput(delimiter='|',
quoting=csv.QUOTE_NONE,
escapechar='\\')),
mapper=ClassificationMapper(),
reducer=parallel.IdentityReducer(),
output_prefix=self.output().path)
def run(self):
sync_path = join(BASE_DIR, self.date_str)
target_bucket = 's3://%s/%s/' % (self.download_bucket, self.date_str)
s3_cmd = [
'aws', '--profile',
config.aws_profile(), 's3', 'sync', sync_path, target_bucket,
'--exclude "*"', '--include "*.zip"', '--include "*schema.json"'
]
common.shell_cmd(' '.join(s3_cmd))
common.shell_cmd('touch %s', self.output().path)
def run(self):
common.shell_cmd('mkdir -p %s', dirname(self.output().path))
input_files = glob.glob(self.input().path + '/*.txt')
parallel.mapreduce(
parallel.Collection.from_glob(
input_files, parallel.CSVDictLineInput(delimiter='|',
quoting=csv.QUOTE_NONE,
escapechar='\\')),
mapper=ClassificationMapper(),
reducer=parallel.IdentityReducer(),
output_prefix=self.output().path)
def run(self):
output_dir = self.output().path
common.shell_cmd("mkdir -p %s", output_dir)
input_dir = self.input()[0].path
supplemental_dir = self.input()[1].path
download_util.extract_and_clean(input_dir, "ISO-8859-1", "UTF-8", "txt")
# One of the files needs to be remapped from one column (submission_number)
# to two columns (pma_number and k_number) depending on the prefix.
file_name = "registration_listing.txt"
output_file = join(output_dir, "remapped_" + file_name)
remap_supplemental_files(join(output_dir, file_name), join(supplemental_dir, file_name), output_file)
def run(self):
output_dir = self.output().path
common.shell_cmd('mkdir -p %s', output_dir)
date = self.batch
if date >= CROSSOVER_XML_START_DATE and date <= CROSSOVER_XML_END_DATE:
url = CROSSOVER_XML_URL
else:
url = CURRENT_XML_BASE_URL
url = url.replace('WEEK', date.strftime('%m%d%Y'))
file_name = 'enforcementreport.xml'
xml_file = '%(output_dir)s/%(file_name)s' % locals()
download_to_file_with_retry(url, xml_file)
def extract_and_clean(input_dir, source_encoding, target_encoding, file_type):
''' A utility function that extracts all of the zip files in a directory and
converts the files from a source encoding to a target encoding.
'''
for zip_filename in glob.glob(input_dir + '/*.zip'):
txt_name = zip_filename.replace('zip', file_type)
txt_name = txt_name.replace('raw', 'extracted')
common.shell_cmd('mkdir -p %s', dirname(txt_name))
cmd = 'unzip -p %s | iconv -f %s -t %s -c > %s'
logging.info('Unzipping and converting %s', zip_filename)
common.shell_cmd(cmd, zip_filename, source_encoding, target_encoding,
txt_name)
def run(self):
output_dir = self.output().path
common.shell_cmd('mkdir -p %s', output_dir)
date = self.batch
if date >= CROSSOVER_XML_START_DATE and date <= CROSSOVER_XML_END_DATE:
url = CROSSOVER_XML_URL
else:
url = CURRENT_XML_BASE_URL
url = url.replace('WEEK', date.strftime('%m%d%Y'))
file_name = 'enforcementreport.xml'
xml_file = '%(output_dir)s/%(file_name)s' % locals()
download_to_file_with_retry(url, xml_file)
def run(self):
sync_path = join(BASE_DIR, self.date_str)
target_bucket = S3_BASE_BUCKET + '%s/' % self.date_str
for data_path in self.output():
s3_cmd = [
'aws', '--profile',
config.aws_profile(), 's3', 'sync', sync_path, target_bucket,
'--exclude "*"', '--include "*.zip"',
'--include "*schema.json"'
]
common.shell_cmd(' '.join(s3_cmd))
common.shell_cmd('touch %s', data_path.path)
def run(self):
logging.info('Extracting: %s', (self.input().path))
extract_dir = dirname(self.input().path)
gsrs_file_name = os.path.basename(self.input().path)
gz_filename = os.path.splitext(gsrs_file_name)[0] + ".gz"
gsrs_file = join(extract_dir, gsrs_file_name)
gz_file = join(extract_dir, gz_filename)
os.rename(gsrs_file, gz_file)
common.shell_cmd('gunzip ' + gz_file)
os.rename(
os.path.splitext(gz_file)[0],
os.path.splitext(gz_file)[0] + ".json")
def run(self):
output_dir = self.output().path
common.shell_cmd('mkdir -p %s', output_dir)
input_dir = self.input()[0].path
supplemental_dir = self.input()[1].path
download_util.extract_and_clean(input_dir, 'ISO-8859-1', 'UTF-8', 'txt')
