def run(self):
# Running this task doesn't delete anything from CKAN itself - so require --force flag to be sent to run it
if not self.force:
raise Exception('Warning: this class does not delete CKAN records. Use --force to run it.')
# Build a dict of all modules and collections
# We then retrieve the appropriate collection from the records module name (AudTable)
# Exclude the MongoDeleteTask though
collections = {cls.module: cls(None).get_collection() for cls in MongoTask.__subclasses__()}
ke_data = KEParser(self.input().open('r'), file_path=self.input().path, schema_file=self.keemu_schema_file)
for record in self.iterate_data(ke_data):
module = record.get('AudTable')
irn = record.get('AudKey')
try:
collection = collections[module]
except KeyError:
log.debug('Skipping eaudit record for %s' % module)
# We do not have a collection for this module - skip to next record
continue
else:
log.info('Deleting record %s(%s)' % (module, irn))
self.delete(collection, irn)
self.mark_complete()
def bulk_update(self, ke_data):
bulk = self.collection.initialize_unordered_bulk_op()
count = 0
for record in self.iterate_data(ke_data):
# Find and replace doc - inserting if it doesn't exist
bulk.find({'_id': record['_id']}).upsert().replace_one(record)
count += 1
# Bulk ops can have out of memory errors (I'm getting for ~400,000+ bulk ops)
# So execute the bulk op in stages, when bulk_op_size is reached
if count % self.bulk_op_size == 0:
log.info('Executing bulk op')
bulk.execute()
bulk = self.collection.initialize_unordered_bulk_op()
try:
bulk.execute()
except InvalidOperation:
# If we do not have any records to execute, ignore error
# They have been executed in ln124
pass
def bulk_update(self, ke_data):
bulk = self.collection.initialize_unordered_bulk_op()
count = 0
for record in self.iterate_data(ke_data):
# Find and replace doc - inserting if it doesn't exist
bulk.find({'_id': record['_id']}).upsert().replace_one(record)
count += 1
# Bulk ops can have out of memory errors (I'm getting for ~400,000+ bulk ops)
# So execute the bulk op in stages, when bulk_op_size is reached
if count % self.bulk_op_size == 0:
log.info('Executing bulk op')
bulk.execute()
bulk = self.collection.initialize_unordered_bulk_op()
try:
bulk.execute()
except InvalidOperation:
# If we do not have any records to execute, ignore error
# They have been executed in ln124
pass
def main():
update_markers = mongo_get_update_markers()
# Make sure the updates have all mongo classes
bulk_tasks = [
MongoCollectionIndexTask,
MongoCollectionEventTask,
MongoCatalogueTask,
MongoTaxonomyTask,
# MongoMultimediaTask,
MongoSiteTask,
UnpublishTask,
MongoDeleteTask
]
def _get_task_names(tasks):
"""
We need to initiate and get the family name, not just the class name
MongoDeleteTask => DeleteTask
@param tasks:
@return:
"""
return [unicode(task(date=0).task_family) for task in tasks]
full_export_date = int(config.get('keemu', 'full_export_date'))
for date, update_marker in update_markers.iteritems():
# If this is the fll export date, MongoDeleteTask is not required
if full_export_date and date == full_export_date:
bulk_task_copy = list(bulk_tasks)
bulk_task_copy.remove(MongoDeleteTask)
bulk_task_names = _get_task_names(bulk_task_copy)
else:
bulk_task_names = _get_task_names(bulk_tasks)
# Assert that for every date we have all the bulk tasks
missing_tasks = list(set(bulk_task_names) - set(update_marker))
assert missing_tasks == [], 'There are missing mongo tasks for date %s: %s' % (date, missing_tasks)
# Get a list of all export files to process
export_dates = [d for d in get_export_file_dates() if d not in update_markers.keys()]
# Run setup_interface_logging to ensure luigi commands
setup_interface_logging()
sch = scheduler.CentralPlannerScheduler()
w = BulkWorker(scheduler=sch)
for export_date in export_dates:
log.info('Processing date %s', export_date)
# We only need to call the mongo delete task, as all other tasks are a requirement
# NB: This doesn't delete anything from CKAN - if that's needed change this to DeleteTask
w.add(MongoDeleteTask(date=export_date, force=True))
