def test_pull_datapackage(storage, descriptor):
# Prepare and call
storage.buckets = ['data___data']
storage.describe.return_value = (
{'fields': [
{'name': 'id', 'type': 'integer'},
{'name': 'city', 'type': 'string'}]})
storage.read.return_value = [
(1, 'London'),
(2, 'Paris'),
]
module.pull_datapackage(descriptor=descriptor, name='name', backend='backend')
# Assert pulled datapackage
dp = DataPackage(descriptor)
assert dp.descriptor == helpers.expand_data_package_descriptor(
{'name': 'name',
'resources': [
{'path': ['data.csv'],
'name': 'data',
'schema':
{'fields': [
{'name': 'id', 'type': 'integer'},
{'name': 'city', 'type': 'string'}]}}]})
def get_fiscal_datapackage(skip_validation=False, source=None):
"""Create the master fiscal datapackage from parts."""
with open(FISCAL_METADATA_FILE) as stream:
fiscal_datapackage = yaml.load(stream.read())
if source:
datapackage = source
datapackage['name'] = slugify(os.getcwd().lstrip(DATA_DIR)).lower()
else:
datapackage = fiscal_datapackage
with open(FISCAL_SCHEMA_FILE) as stream:
schema = yaml.load(stream.read())
datapackage['resources'][0]['schema'] = schema
datapackage['resources'][0].update(mediatype='text/csv')
datapackage['resources'] = [datapackage['resources'][0]]
# TODO: Update the resource properties in the fiscal data-package
with open(FISCAL_MODEL_FILE) as stream:
datapackage['model'] = yaml.load(stream.read())
if not skip_validation:
DataPackage(datapackage, schema='fiscal').validate()
return datapackage
def get(location):
"""
Helper function to retreive data from a data package located at the
provided location.
"""
datapkg = DataPackage(location)
return datapkg.data
def datapackage(source, **options):
errors = []
tables = []
# Prepare datapackage
datapackage = DataPackage(source, **options)
for exception in datapackage.iter_errors():
# Error message should contain datapackage source (often it's path)
message = spec['errors']['datapackage-error']['message']
message = message.format(
error_message='{problem} [{source}]'.format(
problem=str(exception).splitlines()[0],
source=str(source)))
errors.append({
'code': 'datapackage-error',
'message': message,
'row-number': None,
'column-number': None,
})
# Add tables
if not errors:
for resource in datapackage.resources:
path = resource.remote_data_path or resource.local_data_path
tables.append({
'source': path,
'stream': Stream(path, headers=1),
'schema': Schema(resource.descriptor['schema']),
'extra': {
'datapackage': str(source),
},
})
return errors, tables
def assemble(metadata_file):
"""Assemble a data-package from its descriptor parts."""
def read(file):
with open(file) as yaml:
return load(yaml.read())
def add_name(info):
info['name'] = slugify(info['title'], separator='_')
return info
def get_files(filetype):
filename = metadata_file.replace('metadata', filetype)
folder = dirname(metadata_file)
schema_files_pattern = join(folder, filename)
return glob(schema_files_pattern)
descriptor = add_name(read(metadata_file))
resources = [add_name(read(file)) for file in get_files('resource')]
model = get_files('model')
descriptor['resources'] = resources
if model and len(model) == 1:
descriptor['model'] = model.pop()
return DataPackage(descriptor)
def load_data(pkgdir, engine):
dpo = DataPackage(pkgdir)
schema = dpo.resources[0].schema
csvpath = pkgdir + dpo.resources[0].path
data = [row for row in csv.DictReader(open(csvpath))]
table = SchemaTable(engine, 'table', schema)
table.create()
table.load_iter(data)
def __init__(self, datapackage_dir, with_dependencies):
self.datapackage_dir = datapackage_dir
self.with_dependencies = with_dependencies
datapackage_descriptor_file = os.path.join(datapackage_dir,
"datapackage.json")
with open(datapackage_descriptor_file) as f:
descriptor = json.load(f)
self.fix_descriptor(descriptor)
self.datapackage = DataPackage(descriptor,
default_base_path=self.datapackage_dir)
def _get_load_resources(self):
if not hasattr(self, "_load_resources"):
self._load_resources = []
for load_resource in self._parameters["load-resources"]:
if os.path.exists(load_resource["url"]):
datapackage = DataPackage(load_resource["url"])
for resource in datapackage.resources:
if resource.descriptor["name"] == load_resource["resource"]:
self._load_resources.append(resource)
return self._load_resources
def __new__(mcls, name, bases, attrs):
cls = super(BaseMeta, mcls).__new__(mcls, name, bases, attrs)
datapackage = attrs.get('__datapackage__')
if datapackage:
if isinstance(datapackage, basestring):
datapackage = DataPackage(unicode(datapackage))
resource_name = unicode(attrs.get('__resource__'))
metadata = attrs.get('__metadata__', metadata_)
mapper(cls, datapackage, resource_name, metadata)
cls.__queryset__ = SQLAlchemyQuerySet
return cls
def _load_dp(self, path):
dppath = join(path, 'datapackage.json')