# One of the files needs to be remapped from one column (submission_number)
# to two columns (pma_number and k_number) depending on the prefix.
file_name = 'registration_listing.txt'
output_file = join(output_dir, 'remapped_' + file_name)
remap_supplemental_files(join(output_dir, file_name),
join(supplemental_dir, file_name),
output_file)
def run(self):
zip_filename = self.input().path
output_filename = self.output().path
output_dir = dirname(output_filename)
common.shell_cmd('mkdir -p %s' % output_dir)
cmd = 'unzip -o %(zip_filename)s \
-d %(output_dir)s' % locals()
common.shell_cmd(cmd)
# UNII filename varies; find and rename to a standardized name.
# It is now a tab-delimited CSV instead of an XML as before.
for file in glob.glob(join(output_dir, 'UNII*Names*.txt')):
logging.info('Renaming %s', file)
os.rename(file, output_filename)
def _run(self):
shutil.rmtree(self.output().path, ignore_errors=True)
os.makedirs(self.output().path)
# Get all of the endpoints served by this index
# Create an `EndpointExport` object for each endpoint in order to export
# each endpoint properly.
#
# Endpoint exports can be:
# date range based (quarterly output)
# filter based (index serves many endpoints)
# vanilla (endpoint is 1 to 1 with index and it is exported all at once)
for endpoint, index_name in ENDPOINT_INDEX_MAP.items():
endpoint_batches = []
chunks = CUSTOM_CHUNKS.get(endpoint, DEFAULT_CHUNKS)
if endpoint in RANGE_ENDPOINT_MAP:
params = RANGE_ENDPOINT_MAP[endpoint]
params['chunks'] = chunks
endpoint_batches = _make_date_range_endpoint_batch(
endpoint, params)
elif endpoint in FILTERED_ENPOINT_MAP:
params = FILTERED_ENPOINT_MAP[endpoint]
query = EndpointExport.build_term_filter(**params)
endpoint_batches.append(
EndpointExport(endpoint, query=query, chunks=chunks))
else:
endpoint_batches.append(EndpointExport(endpoint,
chunks=chunks))
# This is a hack to overcome the shortcoming of the parallel library of
# only having one mapper process for a tiny, single file input. Since we
# want to execute these endpoint batches in parallel, we write each task
# to its own file. It will create a mapper for each file.
for ep in endpoint_batches:
partition = ep.partition if ep.partition else 'all'
if 'enforcement' in ep.endpoint:
partition = ep.endpoint.replace('enforcement',
'').replace('/', '')
elif 'label' in ep.endpoint:
partition = ep.endpoint.replace('label',
'').replace('/', '')
output_dir = join(self.output().path, index_name)
common.shell_cmd('mkdir -p %s', output_dir)
file_name = join(output_dir, partition + '.json')
with open(file_name, 'w') as json_out:
json_dict = json.dumps(ep.__dict__)
json_out.write(json_dict + '\n')
def extract_and_clean(input_dir, source_encoding, target_encoding, file_type):
''' A utility function that extracts all of the zip files in a directory and
converts the files from a source encoding to a target encoding.
'''
for zip_filename in glob.glob(input_dir + '/*.zip'):
txt_name = zip_filename.replace('zip', file_type)
txt_name = txt_name.replace('raw', 'extracted')
common.shell_cmd('mkdir -p %s', dirname(txt_name))
cmd = 'unzip -p %s | iconv -f %s -t %s -c > %s'
logging.info('Unzipping and converting %s', zip_filename)
common.shell_cmd(cmd,
zip_filename,
source_encoding,
target_encoding,
txt_name)
def run(self):
schema_file = self.get_schemafile()
assert os.path.exists(
schema_file
), 'No schema file available for index %s' % self.index_name
es_client = elasticsearch.Elasticsearch(config.es_host())
endpoints = self.get_endpoints()
