Skip to main content
Help us improve Python packaging – donate today!

zope.interface extension for defining data schemas

Project Description

Zope 3 Schemas


This package is intended to be independently reusable in any Python project. It is maintained by the Zope Toolkit project.

Schemas extend the notion of interfaces to detailed descriptions of Attributes (but not methods). Every schema is an interface and specifies the public fields of an object. A field roughly corresponds to an attribute of a python object. But a Field provides space for at least a title and a description. It can also constrain its value and provide a validation method. Besides you can optionally specify characteristics such as its value being read-only or not required.

Zope 3 schemas were born when Jim Fulton and Martijn Faassen thought about Formulator for Zope 3 and PropertySets while at the Zope 3 sprint at the Zope BBQ in Berlin. They realized that if you strip all view logic from forms then you have something similar to interfaces. And thus schemas were born.

Simple Usage

Let’s have a look at a simple example. First we write an interface as usual, but instead of describing the attributes of the interface with Attribute instances, we now use schema fields:

>>> import zope.interface
>>> import zope.schema
>>> class IBookmark(zope.interface.Interface):
...     title = zope.schema.TextLine(
...         title=u'Title',
...         description=u'The title of the bookmark',
...         required=True)
...     url = zope.schema.URI(
...         title=u'Bookmark URL',
...         description=u'URL of the Bookmark',
...         required=True)

Now we create a class that implements this interface and create an instance of it:

>>> class Bookmark(object):
...     zope.interface.implements(IBookmark)
...     title = None
...     url = None
>>> bm = Bookmark()

We would now like to only add validated values to the class. This can be done by first validating and then setting the value on the object. The first step is to define some data:

>>> title = u'Zope 3 Website'
>>> url = ''

Now we, get the fields from the interface:

>>> title_field = IBookmark.get('title')
>>> url_field = IBookmark.get('url')

Next we have to bind these fields to the context, so that instance-specific information can be used for validation:

>>> title_bound = title_field.bind(bm)
>>> url_bound = url_field.bind(bm)

Now that the fields are bound, we can finally validate the data:

>>> title_bound.validate(title)
>>> url_bound.validate(url)

If the validation is successful, None is returned. If a validation error occurs a ValidationError will be raised; for example:

>>> url_bound.validate(u'')
Traceback (most recent call last):
WrongType: (u'', <type 'str'>, 'url')
>>> url_bound.validate('')
Traceback (most recent call last):

Now that the data has been successfully validated, we can set it on the object:

>>> title_bound.set(bm, title)
>>> url_bound.set(bm, url)

That’s it. You still might think this is a lot of work to validate and set a value for an object. Note, however, that it is very easy to write helper functions that automate these tasks. If correctly designed, you will never have to worry explicitly about validation again, since the system takes care of it automatically.

What is a schema, how does it compare to an interface?

A schema is an extended interface which defines fields. You can validate that the attributes of an object conform to their fields defined on the schema. With plain interfaces you can only validate that methods conform to their interface specification.

So interfaces and schemas refer to different aspects of an object (respectively its code and state).

A schema starts out like an interface but defines certain fields to which an object’s attributes must conform. Let’s look at a stripped down example from the programmer’s tutorial:

>>> import re
>>> class IContact(zope.interface.Interface):
...     """Provides access to basic contact information."""
...     first = zope.schema.TextLine(title=u"First name")
...     last = zope.schema.TextLine(title=u"Last name")
...     email = zope.schema.TextLine(title=u"Electronic mail address")
...     address = zope.schema.Text(title=u"Postal address")
...     postalCode = zope.schema.TextLine(
...         title=u"Postal code",
...         constraint=re.compile("\d{5,5}(-\d{4,4})?$").match)

TextLine is a field and expresses that an attribute is a single line of Unicode text. Text expresses an arbitrary Unicode (“text”) object. The most interesting part is the last attribute specification. It constrains the postalCode attribute to only have values that are US postal codes.