# do we need to do this or is it done in datapackage library?
if not exists(dppath):
raise DpmException(
'No Data Package found at %s. Did not find datapackage.json at %s'
% (path, dppath))
dp = DataPackage(dppath)
return dp
def import_package(storage, descriptor):
"""Import Data Package to storage.
Parameters
----------
storage: object
Storage object.
descriptor: str
Path to descriptor.
"""
# Init maps
tables = []
schemas = []
datamap = {}
mapping = {}
# Init model
model = DataPackage(descriptor)
# Collect tables/schemas/data
for resource in model.resources:
name = resource.metadata.get('name', None)
table = _convert_path(resource.metadata['path'], name)
schema = resource.metadata['schema']
data = resource.iter()
tables.append(table)
schemas.append(schema)
datamap[table] = data
if name is not None:
mapping[name] = table
schemas = _convert_schemas(mapping, schemas)
# Create tables
for table in tables:
if storage.check(table):
storage.delete(table)
storage.create(tables, schemas)
# Write data to tables
for table in storage.tables:
storage.write(table, datamap[table])
def create_offline_client(paths, cachedir='.cached'):
''' Establish an offline client for more up to date assessments than those published
'''
import pandas as pd
from datapackage import DataPackage
all_pkgs = {}
for path in paths:
pkg = DataPackage(path)
for resource in map(format_patch, pkg.resources):
if resource.name not in all_pkgs:
all_pkgs[resource.name] = {
'schema': resource.descriptor['schema'],
'data': []
}
#
try:
all_pkgs[resource.name]['data'] += resource.read(keyed=True)
except Exception as e:
print(
f"datapackage exception while reading from table: '{resource.name}'"
)
print(e.errors)
raise e
#
joined_pkgs = {}
for resource_name, resource in all_pkgs.items():
if resource['data']:
data = pd.DataFrame(resource['data'])
else:
data = pd.DataFrame([],
columns=[
field['name']
for field in resource['schema']['fields']
])
#
for field in resource['schema']['fields']:
if field['type'] == 'datetime':
data[field['name']] = pd.to_datetime(data[field['name']],
utc=True)
elif field['type'] in {'array', 'object'}:
data[field['name']] = data[field['name']].apply(json.dumps)
joined_pkgs[resource_name] = dict(resource, data=data)
return DerivaCompatPkg(joined_pkgs, cachedir=cachedir)
def datavalidate(filepath, print_json):
"""
Validate csv file data, given its path. Print validation report. If the file is
a resource of the datapackage in current dir, will use datapackage.json schema for
validation; otherwise infer the schema automatically.
If no file path is given, validate all resources data in datapackage.json.
"""
inspector = goodtables.Inspector(infer_schema=True)
if exists('datapackage.json'):
dp = DataPackage('datapackage.json')
else:
dp = None
if not filepath and not dp:
echo(
'[ERROR] please provide csv file path or run command inside a datapackage dir.'