# Get all of the endpoints served by this index
# Create an `EndpointExport` object for each endpoint in order to export
# each endpoint properly.
#
# Endpoint exports can be:
# date range based (quarterly output)
# filter based (index serves many endpoints)
# vanilla (endpoint is 1 to 1 with index and it is exported all at once)
endpoint_batches = []
for endpoint in endpoints:
chunks = CUSTOM_CHUNKS.get(endpoint, DEFAULT_CHUNKS)
if endpoint in RANGE_ENDPOINT_MAP:
params = RANGE_ENDPOINT_MAP[endpoint]
params['chunks'] = chunks
endpoint_batches = _make_date_range_endpoint_batch(
endpoint, params)
elif endpoint in FILTERED_ENPOINT_MAP:
params = FILTERED_ENPOINT_MAP[endpoint]
query = EndpointExport.build_term_filter(**params)
endpoint_batches.append(
EndpointExport(endpoint, query=query, chunks=chunks))
else:
endpoint_batches.append(EndpointExport(endpoint,
chunks=chunks))
# Dump each of the `EndpointExport` objects in the list
for ep in endpoint_batches:
# The output_dir will be the same for all outputs, once you factor out
# the endpoint, so we can safely look at the first one only.
output_dir = dirname(dirname(self.output()[0].path))
endpoint_dir = join(output_dir, ep.endpoint[1:])
index_util.dump_index(es_client,
ep.index_name,
ep.endpoint,
join(endpoint_dir, ep.partition),
cleaner=omit_internal_keys,
query=ep.query,
chunks=ep.chunks)
common.shell_cmd('cp %s %s', schema_file, endpoint_dir)
def run(self):
sync_path = join(BASE_DIR, self.date_str)
target_bucket = 's3://%s/%s/' % (self.download_bucket, self.date_str)
s3_cmd = [
'aws',
'--profile',
config.aws_profile(),
's3',
'sync',
sync_path,
target_bucket,
'--exclude "*"',
'--include "*.zip"',
'--include "*schema.json"']
common.shell_cmd(' '.join(s3_cmd))
common.shell_cmd('touch %s', self.output().path)
def run(self):
output_dir = dirname(self.output().path)
common.shell_cmd('mkdir -p %s', output_dir)
end = arrow.get(self.batch)
start = end.shift(days=-6)
id_list = self._fetch_ids(start, end)
if len(id_list) >= 500:
# We have tried to get the entire list of weekly results in one shot, but apparently there are more than 500
# hits and CDRH recall search does not support more than 500 results per search and does not support paging either.
# This occurrence is rare, but when it does happen we need to re-retrieve the results day by day.
for r in arrow.Arrow.range('day', start, end):
id_list = id_list + self._fetch_ids(r, r)
df = pd.DataFrame(data=list(set(id_list)), columns=['id'])
df.to_csv(self.output().path, index=False)
def map(self, key, value, output):
es_client = elasticsearch.Elasticsearch(config.es_host(), timeout=120)
ep = common.ObjectDict(value)
schema_file = join(SCHEMA_DIR, ep.index_name + '_schema.json')
endpoint_dir = join(self.output_dir, ep.endpoint[1:])
target_dir = join(endpoint_dir, ep.partition)
common.shell_cmd('mkdir -p %s', target_dir)
index_util.dump_index(es_client,
ep.index_name,
ep.endpoint,
target_dir,
cleaner=omit_internal_keys,
query=ep.query,
chunks=ep.chunks)
# Copy the current JSON schema to the zip location so that it is included
# in the sync to s3
common.shell_cmd('cp %s %s', schema_file, endpoint_dir)
def map(self, key, value, output):
es_client = elasticsearch.Elasticsearch(config.es_host())
ep = common.ObjectDict(value)
schema_file = join(SCHEMA_DIR, ep.index_name + '_schema.json')
endpoint_dir = join(self.output_dir, ep.endpoint[1:])
target_dir = join(endpoint_dir, ep.partition)
common.shell_cmd('mkdir -p %s', target_dir)
index_util.dump_index(es_client,
ep.index_name,
ep.endpoint,
target_dir,
cleaner=omit_internal_keys,
query=ep.query,
chunks=ep.chunks)
# Copy the current JSON schema to the zip location so that it is included
# in the sync to s3
common.shell_cmd('cp %s %s', schema_file, endpoint_dir)
def map(self, key, value, output):
es_client = elasticsearch.Elasticsearch(config.es_host(), timeout=120)
ep = common.ObjectDict(value)
schema_file = join(SCHEMA_DIR, ep.index_name + '_schema.json')
endpoint_dir = join(self.output_dir, ep.endpoint[1:])
target_dir = join(endpoint_dir, ep.partition)
common.shell_cmd('mkdir -p %s', target_dir)
index_util.dump_index(es_client,
ep.index_name,
ep.endpoint,
target_dir,
cleaner=omit_internal_keys,
query=ep.query,
chunks=ep.chunks)
