def process(self, item):
response = self.get_action('package_show', id=item.remote_id)
data = self.validate(response['result'], schema)
# Fix the remote_id: use real ID instead of not stable name
item.remote_id = data['id']
# Skip if no resource
if not len(data.get('resources', [])):
msg = 'Dataset {0} has no record'.format(item.remote_id)
raise HarvestSkipException(msg)
dataset = self.get_dataset(item.remote_id)
# Core attributes
if not dataset.slug:
dataset.slug = data['name']
dataset.title = data['title']
dataset.description = data['notes']
dataset.license = License.objects(id=data['license_id']).first()
# dataset.license = license or License.objects.get(id='notspecified')
dataset.tags = [t['name'] for t in data['tags'] if t['name']]
dataset.created_at = data['metadata_created']
dataset.last_modified = data['metadata_modified']
dataset.extras['ckan:name'] = data['name']
temporal_start, temporal_end = None, None
spatial_geom = None
for extra in data['extras']:
# GeoJSON representation (Polygon or Point)
if extra['key'] == 'spatial':
spatial_geom = json.loads(extra['value'])
# Textual representation of the extent / location
elif extra['key'] == 'spatial-text':
log.debug('spatial-text value not handled')
print 'spatial-text', extra['value']
# Linked Data URI representing the place name
elif extra['key'] == 'spatial-uri':
log.debug('spatial-uri value not handled')
print 'spatial-uri', extra['value']
# Update frequency
elif extra['key'] == 'frequency':
print 'frequency', extra['value']
# Temporal coverage start
elif extra['key'] == 'temporal_start':
print 'temporal_start', extra['value']
temporal_start = daterange_start(extra['value'])
continue
# Temporal coverage end
elif extra['key'] == 'temporal_end':
print 'temporal_end', extra['value']
temporal_end = daterange_end(extra['value'])
continue
# else:
# print extra['key'], extra['value']
dataset.extras[extra['key']] = extra['value']
if spatial_geom:
dataset.spatial = SpatialCoverage()
if spatial_geom['type'] == 'Polygon':
coordinates = [spatial_geom['coordinates']]
elif spatial_geom['type'] == 'MultiPolygon':
coordinates = spatial_geom['coordinates']
else:
HarvestException('Unsupported spatial geometry')
dataset.spatial.geom = {
'type': 'MultiPolygon',
'coordinates': coordinates
}
if temporal_start and temporal_end:
dataset.temporal_coverage = db.DateRange(
start=temporal_start,
end=temporal_end,
)
# Remote URL
if data.get('url'):
dataset.extras['remote_url'] = data['url']
# Resources
for res in data['resources']:
if res['resource_type'] not in ALLOWED_RESOURCE_TYPES:
continue
try:
resource = get_by(dataset.resources, 'id', UUID(res['id']))
except:
log.error('Unable to parse resource ID %s', res['id'])
continue
if not resource:
resource = Resource(id=res['id'])
dataset.resources.append(resource)
resource.title = res.get('name', '') or ''
resource.description = res.get('description')
resource.url = res['url']
resource.filetype = ('api' if res['resource_type'] == 'api'
else 'remote')
resource.format = res.get('format')
resource.mime = res.get('mimetype')
resource.hash = res.get('hash')
resource.created = res['created']
resource.modified = res['last_modified']
resource.published = resource.published or resource.created
return dataset
def license(self):
return License.objects(id=self.license_id).first()
def process(self, item):
response = self.get_action('package_show', id=item.remote_id)
data = self.validate(response['result'], schema)
# Fix the remote_id: use real ID instead of not stable name
item.remote_id = data['id']
# Skip if no resource
if not len(data.get('resources', [])):
msg = 'Dataset {0} has no record'.format(item.remote_id)
raise HarvestSkipException(msg)
dataset = self.get_dataset(item.remote_id)
# Core attributes
if not dataset.slug:
dataset.slug = data['name']
dataset.title = data['title']
dataset.description = data['notes']
dataset.license = License.objects(id=data['license_id']).first()
# dataset.license = license or License.objects.get(id='notspecified')
dataset.tags = [t['name'] for t in data['tags']]
dataset.created_at = data['metadata_created']
dataset.last_modified = data['metadata_modified']
dataset.extras['ckan:name'] = data['name']
temporal_start, temporal_end = None, None
spatial_geom = None
for extra in data['extras']:
# GeoJSON representation (Polygon or Point)
if extra['key'] == 'spatial':
spatial_geom = json.loads(extra['value'])
# Textual representation of the extent / location
elif extra['key'] == 'spatial-text':
