pytest-voluptuous 1.0.1

Install

Works on python 2.7 and 3.4+:

pip install pytest-voluptuous

Changelog

See CHANGELOG.

Features

• Provides utility schemas (S, Exact and Partial) to cut down boilerplate.

• Implement a pytest hook to provide error details on assert failure.

• Print descriptive validation failure messages.

• Equal and Unordered validators (contributed to voluptuous project, available in 0.10+).

Why?

Because writing:

>>> r = {'info': {'package_url': 'http://pypi.python.org/pypi/pytest'}}
>>> assert 'info' in r
>>> assert 'package_url' in r['info']
>>> assert r['info']['package_url'] == 'http://pypi.python.org/pypi/pytest'

…is just way too annoying.

Why not JSON schema? It’s too verbose, too inconvenient. JSON schema will never match the convenience of a validation library that can utilize the goodies of the platform.

Why voluptuous and not some other library? No special reason - but it’s pretty easy to use and understand. Also, the syntax is quite compact.

Usage

In pytest:

>>> import requests
>>> from pytest_voluptuous import S, Partial, Exact
>>> from voluptuous.validators import All, Length
>>> resp = requests.get('https://pypi.python.org/pypi/pytest/json')
>>> assert S({
...     'info': Partial({
...         'package_url': 'http://pypi.python.org/pypi/pytest',
...         'platform': 'unix',
...         'description': Length(min=100),
...         'downloads': dict,
...         'classifiers': list,
...     }),
...     'releases': {
...         any: list
...     },
...     'urls': list
... }) == resp.json()

Note: if you run this in shell, there’s no pytest magic in play and in case of failure, you’ll just get AssertionError as in:

>>> assert S({'does_not_exist': 1}) == resp.json()
Traceback (most recent call last):
    ...
AssertionError

Don’t worry - the promised magic comes into play when you run the validation in a pytest test.

Use == operator to do exact validation:

>>> data = {'foo': 1, 'bar': True}
>>> S({'foo': 1, 'bar': True}) == data
True

We omit assert in these examples (for easier doctesting).

Use <= to do partial validation (to allow extra keys, that is):

>>> S({'foo': 1}) == data  # not valid
False
>>> S({'foo': 1}) <= data  # valid
True

The operator you choose gets inherited, so with test data of:

>>> data = {
...     'outer1': {
...         'inner1': 1,
...         'inner2': True
...     },
...     'outer2': 'foo'
... }

With == you must provide exact value also in nested context:

>>> S({
...     'outer1': {
...         'inner1': 1,  # this would be valid but...
...         # missing 'inner2'
...     },
...     'outer2': 'foo'
... }) == data
False
>>> S({
...     'outer1': {
...         'inner1': int,  # exact/partial matching
...         'inner2': bool  # is for keys only
...     },
...     'outer2': 'foo'
... }) == data
True

<= implies partial matching:

>>> S({
...     'outer1': {
...         'inner1': int,
...         # 'inner2' missing but that's ok
...     },
...     # 'outer2' is missing too
... }) <= data
True

When you need to mix and match operators, you can loosen matching with Partial:

>>> S({
...     'outer1': Partial({
...         'inner1': int
...         # 'inner2' ok to omit as scope is partial
...     }),
...     'outer2': 'foo'  # can't be missing as outer scope is exact
... }) == data
True

And stricten with Exact:

>>> S({
...     'outer1': Exact({
...         'inner1': int,
...         'inner2': bool
...     }),
...     # 'outer2' can be missing as outer scope is partial
... }) <= data
True

Remember, matching mode is inherited, so you may end up doing stuff like this:

>>> data['outer1']['inner1'] = {'prop': 1}
>>> S({
...     'outer1': Partial({
...         'inner1': Exact({
...             'prop': 1
...         })
...     }),
...     'outer2': 'foo'
... }) == data
True

There is no >=. If you want to declare schema keys that may be missing, use Optional:

>>> from voluptuous.schema_builder import Optional
>>> S({Optional('foo'): str}) == {'extra': 1}
False
>>> S({'foo': str}) == {}
False
>>> S({'foo': str}) <= {}
False
>>> S({Optional('foo'): str}) == {}
True
>>> S({Optional('foo'): str}) <= {'extra': 1}
True

Or, if you want to make all keys optional, override required:

>>> from voluptuous.schema_builder import Required
>>> S({'foo': str}, required=False) == {}
True

In these cases, if you want to require a key:

>>> S({'foo': str, Required('bar'): 1}, required=False) == {}
False
>>> S({'foo': str, Required('bar'): 1}, required=False) == {'bar': 1}
True

That’s it. For available validators, look into voluptuous docs.

Gotchas

Voluptuous 0.9.3 and earlier:

In voluptuous pre-0.10.2 [] matches any list, not an empty list. To declare an empty list, use Equal([]).

Similarly, in voluptuous pre-0.10.2, {} doesn’t always match an empty dict. If you’re inside a Schema({...}, extra=PREVENT_EXTRA) (or Exact), {} does indeed match exactly {}. However, inside Schema({...}, extra=ALLOW_EXTRA) (or ``Partial), it matches any dict (because any extra keys are allowed). To declare an empty dict, use Equal({}).

Voluptuous 0.10.0+:

In voluptuous 0.10.0+ {} and [] evaluate as empty dict and empty list, so you don’t need above workarounds.

Always use dict and list to validate dict or list of any size. It works despite voluptuous version.

Any version:

[str, int] matches any list that contains both strings and ints (in any order and 1-n times). To validate a list of fixed length with those types in it, use ExactSequence([str, int]) and Unordered([str, int]) when the order has no meaning. You can also use values inside these as in ExactSequence([2, 3]).

License

Apache 2.0 licensed. See LICENSE for more details.

Changelog