# Copy the current JSON schema to the zip location so that it is included
# in the sync to s3. flock is required to avoid a race condition when copying the schema file.
common.shell_cmd_quiet('flock --verbose %s cp %s %s', schema_file,
schema_file, endpoint_dir)
def run(self):
output_file = self.output().path
input_file = self.input()[1].path
es = elasticsearch.Elasticsearch(self.es_host)
index_util.start_index_transaction(es, 'recall', self.epoch)
parallel.mapreduce(
input_collection=parallel.Collection.from_sharded(input_file),
mapper=index_util.LoadJSONMapper(self.es_host,
'recall',
'enforcementreport',
self.epoch,
docid_key='@id',
version_key='@version'),
reducer=parallel.NullReducer(),
output_prefix='/tmp/loadjson.recall',
num_shards=1,
map_workers=1)
index_util.commit_index_transaction(es, 'recall')
common.shell_cmd('touch %s', output_file)
def run(self):
output_file = self.output().path
input_file = self.input()[1].path
es = elasticsearch.Elasticsearch(self.es_host)
index_util.start_index_transaction(es, 'recall', self.epoch)
parallel.mapreduce(
input_collection=parallel.Collection.from_sharded(input_file),
mapper=index_util.LoadJSONMapper(self.es_host,
'recall',
'enforcementreport',
self.epoch,
docid_key='@id',
version_key='@version'),
reducer=parallel.NullReducer(),
output_prefix='/tmp/loadjson.recall',
num_shards=1,
map_workers=1)
index_util.commit_index_transaction(es, 'recall')
common.shell_cmd('touch %s', output_file)
def _run(self):
common.shell_cmd('mkdir -p %s', self.output().path)
# Get all of the endpoints served by this index
# Create an `EndpointExport` object for each endpoint in order to export
# each endpoint properly.
#
# Endpoint exports can be:
# date range based (quarterly output)
# filter based (index serves many endpoints)
# vanilla (endpoint is 1 to 1 with index and it is exported all at once)
for endpoint, index_name in ENDPOINT_INDEX_MAP.items():
endpoint_batches = []
chunks = CUSTOM_CHUNKS.get(endpoint, DEFAULT_CHUNKS)
if endpoint in RANGE_ENDPOINT_MAP:
params = RANGE_ENDPOINT_MAP[endpoint]
params['chunks'] = chunks
endpoint_batches = _make_date_range_endpoint_batch(endpoint, params)
elif endpoint in FILTERED_ENPOINT_MAP:
params = FILTERED_ENPOINT_MAP[endpoint]
query = EndpointExport.build_term_filter(**params)
endpoint_batches.append(
EndpointExport(endpoint, query=query, chunks=chunks)
)
else:
endpoint_batches.append(EndpointExport(endpoint, chunks=chunks))
# This is a hack to overcome the shortcoming of the parallel library of
# only having one mapper process for a tiny, single file input. Since we
# want to execute these endpoint batches in parallel, we write each task
# to its own file. It will create a mapper for each file.
for ep in endpoint_batches:
partition = ep.partition if ep.partition else 'all'
if 'enforcement' in ep.endpoint:
partition = ep.endpoint.replace('enforcement', '').replace('/', '')
output_dir = join(self.output().path, index_name)
common.shell_cmd('mkdir -p %s', output_dir)
file_name = join(output_dir, partition + '.json')
with open(file_name, 'w') as json_out:
json_dict = json.dumps(ep.__dict__)
json_out.write(json_dict + '\n')
def _run(self):
output_dir = self.output().path
common.shell_cmd("mkdir -p %s", output_dir)
change_log = csv.reader(open(self.change_log_file, "r"))
batches = collections.defaultdict(list)
for row in change_log:
spl_id, spl_type, spl_date = row
# Only grab the human labels for this index
if spl_type.lower().find("human") != -1:
# All blank dates to be treated as the week of June 1, 2009
if not spl_date:
spl_date = "20090601120000"
date = arrow.get(spl_date, "YYYYMMDDHHmmss")
batches[date.ceil("week")].append(spl_id)