w.run()
w.stop()
def run(self):
mongo_db = mongo_client_db()
collection = MongoCatalogueTask(date=None).collection_name
cites_species = get_cites_species()
# Set cites=true flag
cites_records_cursor = mongo_db[collection].update({'DarScientificName': {'$in': cites_species}}, {'$set': {'cites': True}}, multi=True)
log.info('Updated %s catalogue records as CITES', cites_records_cursor['nModified'])
def on_success(self):
"""
On completion, add indexes
@return: None
"""
self.collection = self.get_collection()
log.info("Adding exportFileDate index")
self.collection.ensure_index('exportFileDate')
def on_success(self):
"""
On completion, add indexes
@return: None
"""
self.collection = self.get_collection()
log.info("Adding exportFileDate index")
self.collection.ensure_index('exportFileDate')
def write(self, df):
log.info("Saving records to CKAN resource %s", self.resource_id)
# Convert all empty/null values to None - so will be NULL values in postgres
# Ensure any float fields with value 0.0 are actually None
for col, np_type in self.columns.iteritems():
if np_type.startswith('float'):
df[col][df[col] == 0.0] = None
else:
# BUGFIX: Multimedia fields are being populated with empty string rather than NULL
df[col][df[col].astype(str) == ''] = None
# Loop through all the dataframe columns, removing internal ones (fields starting with _)
for col in df:
if col.startswith('_'):
df.drop(col, axis=1, inplace=True)
# Convert all NaN to None
df = df.where(pd.notnull(df), None)
# Convert records to dictionary
records = df.to_dict(outtype='records')
datastore_params = {
'resource_id': self.resource_id,
'records': records,
'force': True
# 'primary_key': '_id'
}
# Check that the data doesn't contain invalid chars
try:
json.dumps(datastore_params).encode('ascii')
except UnicodeDecodeError:
# At least one of the records contain invalid chars
# Loop through, validating each of the records
validated_records = []
for i, record in enumerate(datastore_params['records']):
try:
json.dumps(record).encode('ascii')
except UnicodeDecodeError:
log.critical('Error encoding record: %s', ' '.join(['%s=%s' % (field, value) for field, value in record.iteritems() if value]))
else:
validated_records.append(record)
datastore_params['records'] = validated_records
self.remote_ckan.action.datastore_upsert(**datastore_params)
def ckan_delete(remote_ckan, mongo_record):
# To avoid circular imports, import the tasks we need to check here
# Dataset tasks are dependent on the DeleteTask
from ke2mongo.tasks.indexlot import IndexLotDatasetAPITask
from ke2mongo.tasks.artefact import ArtefactDatasetAPITask
from ke2mongo.tasks.specimen import SpecimenDatasetAPITask
# By default, use SpecimenDatasetAPITask
task_cls = SpecimenDatasetAPITask
# Override default class if is Index Lot or Artefact
for t in [IndexLotDatasetAPITask, ArtefactDatasetAPITask]:
if t.record_type == mongo_record['ColRecordType']:
task_cls = t
break
# Get the primary key
for col in task_cls.columns:
if col[1] == task_cls.datastore['primary_key']:
primary_key_field = col
break
# Get the source primary key - this needs to be split on . as we have added the collection name
ke_primary_key = primary_key_field[0].split('.')[1]
# The name of the primary key field used in CKAN
ckan_primary_key = primary_key_field[1]
try:
primary_key_value = mongo_record[ke_primary_key]
except KeyError:
log.error('No value for primary key %s', ke_primary_key)
else:
resource_id = get_resource_id(remote_ckan, task_cls.package['name'])
if resource_id:
try:
# And delete the record from the datastore
log.info('Deleting record from CKAN where %s=%s' %
(ckan_primary_key, primary_key_value))
remote_ckan.action.datastore_delete(
id=resource_id,
filters={ckan_primary_key: primary_key_value},
force=True)
except ckanapi.CKANAPIError:
# We don't care if the record isn't found
log.error('Record not found')
else:
log.error('No resource ID')
