Django data schema is a lightweight Django app for defining the schema for a model, dictionary, or list. By describing a schema on a piece of data, this allows other applications to easily reference fields of models or fields in dictionaries (or their related json fields).

Django data schema also takes care of all conversions under the hood, such as parsing datetime strings, converting strings to numeric values, using default values when values don't exist, and so on.

1. Installation
2. Model Overview
3. Examples

pip install django-data-schema

Django data schema defines three models for building schemas on data. These models are DataSchema, FieldSchema, and FieldOptional.

The DataSchema model provides a model_content_type field that points to a Django ContentType model. This field represents which object this schema is modeling. If the field is None, it is assumed that this schema models an object such as a dictionary or list.

After the enclosing DataSchema has been defined, various FieldSchema models can reference the main data schema. FieldSchema models provide the following attributes:

• field_key: The name of the field. Used to identify a field in a dictionary or model.
• field_position: The position of the field. Used to identify a field in a list.
• uniqueness_order: The order of this field in the uniqueness constraint of the schema. Defaults to None.
• field_type: The type of field. More on the field types below.
• field_format: An optional formatting string for the field. Used differently depending on the field type and documented more below.
• default_value: If the field returns None, this default value will be returned instead.

A FieldSchema object must specify its data type. While data of a given type can be stored in different formats, django-data-schema normalizes the data when accessing it through get_value, described below. The available types are listed in the FieldSchemaType class. These types are listed here, with the type they normalize to:

• FieldSchemaType.DATE: A python date object from the datetime module. Currently returned as a datetime object.
• FieldSchemaType.DATETIME: A python datetime object from the datetime module.
• FieldSchemaType.INT: A python int.
• FieldSchemaType.FLOAT: A python float.
• FieldSchemaType.STRING: A python str.
• FieldSchemaType.BOOLEAN: A python bool.

These fields provide the necessary conversion mechanisms when accessing data via FieldSchema.get_value. Differences in how the get_value function operates are detailed below.

The get_value function has the following behavior on DATE and DATETIME fields:

• If called on a Python int or float value, the numeric value will be passed to the datetime.utcfromtimestamp function.
• If called on a string or unicode value, the string will be stripped of all trailing and leading whitespace. If the string is empty, the default value (or None) will be used. If the string is not empty, it will be passed to dateutil's parse function. If the field_format field is specified on the FieldSchema object, it will be passed to the strptime function instead.
• If called on an aware datetime object (or a string with a timezone), it will be converted to naive UTC time.
• If called on None, the default value (or None) is returned.

The get_value function has the following behavior on INT and FLOAT fields:

• If called on a string or unicode value, the string will be stripped of all non-numeric numbers except for periods. If the string is blank, the default value (or None) will be returned. If not, the string will then be passed to int() or float().
• If called on an int or float, the value will be passed to the int() or float() function.
• No other values can be converted. The field_format parameter is ignored.
• If called on None, the default value (or None) is returned.

The get_value function has the following behavior on a STRING field:

• If called on a string or unicode value, the string will be stripped of all trailing and leading whitespace. If a field_format is specified, the string is then be matched to the regex. If it passes, the string is returned. If not, None is returned and the default value is used (or None).
• All other types are passed to the str() function.
• If called on None, the default value (or None) is returned.

The get_value function has the following behavior on a BOOLEAN field:

• Bool data types will return True or False
• Truthy looking string values return True ('t', 'T', 'true', 'True', 'TRUE', 1, '1')
• Falsy looking string values return False ('f', 'F', 'false', 'False', 'FALSE', 0, '0')
• If called on None, the default value (or None) is returned.

A data schema can be created like the following:

from data_schema import DataSchema, FieldSchema, FieldSchemaType

user_login_schema = DataSchema.objects.create()
user_id_field = FieldSchema.objects.create(
    data_schema=user_login_schema, field_key='user_id', uniqueness_order=1, field_type=FieldSchemaType.STRING)
login_time_field = FieldSchema.objects.create(
    data_schema=user_login_schema, field_key='login_time', field_type=FieldSchemaType.DATETIME)

The above example represents the schema of a user login. In this schema, the user id field provides the uniqueness constraint of the data. The uniqueness constraint can then easily be accessed by simply doing the following.

unique_fields = user_login_schema.get_unique_fields()

The above function returns the unique fields in the order in which they were specified, allowing the user to generate a unique ID for the data.

To obtain values of data using the schema, one can use the get_value function as follows:

data = {
    'user_id': 'my_user_id',
    'login_time': 1396396800,
}

print login_time_field.get_value(data)
2014-04-02 00:00:00

Note that the get_value function looks at the type of data object and uses the proper access method. If the data object is a dict, it accesses it using data[field_key]. If it is an object, it accesses it with getattr(data, field_key). An array is accessed as data[field_position].

Here's another example of parsing datetime objects in an array with a format string.

string_time_field_schema = FieldSchema.objects.create(
    data_schema=data_schema, field_key='time', field_position=1, field_type=FieldSchemaType.DATETIME, field_format='%Y-%m-%d %H:%M:%S')

print string_time_field_schema.get_value(['value', '2013-04-12 12:12:12'])
2013-04-12 12:12:12

Note that if you are parsing numerical fields, Django data schema will strip out any non-numerical values, allowing the user to get values of currency-based numbers and other formats.

revenue_field_schema = FieldSchema.objects.create(
    data_schema=data_schema, field_key='revenue', field_type=FieldSchemaType.FLOAT)

print revenue_field_schema.get_value({'revenue': '$15,000,456.23'})
15000456.23

Note that FieldSchema objects have an analogous set_value method for setting the value of a field. The set_value method does not do any data conversions, so when calling this method, be sure to use a value that is in the correct format.