for batch_date, ids in batches.items():
batch_file = "%s.ids" % batch_date.format("YYYYMMDD")
batch_out = open(join(output_dir, batch_file), "w")
unique_ids = list(set(ids))
batch_out.write("\n".join(unique_ids))
def _run(self):
output_dir = self.output().path
common.shell_cmd('mkdir -p %s', output_dir)
change_log = csv.reader(open(self.change_log_file, 'r'))
batches = collections.defaultdict(list)
for row in change_log:
spl_id, spl_type, spl_date = row
# Only grab the human labels for this index
if spl_type.lower().find('human') != -1:
# All blank dates to be treated as the week of June 1, 2009
if not spl_date:
spl_date = '20090601120000'
date = arrow.get(spl_date, 'YYYYMMDDHHmmss')
batches[date.ceil('week')].append(spl_id)
for batch_date, ids in batches.items():
batch_file = '%s.ids' % batch_date.format('YYYYMMDD')
batch_out = open(join(output_dir, batch_file), 'w')
unique_ids = list(set(ids))
batch_out.write('\n'.join(unique_ids))
def run(self):
for filename in glob.glob(SPL_S3_DIR + '/*/*.xml'):
src_dir = dirname(filename)
barcode_target = join(src_dir, 'barcodes')
xml_out = join(barcode_target, 'otc-bars.xml')
json_out = xml_out.replace('.xml', '.json')
if not os.path.exists(xml_out):
common.shell_cmd('mkdir -p %s', barcode_target)
logging.info('Zbarimg on directory %s', src_dir)
cmd = 'find %(src_dir)s -name "*.jpg" -size +0\
-exec zbarimg -q --xml {} \; > \
%(xml_out)s' % locals()
os.system(cmd)
if common.is_older(json_out, xml_out):
logging.info('%s does not exist, producing...', json_out)
process_barcodes.XML2JSON(xml_out)
else:
logging.debug('%s already exists, skipping', xml_out)
common.shell_cmd('touch %s', self.output().path)
def run(self):
for filename in glob.glob(SPL_S3_DIR + '/*/*.xml'):
src_dir = dirname(filename)
barcode_target = join(src_dir, 'barcodes')
xml_out = join(barcode_target, 'otc-bars.xml')
json_out = xml_out.replace('.xml', '.json')
if not os.path.exists(xml_out):
common.shell_cmd('mkdir -p %s', barcode_target)
logging.info('Zbarimg on directory %s', src_dir)
cmd = 'find %(src_dir)s -name "*.jpg" -size +0\
-exec zbarimg -q --xml {} \; > \
%(xml_out)s' % locals()
os.system(cmd)
if common.is_older(json_out, xml_out):
logging.info('%s does not exist, producing...', json_out)
process_barcodes.XML2JSON(xml_out)
else:
logging.debug('%s already exists, skipping', xml_out)
common.shell_cmd('touch %s', self.output().path)
def _run(self):
output_dir = self.output().path
common.shell_cmd('mkdir -p %s', output_dir)
change_log = csv.reader(open(self.change_log_file, 'r'))
batches = collections.defaultdict(list)
for row in change_log:
spl_id, spl_type, spl_date = row
# Only grab the human labels for this index
if spl_type.lower().find('human') != -1:
# All blank dates to be treated as the week of June 1, 2009
if not spl_date:
spl_date = '20090601120000'
date = arrow.get(spl_date, 'YYYYMMDDHHmmss')
batches[date.ceil('week')].append(spl_id)
for batch_date, ids in batches.items():
batch_file = '%s.ids' % batch_date.format('YYYYMMDD')
batch_out = open(join(output_dir, batch_file), 'w')
unique_ids = list(set(ids))
batch_out.write('\n'.join(unique_ids))
def run(self):
output_dir = self.output().path
common.shell_cmd('mkdir -p %s', output_dir)
input_dir = self.input()[0].path
supplemental_dir = self.input()[1].path
download_util.extract_and_clean(input_dir, 'ISO-8859-1', 'UTF-8', 'txt')
# One of the files needs to be remapped from one column (submission_number)
# to two columns (pma_number and k_number) depending on the prefix.
file_name = 'registration_listing.txt'
output_file = join(output_dir, 'remapped_' + file_name)
remap_supplemental_files(join(output_dir, file_name),
join(supplemental_dir, file_name),
output_file)