Now we want a class that adheres to the IContact schema:

>>> class Contact(object):
...     zope.interface.implements(IContact)
...     def __init__(self, first, last, email, address, pc):
...         self.first = first
...         self.last = last
... = email
...         self.address = address
...         self.postalCode = pc

Now you can see if an instance of Contact actually implements the schema:

>>> someone = Contact(u'Tim', u'Roberts', u'tim@roberts', u'',
...                   u'12032-3492')
>>> for field in zope.schema.getFields(IContact).values():
...     bound = field.bind(someone)
...     bound.validate(bound.get(someone))

Data Modeling Concepts

The zope.schema package provides a core set of field types, including single- and multi-line text fields, binary data fields, integers, floating-point numbers, and date/time values.

Selection issues; field type can specify:

  • “Raw” data value

    Simple values not constrained by a selection list.

  • Value from enumeration (options provided by schema)

    This models a single selection from a list of possible values specified by the schema. The selection list is expected to be the same for all values of the type. Changes to the list are driven by schema evolution.

    This is done by mixing-in the IEnumerated interface into the field type, and the Enumerated mix-in for the implementation (or emulating it in a concrete class).

  • Value from selection list (options provided by an object)

    This models a single selection from a list of possible values specified by a source outside the schema. The selection list depends entirely on the source of the list, and may vary over time and from object to object. Changes to the list are not related to the schema, but changing how the list is determined is based on schema evolution.

    There is not currently a spelling of this, but it could be facilitated using alternate mix-ins similar to IEnumerated and Enumerated.

  • Whether or not the field is read-only

    If a field value is read-only, it cannot be changed once the object is created.

  • Whether or not the field is required

    If a field is designated as required, assigned field values must always be non-missing. See the next section for a description of missing values.

  • A value designated as missing

    Missing values, when assigned to an object, indicate that there is ‘no data’ for that field. Missing values are analogous to null values in relational databases. For example, a boolean value can be True, False, or missing, in which case its value is unknown.

    While Python’s None is the most likely value to signify ‘missing’, some fields may use different values. For example, it is common for text fields to use the empty string (‘’) to signify that a value is missing. Numeric fields may use 0 or -1 instead of None as their missing value.

    A field that is ‘required’ signifies that missing values are invalid and should not be assigned.

  • A default value

    Default field values are assigned to objects when they are first created.

Fields and Widgets

Widgets are components that display field values and, in the case of writable fields, allow the user to edit those values.


  • Display current field values, either in a read-only format, or in a format that lets the user change the field value.
  • Update their corresponding field values based on values provided by users.
  • Manage the relationships between their representation of a field value and the object’s field value. For example, a widget responsible for editing a number will likely represent that number internally as a string. For this reason, widgets must be able to convert between the two value formats. In the case of the number-editing widget, string values typed by the user need to be converted to numbers such as int or float.
  • Support the ability to assign a missing value to a field. For example, a widget may present a None option for selection that, when selected, indicates that the object should be updated with the field’s missing value.



This document highlights unusual and subtle aspects of various fields and field classes, and is not intended to be a general introduction to schema fields. Please see README.txt for a more general introduction.

While many field types, such as Int, TextLine, Text, and Bool are relatively straightforward, a few have some subtlety. We will explore the general class of collections and discuss how to create a custom creation field; discuss Choice fields, vocabularies, and their use with collections; and close with a look at the standard approach to using these fields to find views (“widgets”).


Normal fields typically describe the API of the attribute – does it behave as a Python Int, or a Float, or a Bool – and various constraints to the model, such as a maximum or minimum value. Collection fields have additional requirements because they contain other types, which may also be described and constrained.

For instance, imagine a list that contains non-negative floats and enforces uniqueness. In a schema, this might be written as follows:

>>> from zope.interface import Interface
>>> from zope.schema import List, Float
>>> class IInventoryItem(Interface):
...     pricePoints = List(
...         title=u"Price Points",
...         unique=True,
...         value_type=Float(title=u"Price", min=0.0)
...     )

This indicates several things.

  • pricePoints is an attribute of objects that implement IInventoryItem.
  • The contents of pricePoints can be accessed and manipulated via a Python list API.
  • Each member of pricePoints must be a non-negative float.
  • Members cannot be duplicated within pricePoints: each must be must be unique.
  • The attribute and its contents have descriptive titles. Typically these would be message ids.