log.debug('spatial-text value not handled')
print 'spatial-text', extra['value']
# Linked Data URI representing the place name
elif extra['key'] == 'spatial-uri':
log.debug('spatial-uri value not handled')
print 'spatial-uri', extra['value']
# Update frequency
elif extra['key'] == 'frequency':
print 'frequency', extra['value']
# Temporal coverage start
elif extra['key'] == 'temporal_start':
print 'temporal_start', extra['value']
temporal_start = daterange_start(extra['value'])
continue
# Temporal coverage end
elif extra['key'] == 'temporal_end':
print 'temporal_end', extra['value']
temporal_end = daterange_end(extra['value'])
continue
# else:
# print extra['key'], extra['value']
dataset.extras[extra['key']] = extra['value']
if spatial_geom:
dataset.spatial = SpatialCoverage()
if spatial_geom['type'] == 'Polygon':
coordinates = [spatial_geom['coordinates']]
elif spatial_geom['type'] == 'MultiPolygon':
coordinates = spatial_geom['coordinates']
else:
HarvestException('Unsupported spatial geometry')
dataset.spatial.geom = {
'type': 'MultiPolygon',
'coordinates': coordinates
}
if temporal_start and temporal_end:
dataset.temporal_coverage = db.DateRange(
start=temporal_start,
end=temporal_end,
)
# Remote URL
if data.get('url'):
dataset.extras['remote_url'] = data['url']
# Resources
for res in data['resources']:
if res['resource_type'] not in ALLOWED_RESOURCE_TYPES:
continue
try:
resource = get_by(dataset.resources, 'id', UUID(res['id']))
except:
log.error('Unable to parse resource ID %s', res['id'])
continue
if not resource:
resource = Resource(id=res['id'])
dataset.resources.append(resource)
resource.title = res.get('name', '') or ''
resource.description = res.get('description')
resource.url = res['url']
resource.filetype = ('api' if res['resource_type'] == 'api'
else 'remote')
resource.format = res.get('format')
resource.mime = res.get('mimetype')
resource.hash = res.get('hash')
resource.created = res['created']
resource.modified = res['last_modified']
resource.published = resource.published or resource.created
return dataset
def license(self):
return License.objects(id=self.license_id).first()
def remote_datasets(self):
response = self.get('package_list')
for name in response['result']:
details = self.get('package_show', {'id': name})['result']
dataset = self.get_harvested(Dataset, details['id'])
# Core attributes
dataset.slug = details['name']
dataset.title = details['title']
dataset.description = details.get('notes', 'No description')
dataset.license = License.objects(id=details['license_id']).first() or License.objects.get(id='notspecified')
dataset.tags = [tag['name'].lower() for tag in details['tags']]
dataset.frequency = self.map('frequency', details) or 'unknown'
dataset.created_at = parse(details['metadata_created'])
dataset.last_modified = parse(details['metadata_modified'])
if any_field(details, 'territorial_coverage', 'territorial_coverage_granularity'):
coverage = TerritorialCoverage(
codes=[code.strip() for code in details.get('territorial_coverage', '').split(',') if code.strip()],
granularity=self.map('territorial_coverage_granularity', details),
)
dataset.extras['territorial_coverage'] = coverage
try:
dataset.spatial = territorial_to_spatial(dataset)
except Exception as e:
print 'Error while processing spatial coverage for {0}:'.format(dataset.title), e
if all_field(details, 'temporal_coverage_from', 'temporal_coverage_to'):
try:
dataset.temporal_coverage = db.DateRange(
start=daterange_start(details.get('temporal_coverage_from')),
end=daterange_end(details.get('temporal_coverage_to')),
)
except:
log.error('Unable to parse temporal coverage for dataset %s', details['id'])
# Organization
if details.get('organization'):
dataset.organization = self.get_harvested(Organization, details['organization']['id'], False)
else:
# Need to fetch user from roles
roles = self.get('roles_show', {'domain_object': name})['result']['roles']
for role in roles:
if role['role'] == 'admin' and role['context'] == 'Package':
dataset.owner = self.get_harvested(User, role['user_id'])
break
# Supplier
if details.get('supplier_id'):
dataset.supplier = self.get_harvested(Organization, details['supplier_id'], False)
# Remote URL
if details.get('url'):
dataset.extras['remote_url'] = details['url']
# Extras
if 'extras' in details:
extra_mapping = self.harvester.mapping.get('from_extras', {})
for extra in details['extras']:
if extra['key'] in self.harvester.mapping:
value = self.harvester.mapping[extra['key']].get(extra['value'])