def delete(self, collection, irn):
# If this is an ecatalogue record, try and delete from CKAN
if collection.name == 'ecatalogue':
# Load the record from mongo
mongo_record = collection.find_one({'_id': int(irn)})
if mongo_record:
ckan_delete(self.remote_ckan, mongo_record)
else:
log.info('Record %s does not exist. SKipping delete.' % irn)
# And call the Mongo Delete task delete() method to remove the record from mongodb
super(DeleteAPITask, self).delete(collection, irn)
def run(self):
mongo_db = mongo_client_db()
collection = MongoCatalogueTask(date=None).collection_name
cites_species = get_cites_species()
# Set cites=true flag
cites_records_cursor = mongo_db[collection].update(
{'DarScientificName': {
'$in': cites_species
}}, {'$set': {
'cites': True
}},
multi=True)
log.info('Updated %s catalogue records as CITES',
cites_records_cursor['nModified'])
def get_or_create_resource(self):
"""
Either load a resource object
Or if it doesn't exist, create the dataset package, and datastore
@param package: params to create the package
@param datastore: params to create the datastore
@return: CKAN resource ID
"""
resource_id = None
try:
# If the package exists, retrieve the resource
ckan_package = self.remote_ckan.action.package_show(id=self.package['name'])
# Does a resource of the same name already exist for this dataset?
# If it does, assign to resource_id
for resource in ckan_package['resources']:
if resource['name'] == self.datastore['resource']['name']:
self.validate_resource(resource)
resource_id = resource['id']
except ckanapi.NotFound:
log.info("Package %s not found - creating", self.package['name'])
# Create the package
ckan_package = self.remote_ckan.action.package_create(**self.package)
# If we don't have the resource ID, create
if not resource_id:
log.info("Resource %s not found - creating", self.datastore['resource']['name'])
self.datastore['fields'] = [{'id': col, 'type': self.numpy_to_ckan_type(np_type)} for col, np_type in self.get_output_columns().iteritems()]
self.datastore['resource']['package_id'] = ckan_package['id']
if self.indexed_fields:
# Create BTREE indexes for all specified indexed fields
self.datastore['indexes'] = [col['id'] for col in self.datastore['fields'] if col['id'] in self.indexed_fields]
else:
# Create BTREE indexes for all citext fields
self.datastore['indexes'] = [col['id'] for col in self.datastore['fields'] if col['type'] == 'citext']
# API call to create the datastore
resource_id = self.remote_ckan.action.datastore_create(**self.datastore)['resource_id']
# If this has geospatial fields, create geom columns
if self.geospatial_fields:
log.info("Creating geometry columns for %s", resource_id)
self.geospatial_fields['resource_id'] = resource_id
self.remote_ckan.action.create_geom_columns(**self.geospatial_fields)
log.info("Created datastore resource %s", resource_id)
return resource_id
def ckan_delete(remote_ckan, mongo_record):
# To avoid circular imports, import the tasks we need to check here
# Dataset tasks are dependent on the DeleteTask
from ke2mongo.tasks.indexlot import IndexLotDatasetAPITask
from ke2mongo.tasks.artefact import ArtefactDatasetAPITask
from ke2mongo.tasks.specimen import SpecimenDatasetAPITask
# By default, use SpecimenDatasetAPITask
task_cls = SpecimenDatasetAPITask
# Override default class if is Index Lot or Artefact
for t in [IndexLotDatasetAPITask, ArtefactDatasetAPITask]:
if t.record_type == mongo_record['ColRecordType']:
task_cls = t
break
# Get the primary key
for col in task_cls.columns:
if col[1] == task_cls.datastore['primary_key']:
primary_key_field = col
break
# Get the source primary key - this needs to be split on . as we have added the collection name
ke_primary_key = primary_key_field[0].split('.')[1]
# The name of the primary key field used in CKAN
ckan_primary_key = primary_key_field[1]
try:
primary_key_value = mongo_record[ke_primary_key]
except KeyError:
log.error('No value for primary key %s', ke_primary_key)
else:
resource_id = get_resource_id(remote_ckan, task_cls.package['name'])
if resource_id:
try:
# And delete the record from the datastore
log.info('Deleting record from CKAN where %s=%s' % (ckan_primary_key, primary_key_value))
remote_ckan.action.datastore_delete(id=resource_id, filters={ckan_primary_key: primary_key_value}, force=True)
except ckanapi.CKANAPIError:
# We don't care if the record isn't found
log.error('Record not found')
else:
log.error('No resource ID')
def on_success(self):
"""
On completion, add indexes
@return: None
"""
self.collection = self.get_collection()
log.info("Adding ecatalogue indexes")
self.collection.ensure_index('ColRecordType')
# Only include active records - not Stubs etc.,
self.collection.ensure_index('SecRecordStatus')
# Add index on RegRegistrationParentRef - select records with the same parent
self.collection.ensure_index('RegRegistrationParentRef')
# Need to filter on web publishable
self.collection.ensure_index('AdmPublishWebNoPasswordFlag')
# Exclude records if they do not have a GUID
self.collection.ensure_index('AdmGUIDPreferredValue')
# Add embargo date index
self.collection.ensure_index('RealEmbargoDate')
super(MongoCatalogueTask, self).on_success()
def run(self):
# Do not run if this is a full export date - all non-publishable records will
# Already have been removed
if int(self.full_export_date) == int(self.date):
log.info("No records to unpublish for full exports")
self.mark_complete()
return
collection = self.output().get_collection('ecatalogue')
# We only care about records who's status has changed in the past week (6 days to be sure)
date_object = datetime.strptime(str(self.date), '%Y%m%d')
q = dict(AdmPublishWebNoPasswordFlag='N',
exportFileDate=self.date,
ISODateInserted={'$gte': date_object - timedelta(days=6)})
cursor = collection.find(q)
log.info('%s records to unpublish', cursor.count())
for record in cursor:
ckan_delete(self.remote_ckan, record)
# And mark the object as complete
self.mark_complete()
def run(self):
count = 0
host = config.get('mongo', 'host')
db = config.get('mongo', 'database')
def _fill_field(field_arr, field_type):
if field_type.startswith('string'):
field_arr = field_arr.astype(np.str).filled('')
elif field_type == 'bool':
field_arr = field_arr.astype(np.str).filled(None)
elif field_type.startswith('int'):
field_arr = field_arr.filled(0)
elif field_type.startswith('float'):
field_arr = field_arr.filled(np.NaN)
else:
raise Exception('Unknown field type %s' % field_type)
return field_arr
with Monary(host) as m:
log.info("Querying Monary")
# Get field definitions for default collection
query_fields, df_cols, field_types = zip(*self.get_collection_source_columns(self.collection_name))
catalogue_blocks = m.block_query(db, self.collection_name, self.query, query_fields, field_types, block_size=self.block_size)
log.info("Processing Monary data")
for catalogue_block in catalogue_blocks:
# Bit of a hack: fill fields with a blank value (depending on type)
# So the masked value doesn't get used. As the masked is shared between
# each block, if a field is empty it is getting populated by previous values
catalogue_block = [_fill_field(arr, field_types[i]) for i, arr in enumerate(catalogue_block)]
# Create a pandas data frame with block of records
# Columns use the name from the output columns - but must be in the same order as query_fields
# Which is why we're using tuples for the columns
df = pd.DataFrame(np.matrix(catalogue_block).transpose(), columns=df_cols)
# Loop through all the columns and ensure hidden integer fields are cast as int32
# For example, taxonomy_irn is used to join with taxonomy df
for i, df_col in enumerate(df_cols):
if field_types[i].startswith('int'):
df[df_col] = df[df_col].astype(field_types[i])
df = self.process_dataframe(m, df)
# Output the dataframe
self.output().write(df)
row_count, col_count = df.shape
count += row_count
log.info("\t %s records", count)
# After running, update mongo
self.mongo_target.touch()
def on_success(self):