This declaration creates a field that implements a number of interfaces, among them these:

>>> from zope.schema.interfaces import IList, ISequence, ICollection
>>> IList.providedBy(IInventoryItem['pricePoints'])
>>> ISequence.providedBy(IInventoryItem['pricePoints'])
>>> ICollection.providedBy(IInventoryItem['pricePoints'])

Creating a custom collection field

Ideally, custom collection fields have interfaces that inherit appropriately from either zope.schema.interfaces.ISequence or zope.schema.interfaces.IUnorderedCollection. Most collection fields should be able to subclass zope.schema._field.AbstractCollection to get the necessary behavior. Notice the behavior of the Set field in zope.schema._field: this would also be necessary to implement a Bag.

Choices and Vocabularies

Choice fields are the schema way of spelling enumerated fields and more. By providing a dynamically generated vocabulary, the choices available to a choice field can be contextually calculated.

Simple choices do not have to explicitly use vocabularies:

>>> from zope.schema import Choice
>>> f = Choice((640, 1028, 1600))
>>> f.validate(640)
>>> f.validate(960)
Traceback (most recent call last):
ConstraintNotSatisfied: 960
>>> f.validate('bing')
Traceback (most recent call last):
ConstraintNotSatisfied: bing

More complex choices will want to use registered vocabularies. Vocabularies have a simple interface, as defined in zope.schema.interfaces.IBaseVocabulary. A vocabulary must minimally be able to determine whether it contains a value, to create a term object for a value, and to return a query interface (or None) to find items in itself. Term objects are an abstraction that wraps a vocabulary value.

The Zope application server typically needs a fuller interface that provides “tokens” on its terms: ASCII values that have a one-to-one relationship to the values when the vocabulary is asked to “getTermByToken”. If a vocabulary is small, it can also support the IIterableVocabulary interface.

If a vocabulary has been registered, then the choice merely needs to pass the vocabulary identifier to the “vocabulary” argument of the choice during instantiation.

A start to a vocabulary implementation that may do all you need for many simple tasks may be found in zope.schema.vocabulary.SimpleVocabulary. Because registered vocabularies are simply callables passed a context, many registered vocabularies can simply be functions that rely on SimpleVocabulary:

>>> from zope.schema.vocabulary import SimpleVocabulary
>>> def myDynamicVocabulary(context):
...     v = dynamic_context_calculation_that_returns_an_iterable(context)
...     return SimpleVocabulary.fromValues(v)

The vocabulary interface is simple enough that writing a custom vocabulary is not too difficult itself.

Choices and Collections

Choices are a field type and can be used as a value_type for collections. Just as a collection of an “Int” value_type constrains members to integers, so a choice-based value type constrains members to choices within the Choice’s vocabulary. Typically in the Zope application server widgets are found not only for the collection and the choice field but also for the vocabulary on which the choice is based.

Using Choice and Collection Fields within a Widget Framework

While fields support several use cases, including code documentation and data description and even casting, a significant use case influencing their design is to support form generation – generating widgets for a field. Choice and collection fields are expected to be used within widget frameworks. The approach typically (but configurably) uses multiple dispatches to find widgets on the basis of various aspects of the fields.

Widgets for all fields are found by looking up a browser view of the field providing an input or display widget view. Typically there is only a single “widget” registered for Choice fields. When it is looked up, it performs another dispatch – another lookup – for a widget registered for both the field and the vocabulary. This widget typically has enough information to render without a third dispatch.

Collection fields may fire several dispatches. The first is the usual lookup by field. A single “widget” should be registered for ICollection, which does a second lookup by field and value_type constraint, if any, or, theoretically, if value_type is None, renders some absolutely generic collection widget that allows input of any value imaginable: a check-in of such a widget would be unexpected. This second lookup may find a widget that knows how to render, and stop. However, the value_type may be a choice, which will usually fire a third dispatch: a search for a browser widget for the collection field, the value_type field, and the vocabulary. Further lookups may even be configured on the basis of uniqueness and other constraints.