else:
value = extra['value']
if extra['key'] in extra_mapping:
setattr(dataset, extra_mapping[extra['key']], value)
else:
dataset.extras[extra['key']] = value
# Resources
for res in details['resources']:
try:
resource = get_by(dataset.resources, 'id', UUID(res['id']))
except:
log.error('Unable to parse resource %s', res['id'])
continue
if not resource:
resource = Resource(id=res['id'])
dataset.resources.append(resource)
resource.title = res.get('name', '') or ''
resource.url = res['url']
resource.description = res.get('description')
resource.format = res.get('format')
resource.hash = res.get('hash')
resource.created = parse(res['created'])
resource.modified = parse(res['revision_timestamp'])
resource.published = resource.published or resource.created
yield dataset
if dataset.id:
followers = self.get('dataset_follower_list', {'id': name})['result']
for follower in followers:
user = self.get_harvested(User, follower['id'], False)
if user:
follow, created = FollowDataset.objects.get_or_create(follower=user, following=dataset)
def remote_datasets(self):
response = self.get('package_list')
for name in response['result']:
details = self.get('package_show', {'id': name})['result']
dataset = self.get_harvested(Dataset, details['id'])
# Core attributes
dataset.slug = details['name']
dataset.title = details['title']
dataset.description = details.get('notes', 'No description')
dataset.license = License.objects(
id=details['license_id']).first() or License.objects.get(
id='notspecified')
dataset.tags = [tag['name'].lower() for tag in details['tags']]
dataset.frequency = self.map('frequency', details) or 'unknown'
dataset.created_at = parse(details['metadata_created'])
dataset.last_modified = parse(details['metadata_modified'])
if any_field(details, 'territorial_coverage',
'territorial_coverage_granularity'):
coverage = TerritorialCoverage(
codes=[
code.strip() for code in details.get(
'territorial_coverage', '').split(',')
if code.strip()
],
granularity=self.map('territorial_coverage_granularity',
details),
)
dataset.extras['territorial_coverage'] = coverage
try:
dataset.spatial = territorial_to_spatial(dataset)
except Exception as e:
print 'Error while processing spatial coverage for {0}:'.format(
dataset.title), e
if all_field(details, 'temporal_coverage_from',
'temporal_coverage_to'):
try:
dataset.temporal_coverage = db.DateRange(
start=daterange_start(
details.get('temporal_coverage_from')),
end=daterange_end(details.get('temporal_coverage_to')),
)
except:
log.error(
'Unable to parse temporal coverage for dataset %s',
details['id'])
# Organization
if details.get('organization'):
dataset.organization = self.get_harvested(
Organization, details['organization']['id'], False)
else:
# Need to fetch user from roles
roles = self.get('roles_show',
{'domain_object': name})['result']['roles']
for role in roles:
if role['role'] == 'admin' and role['context'] == 'Package':
dataset.owner = self.get_harvested(
User, role['user_id'])
break
# Supplier
if details.get('supplier_id'):
dataset.supplier = self.get_harvested(Organization,
details['supplier_id'],
False)
# Remote URL
if details.get('url'):
dataset.extras['remote_url'] = details['url']
# Extras
if 'extras' in details:
extra_mapping = self.harvester.mapping.get('from_extras', {})
for extra in details['extras']:
if extra['key'] in self.harvester.mapping:
value = self.harvester.mapping[extra['key']].get(
extra['value'])
else:
value = extra['value']
if extra['key'] in extra_mapping:
setattr(dataset, extra_mapping[extra['key']], value)
else:
dataset.extras[extra['key']] = value
# Resources
for res in details['resources']:
try:
resource = get_by(dataset.resources, 'id', UUID(res['id']))
except:
log.error('Unable to parse resource %s', res['id'])
continue
if not resource:
resource = Resource(id=res['id'])
dataset.resources.append(resource)
resource.title = res.get('name', '') or ''
resource.url = res['url']
resource.description = res.get('description')
resource.format = res.get('format')
resource.hash = res.get('hash')
resource.created = parse(res['created'])
resource.modified = parse(res['revision_timestamp'])
resource.published = resource.published or resource.created
yield dataset
if dataset.id:
followers = self.get('dataset_follower_list',
{'id': name})['result']
for follower in followers:
user = self.get_harvested(User, follower['id'], False)
if user:
follow, created = FollowDataset.objects.get_or_create(
follower=user, following=dataset)