log.info("Import CSV file with:")
log.info("COPY \"{resource_id}\" (\"{cols}\") FROM '{path}' DELIMITER ',' CSV ENCODING 'UTF8';".format(
resource_id=self.resource_id,
cols='","'.join(col for col in self.get_output_columns()),
path=self.path
))
log.info("And update full text index:")
log.info("paster update-fulltext -i \"{resource_id}\" -c /vagrant/etc/default/development.ini".format(
resource_id=self.resource_id,
))
return super(DatasetCSVTask, self).complete()
def on_success(self):
log.info("Import CSV file with:")
log.info(
"COPY \"{resource_id}\" (\"{cols}\") FROM '{path}' DELIMITER ',' CSV ENCODING 'UTF8';"
.format(resource_id=self.resource_id,
cols='","'.join(col for col in self.get_output_columns()),
path=self.path))
log.info("And update full text index:")
log.info(
"paster update-fulltext -i \"{resource_id}\" -c /vagrant/etc/default/development.ini"
.format(resource_id=self.resource_id, ))
return super(DatasetCSVTask, self).complete()
def batch_insert(self, ke_data):
def _insert(batch):
try:
self.collection.insert(batch)
except DuplicateKeyError:
# Duplicate key error - KE export does duplicate some records
# So switch to bulk upsert for this operation
log.error('Duplicate key error - switching to upsert')
bulk = self.collection.initialize_unordered_bulk_op()
for batch_record in batch:
bulk.find({
'_id': batch_record['_id']
}).upsert().replace_one(batch_record)
bulk.execute()
batch = []
for record in self.iterate_data(ke_data):
if self.batch_size:
batch.append(record)
# If the batch length equals the batch size, commit and clear the batch
if len(batch) % self.batch_size == 0:
log.info('Submitting batch')
_insert(batch)
batch = []
else:
self.collection.insert(record)
# Add any records remaining in the batch
if batch:
_insert(batch)
def run(self):
# Do not run if this is a full export date - all non-publishable records will
# Already have been removed
if int(self.full_export_date) == int(self.date):
log.info("No records to unpublish for full exports")
self.mark_complete()
return
collection = self.output().get_collection('ecatalogue')
# We only care about records who's status has changed in the past week (6 days to be sure)
date_object = datetime.strptime(str(self.date), '%Y%m%d')
q = dict(
AdmPublishWebNoPasswordFlag='N',
exportFileDate=self.date,
ISODateInserted={'$gte': date_object - timedelta(days=6)}
)
cursor = collection.find(q)
log.info('%s records to unpublish', cursor.count())
for record in cursor:
ckan_delete(self.remote_ckan, record)
# And mark the object as complete
self.mark_complete()
def batch_insert(self, ke_data):
def _insert(batch):
try:
self.collection.insert(batch)
except DuplicateKeyError:
# Duplicate key error - KE export does duplicate some records
# So switch to bulk upsert for this operation
log.error('Duplicate key error - switching to upsert')
bulk = self.collection.initialize_unordered_bulk_op()
for batch_record in batch:
bulk.find({'_id': batch_record['_id']}).upsert().replace_one(batch_record)
bulk.execute()
batch = []
for record in self.iterate_data(ke_data):
if self.batch_size:
batch.append(record)
# If the batch length equals the batch size, commit and clear the batch
if len(batch) % self.batch_size == 0:
log.info('Submitting batch')
_insert(batch)
batch = []
else:
self.collection.insert(record)
# Add any records remaining in the batch
if batch:
_insert(batch)
def iterate_data(self, ke_data):
"""
Iterate through the data
@return:
"""
for record in ke_data:
status = ke_data.get_status()
if status:
log.info(status)
# Use the IRN as _id
record['_id'] = record['irn']
try:
# Do not process if unprocessed flag is set
if not self.unprocessed:
record = self.process_record(record)
except InvalidRecordException:
continue
else:
yield record
def run(self):
# Running this task doesn't delete anything from CKAN itself - so require --force flag to be sent to run it
if not self.force:
raise Exception(
'Warning: this class does not delete CKAN records. Use --force to run it.'
