A Python library for working with Data Packages.
Python HTML
Latest commit b199600 Dec 13, 2016 @roll roll committed on GitHub v0.8.5



Gitter Build Status Windows Build Status Test Coverage Support Python versions 2.7, 3.3, 3.4 and 3.5

A model for working with Data Packages.


pip install datapackage


Reading a Data Package and its resource

import datapackage

dp = datapackage.DataPackage('http://data.okfn.org/data/core/gdp/datapackage.json')
brazil_gdp = [{'Year': int(row['Year']), 'Value': float(row['Value'])}
              for row in dp.resources[0].data if row['Country Code'] == 'BRA']

max_gdp = max(brazil_gdp, key=lambda x: x['Value'])
min_gdp = min(brazil_gdp, key=lambda x: x['Value'])
percentual_increase = max_gdp['Value'] / min_gdp['Value']

msg = (
    'The highest Brazilian GDP occured in {max_gdp_year}, when it peaked at US$ '
    '{max_gdp:1,.0f}. This was {percentual_increase:1,.2f}% more than its '
    'minimum GDP in {min_gdp_year}.'

# The highest Brazilian GDP occured in 2011, when it peaked at US$ 2,615,189,973,181. This was 172.44% more than its minimum GDP in 1960.

Validating a Data Package

import datapackage

dp = datapackage.DataPackage('http://data.okfn.org/data/core/gdp/datapackage.json')
except datapackage.exceptions.ValidationError as e:
    # Handle the ValidationError

Retrieving all validation errors from a Data Package

import datapackage

# This descriptor has two errors:
#   * It has no "name", which is required;
#   * Its resource has no "data", "path" or "url".
descriptor = {
    'resources': [

dp = datapackage.DataPackage(descriptor)

for error in dp.iter_errors():
    # Handle error

Creating a Data Package

import datapackage

dp = datapackage.DataPackage()
dp.descriptor['name'] = 'my_sleep_duration'
dp.descriptor['resources'] = [
    {'name': 'data'}

resource = dp.resources[0]
resource.descriptor['data'] = [
    7, 8, 5, 6, 9, 7, 8

with open('datapackage.json', 'w') as f:
# {"name": "my_sleep_duration", "resources": [{"data": [7, 8, 5, 6, 9, 7, 8], "name": "data"}]}

Using a schema that's not in the local cache

import datapackage
import datapackage.registry

# This constant points to the official registry URL
# You can use any URL or path that points to a registry CSV
registry_url = datapackage.registry.Registry.DEFAULT_REGISTRY_URL
registry = datapackage.registry.Registry(registry_url)

descriptor = {}  # The datapackage.json file
schema = registry.get('tabular')  # Change to your schema ID

dp = datapackage.DataPackage(descriptor, schema)

Push/pull Data Package to storage

Package provides push_datapackage and pull_datapackage utilities to push and pull to/from storage.

This functionality requires jsontableschema storage plugin installed. See plugins section of jsontableschema docs for more information. Let's imagine we have installed jsontableschema-mystorage (not a real name) plugin.

Then we could push and pull datapackage to/from the storage:

All parameters should be used as keyword arguments.

from datapackage import push_datapackage, pull_datapackage

# Push
    backend='mystorage', **<mystorage_options>)

# Import
    descriptor='descriptor_path', name='datapackage_name',
    backend='mystorage', **<mystorage_options>)

Options could be a SQLAlchemy engine or a BigQuery project and dataset name etc. Detailed description you could find in a concrete plugin documentation.

See concrete examples in plugins section of jsontableschema docs.

Developer notes

These notes are intended to help people that want to contribute to this package itself. If you just want to use it, you can safely ignore them.

Updating the local schemas cache

We cache the schemas from https://github.com/dataprotocols/schemas using git-subtree. To update it, use:

git subtree pull --prefix datapackage/schemas https://github.com/dataprotocols/schemas.git master --squash