)
sys.exit(1)
if filepath:
schema = None
if dp:
# Try to find schema in the datapackage.json
for resource in dp.resources:
if resource.local_data_path == abspath(filepath):
#import ipdb; ipdb.sset_trace()
schema = resource.descriptor.get('schema')
break
report = inspector.inspect(filepath, schema=schema)
else:
# Validate whole datapackage
dprclient.validate_metadata(dp)
report = dprclient.validate_data(dp)
dprclient.print_inspection_report(report, print_json)
if not report['valid']:
sys.exit(1)
def __init__(self, dp_url):
self.datapackage = DataPackage(dp_url)
self.resource = self.datapackage.resources[0]
descriptor = self.resource.descriptor
self.type_name = descriptor['name']
self._schema = descriptor['schema']
fields = self._schema['fields']
try:
self.keys = self._schema['primaryKey']
except KeyError:
logger.exception('Failed to load %s', dp_url)
raise
if isinstance(self.keys, str):
self.keys = [self.keys]
self.date_fields = {}
self.range_structure = {}
for field in fields:
if field.get("es:time-range"):
self.date_fields[field["es:time-range"]] = field["name"]
try:
self.scoring_column = next(iter(
filter(lambda f: 'es:score-column' in f, fields),
))['name']
except StopIteration:
self.scoring_column = '<none>'
self._mapping_generator = MappingGenerator()
try:
self.mapping, self.search_fields = self.build_mapping(self._schema)
except: #noqa
logger.exception('Failed to load %s', dp_url)
raise
def test_support_criteria_parser():
dp = DataPackage({
"name":
"support-criteria",
"resources": [{
"name": "criteria",
"dialect": {
"delimiter": ",",
"doubleQuote": True,
"lineTerminator": "\r\n",
"quoteChar": '"',
"skipInitialSpace": False
},
"encoding": "utf-8",
"format": "csv",
"path": "tests/support/criteria.csv",
"schema": {
"fields": [
# the original support-criteria fields
{
"format": "%Y-%m-%d",
"name": "date",
"type": "date"
},
{
"name": "title",
"type": "string"
},
{
"name": "paper_type",
"type": "string"
},
{
"name": "office",
"type": "string"
},
{
"format": "uri",
"name": "pdf_url",
"type": "string"
},
# the expected data from the parser
{
"name": "expected_purpose",
"type": "string"
}
]
}
}]
})
i = 0
num_parsed = 0
num_expected_purposes = 0
for i, row in enumerate(dp.resources[0].iter(keyed=True)):
parsed_row = parse_row(row)
if len(parsed_row["purpose"]
) > 0 and parsed_row["purpose"] != row["title"]:
num_parsed += 1
row_expected_purpose = row["expected_purpose"] if row[
"expected_purpose"] else ""
if len(row_expected_purpose) > 0:
num_expected_purposes += 1
assert parsed_row["purpose"] == row_expected_purpose, "{}".format({
"i":
i,
"row":
row,
"parsed_row":
parsed_row,
"expected_purpose":
row_expected_purpose
})
assert i == 358
assert num_expected_purposes > 20, "not enough purposes were checked, might be something wrong with the criteria.csv file"
assert num_parsed > 190, "not enough parsed rows"
def simpsons_datapackage(self, simpsons_descriptor_path):
datapackage = DataPackage(descriptor=simpsons_descriptor_path)
for r in datapackage.resources:
sanitize_resource_schema(r)
return datapackage
def __init__(self, datapackage, databackend):
if isinstance(datapackage, basestring):
datapackage = DataPackage(unicode(datapackage))
self.datapackage = datapackage
self.models_maker = ModelsMaker(datapackage, backend=databackend)
self._resources = {}
def simpsons_broken_datapackage(self, simpsons_broken_descriptor_path):
return DataPackage(descriptor=simpsons_broken_descriptor_path)
pipeline.register_processor('schema', options={'schema': schema})
valid, report = pipeline.run()
return valid, report
def validate_schema(package):
try:
package.validate()
return []
except (ValidationError, InvalidSchemaError, SchemaValidationError):
for error in package.iter_errors():
yield error.message
if __name__ == '__main__':
package_ = DataPackage(SOURCE_SCHEMA)
encoding_ = detect_encoding(SOURCE_DATA)
errors_ = list(validate_schema(package_))
source_df = read_excel(SOURCE_DATA, header=4, skiprows=range(0, 3))
comment_lines = source_df.index[COMMENT_LINES]
for i in COMMENT_LINES:
print(list(source_df.iloc[i]))
source_df = source_df.drop(comment_lines)
source_csv = SOURCE_DATA.replace('xls', 'csv')
source_df.to_csv(source_csv)
odo(source_df, source_csv)
if errors_:
for message in errors_:
def test_assemble_fiscal_datapackage_returns_a_valid_fiscal_descriptor():
datapackage = assemble_fiscal_datapackage()