)
# Build a dict of all modules and collections
# We then retrieve the appropriate collection from the records module name (AudTable)
# Exclude the MongoDeleteTask though
collections = {
cls.module: cls(None).get_collection()
for cls in MongoTask.__subclasses__()
}
ke_data = KEParser(self.input().open('r'),
file_path=self.input().path,
schema_file=self.keemu_schema_file)
for record in self.iterate_data(ke_data):
module = record.get('AudTable')
irn = record.get('AudKey')
try:
collection = collections[module]
except KeyError:
log.debug('Skipping eaudit record for %s' % module)
# We do not have a collection for this module - skip to next record
continue
else:
log.info('Deleting record %s(%s)' % (module, irn))
self.delete(collection, irn)
self.mark_complete()
def iterate_data(self, ke_data):
"""
Iterate through the data
@return:
"""
for record in ke_data:
status = ke_data.get_status()
if status:
log.info(status)
# Use the IRN as _id
record['_id'] = record['irn']
try:
# Do not process if unprocessed flag is set
if not self.unprocessed:
record = self.process_record(record)
except InvalidRecordException:
continue
else:
yield record
def run(self):
if int(self.full_export_date) == int(self.date):
log.info("No records to delete for full exports")
self.mark_complete()
return
super(DeleteAPITask, self).run()
def get_or_create_resource(self):
"""
Either load a resource object
Or if it doesn't exist, create the dataset package, and datastore
@param package: params to create the package
@param datastore: params to create the datastore
@return: CKAN resource ID
"""
resource_id = None
try:
# If the package exists, retrieve the resource
ckan_package = self.remote_ckan.action.package_show(
id=self.package['name'])
# Does a resource of the same name already exist for this dataset?
# If it does, assign to resource_id
for resource in ckan_package['resources']:
if resource['name'] == self.datastore['resource']['name']:
self.validate_resource(resource)
resource_id = resource['id']
except ckanapi.NotFound:
log.info("Package %s not found - creating", self.package['name'])
# Create the package
ckan_package = self.remote_ckan.action.package_create(
**self.package)
# If we don't have the resource ID, create
if not resource_id:
log.info("Resource %s not found - creating",
self.datastore['resource']['name'])
self.datastore['fields'] = [{
'id':
col,
'type':
self.numpy_to_ckan_type(np_type)
} for col, np_type in self.get_output_columns().iteritems()]
self.datastore['resource']['package_id'] = ckan_package['id']
if self.indexed_fields:
# Create BTREE indexes for all specified indexed fields
self.datastore['indexes'] = [
col['id'] for col in self.datastore['fields']
if col['id'] in self.indexed_fields
]
else:
# Create BTREE indexes for all citext fields
self.datastore['indexes'] = [
col['id'] for col in self.datastore['fields']
if col['type'] == 'citext'
]
# API call to create the datastore
resource_id = self.remote_ckan.action.datastore_create(
**self.datastore)['resource_id']
# If this has geospatial fields, create geom columns
if self.geospatial_fields:
log.info("Creating geometry columns for %s", resource_id)
self.geospatial_fields['resource_id'] = resource_id
self.remote_ckan.action.create_geom_columns(
**self.geospatial_fields)
log.info("Created datastore resource %s", resource_id)
return resource_id
def main():
update_markers = mongo_get_update_markers()
# Make sure the updates have all mongo classes
bulk_tasks = [
MongoCollectionIndexTask,
MongoCollectionEventTask,
MongoCatalogueTask,