# There are a handful of files with floats for keys
# This step can be removed once it is fixed on the source system.
for fix_file in self.problem_files:
with open(join(output_dir, fix_file), 'r') as needs_fixing:
lines = needs_fixing.readlines()
with open(join(output_dir, fix_file), 'w') as gets_fixing:
for line in lines:
gets_fixing.write(re.sub(r'\.0', '', line))
def run(self):
output_dir = self.output().path
common.shell_cmd('mkdir -p %s', output_dir)
for i in range(len(self.input())):
input_dir = self.input()[i].path
for zip_filename in glob.glob(input_dir + '/*.zip'):
txt_name = zip_filename.replace('zip', 'txt')
txt_name = txt_name.replace('raw', 'extracted')
common.shell_cmd('mkdir -p %s', dirname(txt_name))
cmd = 'unzip -p %s | iconv -f "ISO-8859-1//TRANSLIT" -t UTF8 -c > %s'
logging.info('Unzipping and converting %s', zip_filename)
common.shell_cmd(cmd, zip_filename, txt_name)
import arrow
import simplejson as json
from openfda.tasks import DependencyTriggeredTask
from openfda import common, config, parallel, spl
from openfda.annotation_table import unii_harmonization
from openfda.spl import process_barcodes, extract
RUN_DIR = dirname(dirname(os.path.abspath(__file__)))
data_dir = config.data_dir('harmonization')
BATCH_DATE = arrow.utcnow().ceil('week').format('YYYYMMDD')
BASE_DIR = config.data_dir('harmonization/batches/%s' % BATCH_DATE)
SPL_S3_DIR = config.data_dir('spl/s3_sync')
TMP_DIR = config.tmp_dir()
common.shell_cmd('mkdir -p %s', data_dir)
common.shell_cmd('mkdir -p %s', BASE_DIR)
common.shell_cmd('mkdir -p %s', TMP_DIR)
SPL_SET_ID_INDEX = join(BASE_DIR, 'spl_index.db')
DAILYMED_PREFIX = 'ftp://public.nlm.nih.gov/nlmdata/.dailymed/'
PHARM_CLASS_DOWNLOAD = \
DAILYMED_PREFIX + 'pharmacologic_class_indexing_spl_files.zip'
RXNORM_DOWNLOAD = \
DAILYMED_PREFIX + 'rxnorm_mappings.zip'
NDC_DOWNLOAD_PAGE = \
'http://www.fda.gov/drugs/informationondrugs/ucm142438.htm'
def run(self):
input_dir = join(self.input().path, 'events')
output_dir = self.output().path
fh_dict = {}
common.shell_cmd('mkdir -p %s', output_dir)
# Headers need to be written to the start of each partition. The patient
# and foitext headers are manually created. The headers for mdrfoi and
# foidev are detected from the source and placed into the header dictionary.
header = {}
header['patient'] = PATIENT_KEYS
header['foitext'] = TEXT_KEYS
header['mdrfoi'] = MDR_KEYS
header['foidev'] = DEVICE_KEYS
for i in range(PARTITIONS):
for category in CATEGORIES:
filename = str(i) + '.' + category + '.txt'
filename = join(output_dir, filename)
logging.info('Creating file handles for writing %s', filename)
output_handle = open(filename, 'w')
csv_writer = csv.writer(output_handle, delimiter='|')
csv_writer.writerow(header[category])
fh_dict[category + str(i)] = output_handle
# Because we download all zips from the site, we need to ignore some of the
# files for the partitioning process. Remove if files are excluded from
# download.
for filename in glob.glob(input_dir + '/*.txt'):
logging.info('Processing: %s', filename)
skip = False
for ignore in IGNORE_FILES:
if ignore in filename:
skip = True
if skip:
logging.info('Skipping: %s', filename)
continue
for category in CATEGORIES:
if category in filename:
file_category = category
# MAUDE files do not escape quote characters, we just hope that no
# pipe characters occur in records...
file_handle = csv.reader(open(filename, 'r'),
quoting=csv.QUOTE_NONE,
delimiter='|')
partioned = collections.defaultdict(list)
for i, row in enumerate(file_handle):
# skip header rows
if (i == 0) and ('MDR_REPORT_KEY' in row): continue
# Only work with rows that have data and the first column is a number
if row and row[0].isdigit():
partioned[int(row[0]) % PARTITIONS].append(row)
else:
logging.warn('Skipping row: %s', row)
for partnum, rows in partioned.iteritems():
output_handle = fh_dict[file_category + str(partnum)]
csv_writer = csv.writer(output_handle, delimiter='|')
logging.info('Writing: %s %s %s %s', partnum,
file_category + str(partnum), output_handle,
len(rows))
csv_writer.writerows(rows)
def run(self):
output_dir = self.output().path
common.shell_cmd('mkdir -p %s', output_dir)
input_dir = self.local_dir
for zip_filename in glob.glob(input_dir + '/*.zip'):
common.shell_cmd('unzip -ou %s -d %s', zip_filename, output_dir)
def run(self):
output_dir = self.output().path
common.shell_cmd('mkdir -p %s', output_dir)
input_dir = self.input().path
download_util.extract_and_clean(input_dir, 'ISO-8859-1', 'UTF-8', 'txt')
import elasticsearch
import luigi
import requests
import simplejson as json
from openfda import common, config, download_util, elasticsearch_requests, index_util, parallel
from openfda.device_harmonization.pipeline import (Harmonized2OpenFDA,
DeviceAnnotateMapper)
from openfda.device_pma import transform
from openfda.tasks import AlwaysRunTask
RUN_DIR = dirname(dirname(os.path.abspath(__file__)))
# A directory for holding files that track Task state
META_DIR = config.data_dir('device_pma/meta')
common.shell_cmd('mkdir -p %s', META_DIR)
DEVICE_PMA_ZIP = 'https://www.accessdata.fda.gov/premarket/ftparea/pma.zip'
class DownloadPMA(luigi.Task):
def requires(self):
return []
def output(self):
return luigi.LocalTarget(config.data_dir('device_pma/raw'))
def run(self):
output_filename = join(self.output().path, DEVICE_PMA_ZIP.split('/')[-1])
common.download(DEVICE_PMA_ZIP, output_filename)
class ExtractAndCleanDownloadsPMA(luigi.Task):
import csv
import re
import logging
import os
from os.path import dirname, join
import arrow
import datetime
import luigi
from openfda import common, config, index_util, parallel
RUN_DIR = dirname(dirname(os.path.abspath(__file__)))
BASE_DIR = config.data_dir('caers')
common.shell_cmd('mkdir -p %s', BASE_DIR)
S3_BUCKET = 's3://openfda-data-caers/'
S3_LOCAL_DIR = config.data_dir('caers/s3_sync')
# TODO(hansnelsen): initiate and resolve naming convention for this file and
# s3 bucket. Currently, the file is downloaded from
# s3://openfda-lonnie/caers/ (the naming of this file is
# not consistent). The pipeline engineer downloads it, renames
# it and then uploaded manually to the above bucket.
CAERS_FILE = 'caers.csv'
logging.info(S3_LOCAL_DIR, 'dir')
common.shell_cmd('mkdir -p %s', S3_LOCAL_DIR)
RENAME_MAP = {
'Report #': 'report_number',
'Created Date': 'date_created',
'Event Start Date': 'date_started',
def test_shell_cmd(self):
tmpFile = '/tmp/'+(''.join(random.choice(string.ascii_uppercase + string.digits) for _ in range(32)))
common.shell_cmd('touch %(tmpFile)s' % locals())
assert len(common.shell_cmd('ls %(tmpFile)s' % locals())) > 0
assert common.shell_cmd('ls %(tmpFile)s' % locals()).startswith(tmpFile)
def run(self):
input_dir = join(self.input().path, 'events')
output_dir = self.output().path
fh_dict = {}
common.shell_cmd('mkdir -p %s', output_dir)
# Headers need to be written to the start of each partition. The patient
# and foitext headers are manually created. The headers for mdrfoi and
# foidev are detected from the source and placed into the header dictionary.
header = {}
header['patient'] = PATIENT_KEYS
header['foitext'] = TEXT_KEYS
header['mdrfoi'] = MDR_KEYS
header['foidev'] = DEVICE_KEYS
for i in range(PARTITIONS):
for category in CATEGORIES:
filename = str(i) + '.' + category + '.txt'
filename = join(output_dir, filename)
logging.info('Creating file handles for writing %s', filename)
output_handle = open(filename, 'w')
csv_writer = csv.writer(output_handle, delimiter='|')
csv_writer.writerow(header[category])
fh_dict[category + str(i)] = output_handle
# Because we download all zips from the site, we need to ignore some of the
# files for the partitioning process. Remove if files are excluded from
# download.
for filename in glob.glob(input_dir + '/*.txt'):
logging.info('Processing: %s', filename)
skip = False
for ignore in IGNORE_FILES:
if ignore in filename:
skip = True
if skip:
logging.info('Skipping: %s', filename)
continue
for category in CATEGORIES:
if category in filename:
file_category = category
# MAUDE files do not escape quote characters, we just hope that no
# pipe characters occur in records...
file_handle = csv.reader(open(filename, 'r'),
quoting=csv.QUOTE_NONE,
delimiter='|')
partioned = collections.defaultdict(list)
for i, row in enumerate(file_handle):
# skip header rows
if (i == 0) and ('MDR_REPORT_KEY' in row): continue
# Only work with rows that have data and the first column is a number
if row and row[0].isdigit():
partioned[int(row[0]) % PARTITIONS].append(row)
else:
logging.warn('Skipping row: %s', row)
for partnum, rows in partioned.iteritems():
output_handle = fh_dict[file_category + str(partnum)]
csv_writer = csv.writer(output_handle, delimiter='|')
logging.info('Writing: %s %s %s %s',
partnum,
file_category + str(partnum),
output_handle,
len(rows))
csv_writer.writerows(rows)
from bs4 import BeautifulSoup
import elasticsearch
import luigi
import pandas
import requests
import simplejson as json
import urllib2
from openfda import common, config, elasticsearch_requests, index_util, parallel
from openfda import download_util
from openfda.index_util import AlwaysRunTask, ResetElasticSearch
from openfda.device_harmonization.pipeline import Harmonized2OpenFDA, DeviceAnnotateMapper
RUN_DIR = dirname(dirname(os.path.abspath(__file__)))
BASE_DIR = "./data/registration"
common.shell_cmd("mkdir -p %s", BASE_DIR)
# A directory for holding files that track Task state
META_DIR = join(BASE_DIR, "meta")
common.shell_cmd("mkdir -p %s", META_DIR)
DEVICE_REG_PAGE = (
"http://www.fda.gov/MedicalDevices/"
"DeviceRegulationandGuidance/HowtoMarketYourDevice/"
"RegistrationandListing/ucm134495.htm"
)
S3_BUCKET = "s3://openfda-data-reglist/"
S3_LOCAL_DIR = join(BASE_DIR, "s3_sync")
common.shell_cmd("mkdir -p %s", S3_LOCAL_DIR)
import arrow
import elasticsearch
import luigi
from openfda import common, elasticsearch_requests, index_util, parallel
from openfda.annotation_table.pipeline import CombineHarmonization
from openfda.index_util import AlwaysRunTask
from openfda.spl import annotate
from openfda.parallel import IdentityReducer
RUN_DIR = dirname(dirname(os.path.abspath(__file__)))
BASE_DIR = './data'
META_DIR = join(BASE_DIR, 'spl/meta')
# Ensure meta directory is available for task tracking
common.shell_cmd('mkdir -p %s', META_DIR)
SPL_JS = join(RUN_DIR, 'spl/spl_to_json.js')
LOINC = join(RUN_DIR, 'spl/data/sections.csv')
SPL_S3_BUCKET = 's3://openfda.spl.data/data/'
SPL_S3_LOCAL_DIR = join(BASE_DIR, 'spl/s3_sync')
SPL_S3_CHANGE_LOG = join(SPL_S3_LOCAL_DIR, 'change_log/SPLDocuments.csv')
SPL_BATCH_DIR = join(META_DIR, 'batch')
SPL_PROCESS_DIR = join(BASE_DIR, 'spl/batches')
common.shell_cmd('mkdir -p %s', SPL_S3_LOCAL_DIR)
common.shell_cmd('mkdir -p %s', SPL_PROCESS_DIR)
ES_HOST = luigi.Parameter('localhost:9200', is_global=True)
SPL_S3_PROFILE = luigi.Parameter(default='openfda', is_global=True)
import arrow
import elasticsearch
import luigi
from openfda import config, common, elasticsearch_requests, index_util, parallel
from openfda.annotation_table.pipeline import CombineHarmonization
from openfda.tasks import AlwaysRunTask
from openfda.spl import annotate
from openfda.parallel import IdentityReducer
RUN_DIR = dirname(dirname(os.path.abspath(__file__)))
META_DIR = config.data_dir("spl/meta")
# Ensure meta directory is available for task tracking
common.shell_cmd("mkdir -p %s", META_DIR)
SPL_JS = join(RUN_DIR, "spl/spl_to_json.js")
LOINC = join(RUN_DIR, "spl/data/sections.csv")
SPL_S3_BUCKET = "s3://openfda-data-spl/data/"
SPL_S3_LOCAL_DIR = config.data_dir("spl/s3_sync")
SPL_S3_CHANGE_LOG = join(SPL_S3_LOCAL_DIR, "change_log/SPLDocuments.csv")
SPL_BATCH_DIR = join(META_DIR, "batch")
SPL_PROCESS_DIR = config.data_dir("spl/batches")
common.shell_cmd("mkdir -p %s", SPL_S3_LOCAL_DIR)
common.shell_cmd("mkdir -p %s", SPL_PROCESS_DIR)
class SyncS3SPL(luigi.Task):