# The validation raises an exception if the validation fails
assert DataPackage(datapackage, schema='fiscal').validate() is None
def load_fdp_to_db(self, package, callback=noop):
"""
Load an FDP to the database, create a babbage model and save it as well
:param package: URL for the datapackage.json
:param callback: callback to use to send progress updates
"""
self.callback = callback
self.package = package
# Load and validate the datapackage
self.status_update(status=STATUS_LOADING_DATAPACKAGE)
self.dpo = DataPackage(package)
self.status_update(status=STATUS_VALIDATING_DATAPACKAGE)
self.dpo.validate()
self.status_update(status=STATUS_LOADING_RESOURCE)
resource = self.dpo.resources[0]
schema = resource.descriptor['schema']
# Use the cube manager to get the table name
self.datapackage_name = self.dpo.descriptor['name']
datapackage_owner = self.dpo.descriptor['owner']
datapackage_author = self.dpo.descriptor['author']
# Get the full name from the author field, and rewrite it without the email
self.fullname, email_addr = email.utils.parseaddr(datapackage_author)
email_addr = email_addr.split('@')[0] + '@not.shown'
self.dpo.descriptor['author'] = '{0} <{1}>'.format(
self.fullname, email_addr)
self.dpo.descriptor.setdefault('private', True)
self.model_name = "{0}:{1}".format(datapackage_owner,
self.datapackage_name)
table_name = table_name_for_package(datapackage_owner,
self.datapackage_name)
try:
all_fields = set()
field_translation = {}
# Process schema - slugify field names
for field in schema['fields']:
name = database_name(field['name'], all_fields)
all_fields.add(name)
translated_field = {'name': name, 'type': field['type']}
field_translation[field['name']] = translated_field
storage_schema = {
'fields': [{
'type': f['type'],
'name': field_translation[f['name']]['name'],
'format': f.get('format', 'default')
} for f in schema['fields']],
# Babbage likes just one primary key
'primaryKey':
'_id'
}
# Add Primary key to schema
storage_schema['fields'].insert(0, {
'name': '_id',
'type': 'integer'
})
# Create Babbage Model
self.status_update(status=STATUS_CREATING_BABBAGE_MODEL)
self.model = fdp_to_model(self.dpo, table_name, resource,
field_translation)
if self.check_hashes(resource):
# Create indexes
indexes = set()
primary_keys = schema.get('primaryKey', [])
for dim in self.dpo.descriptor.get('model',
{}).get('dimensions',
{}).values():
attributes = dim.get('attributes', {})
for attribute in attributes.values():
source = attribute.get('source')
if source in primary_keys:
indexes.add((field_translation[source]['name'], ))
labelfor = attribute.get('labelfor')
if labelfor is not None:
labelfor = attributes.get(labelfor, {})
labelfor_source = labelfor.get('source')
if labelfor_source in primary_keys:
indexes.add((
field_translation[labelfor_source]['name'],
field_translation[source]['name'],
))
indexes = list(indexes)
logging.error('INDEXES: %r', indexes)
#
# if dim['label'] in primary_keys:
# key_field = dim['attributes'][dim['key_attribute']]['label']
# key_field = field_translation[key_field]['name']
# indexes.append((key_field,))
#
# label_field = dim['attributes'].get(dim.get('label_attribute'), {}).get('label')
# if label_field is not None:
# label_field = field_translation[label_field]['name']
# if label_field != key_field:
# indexes.append((key_field, label_field))
# Load 1st resource data into DB
# We use the prefix name so that JTS-SQL doesn't load all table data into memory
storage = Storage(self.engine, prefix=table_name)
faux_table_name = ''
if faux_table_name in storage.buckets:
self.status_update(status=STATUS_DELETING_TABLE)
storage.delete(faux_table_name)
self.status_update(status=STATUS_CREATING_TABLE)
indexes_fields = None
if indexes:
indexes_fields = [indexes]
storage.create(faux_table_name,
storage_schema,
indexes_fields=indexes_fields)
self.status_update(status=STATUS_LOADING_DATA_READY)
row_processor = RowProcessor(resource.iter(keyed=True),
self.status_update, schema,
self.dpo.descriptor)
storage.write(faux_table_name, row_processor.iter())
cache = get_os_cache()
if cache is not None:
logging.info('Clearing cache for context=%s',
self.model_name)
cache.clear(self.model_name)
response = {
'model_name': self.model_name,
'babbage_model': self.model,
'package': self.dpo.descriptor
}
self.status_update(status=STATUS_DONE, data=response)
except Exception as e:
logging.exception('LOADING FAILED')
self.status_update(status=STATUS_FAIL,
error=traceback.format_exc())
return False
return True