MongoTaxonomyTask,
# MongoMultimediaTask,
MongoSiteTask,
UnpublishTask,
MongoDeleteTask
]
def _get_task_names(tasks):
"""
We need to initiate and get the family name, not just the class name
MongoDeleteTask => DeleteTask
@param tasks:
@return:
"""
return [unicode(task(date=0).task_family) for task in tasks]
full_export_date = int(config.get('keemu', 'full_export_date'))
for date, update_marker in update_markers.iteritems():
# If this is the fll export date, MongoDeleteTask is not required
if full_export_date and date == full_export_date:
bulk_task_copy = list(bulk_tasks)
bulk_task_copy.remove(MongoDeleteTask)
bulk_task_names = _get_task_names(bulk_task_copy)
else:
bulk_task_names = _get_task_names(bulk_tasks)
# Assert that for every date we have all the bulk tasks
missing_tasks = list(set(bulk_task_names) - set(update_marker))
assert missing_tasks == [], 'There are missing mongo tasks for date %s: %s' % (
date, missing_tasks)
# Get a list of all export files to process
export_dates = [
d for d in get_export_file_dates() if d not in update_markers.keys()
]
# Run setup_interface_logging to ensure luigi commands
setup_interface_logging()
sch = scheduler.CentralPlannerScheduler()
w = BulkWorker(scheduler=sch)
for export_date in export_dates:
log.info('Processing date %s', export_date)
# We only need to call the mongo delete task, as all other tasks are a requirement
# NB: This doesn't delete anything from CKAN - if that's needed change this to DeleteTask
w.add(MongoDeleteTask(date=export_date, force=True))
w.run()
w.stop()
def run(self):
count = 0
host = config.get('mongo', 'host')
db = config.get('mongo', 'database')
def _fill_field(field_arr, field_type):
if field_type.startswith('string'):
field_arr = field_arr.astype(np.str).filled('')
elif field_type == 'bool':
field_arr = field_arr.astype(np.str).filled(None)
elif field_type.startswith('int'):
field_arr = field_arr.filled(0)
elif field_type.startswith('float'):
field_arr = field_arr.filled(np.NaN)
else:
raise Exception('Unknown field type %s' % field_type)
return field_arr
with Monary(host) as m:
log.info("Querying Monary")
# Get field definitions for default collection
query_fields, df_cols, field_types = zip(
*self.get_collection_source_columns(self.collection_name))
catalogue_blocks = m.block_query(db,
self.collection_name,
self.query,
query_fields,
field_types,
block_size=self.block_size)
log.info("Processing Monary data")
for catalogue_block in catalogue_blocks:
# Bit of a hack: fill fields with a blank value (depending on type)
# So the masked value doesn't get used. As the masked is shared between
# each block, if a field is empty it is getting populated by previous values
catalogue_block = [
_fill_field(arr, field_types[i])
for i, arr in enumerate(catalogue_block)
]
# Create a pandas data frame with block of records
# Columns use the name from the output columns - but must be in the same order as query_fields
# Which is why we're using tuples for the columns
df = pd.DataFrame(np.matrix(catalogue_block).transpose(),
columns=df_cols)
# Loop through all the columns and ensure hidden integer fields are cast as int32
# For example, taxonomy_irn is used to join with taxonomy df
for i, df_col in enumerate(df_cols):
if field_types[i].startswith('int'):
df[df_col] = df[df_col].astype(field_types[i])
df = self.process_dataframe(m, df)
# Output the dataframe
self.output().write(df)
row_count, col_count = df.shape
count += row_count
log.info("\t %s records", count)
# After running, update mongo
self.mongo_target.touch()