This level of indirection may be unnecessary for some applications, and can be disabled with simple ZCML changes within



Sources are designed with three concepts:

  • The source itself - an iterable

    This can return any kind of object it wants. It doesn’t have to care for browser representation, encoding, …

  • A way to map a value from the iterable to something that can be used for form values - this is called a token. A token is commonly a (unique) 7bit representation of the value.

  • A way to map a value to something that can be displayed to the user - this is called a title

The last two elements are dispatched using a so called term. The ITitledTokenizedTerm interface contains a triple of (value, token, title).

Additionally there are some lookup functions to perform the mapping between values and terms and tokens and terms.

Sources that require context use a special factory: a context source binder that is called with the context and instanciates the source when it is actually used.

Sources in Fields

A choice field can be constructed with a source or source name. When a source is used, it will be used as the source for valid values.

Create a source for all odd numbers.

>>> from zope import interface
>>> from zope.schema.interfaces import ISource, IContextSourceBinder
>>> class MySource(object):
...     interface.implements(ISource)
...     divisor = 2
...     def __contains__(self, value):
...         return bool(value % self.divisor)
>>> my_source = MySource()
>>> 1 in my_source
>>> 2 in my_source
>>> from zope.schema import Choice
>>> choice = Choice(__name__='number', source=my_source)
>>> bound = choice.bind(object())
>>> bound.vocabulary

If a IContextSourceBinder is passed as the source argument to Choice, it’s bind method will be called with the context as its only argument. The result must implement ISource and will be used as the source.

>>> def my_binder(context):
...     print "Binder was called."
...     source = MySource()
...     source.divisor = context.divisor
...     return source
>>> interface.directlyProvides(my_binder, IContextSourceBinder)
>>> class Context(object):
...     divisor = 3
>>> choice = Choice(__name__='number', source=my_binder)
>>> bound = choice.bind(Context())
Binder was called.
>>> bound.vocabulary
>>> bound.vocabulary.divisor

When using IContextSourceBinder together with default value, it’s impossible to validate it on field initialization. Let’s check if initalization doesn’t fail in that case.

>>> choice = Choice(__name__='number', source=my_binder, default=2)
>>> bound = choice.bind(Context())
Binder was called.
>>> bound.validate(bound.default)
>>> bound.validate(3)
Traceback (most recent call last):
ConstraintNotSatisfied: 3

It’s developer’s responsibility to provide a default value that fits the constraints when using context-based sources.

Schema Validation

There are two helper methods to verify schemas and interfaces:

first validates via the zope.schema field validators. If that succeeds the invariants are checked.
only validateds via the zope.schema field validators. The invariants are not checked.

Create an interface to validate against:

>>> import zope.interface
>>> import zope.schema
>>> class ITwoInts(zope.interface.Interface):
...     a = zope.schema.Int(max=10)
...     b = zope.schema.Int(min=5)
...     @zope.interface.invariant
...     def a_greater_b(obj):
...         print "Checking if a > b"
...         if obj.a <= obj.b:
...             raise zope.interface.Invalid("%s<=%s" % (obj.a, obj.b))

Create a silly model:

>>> class TwoInts(object):
...     pass

Create an instance of TwoInts but do not set attributes. We get two errors:

>>> ti = TwoInts()
>>> r = zope.schema.getValidationErrors(ITwoInts, ti)
>>> r.sort()
>>> r
[('a', SchemaNotFullyImplemented(...AttributeError...)),
 ('b', SchemaNotFullyImplemented(...AttributeError...))]
>>> r[0][1].args[0].args
("'TwoInts' object has no attribute 'a'",)
>>> r[1][1].args[0].args
("'TwoInts' object has no attribute 'b'",)

The getSchemaValidationErrors function returns the same result:

>>> r = zope.schema.getSchemaValidationErrors(ITwoInts, ti)
>>> r.sort()
>>> r
[('a', SchemaNotFullyImplemented(...AttributeError...)),
 ('b', SchemaNotFullyImplemented(...AttributeError...))]
>>> r[0][1].args[0].args
("'TwoInts' object has no attribute 'a'",)
>>> r[1][1].args[0].args
("'TwoInts' object has no attribute 'b'",)

Note that see no error from the invariant because the invariants are not vaildated if there are other schema errors.

When we set a valid value for a we still get the same error for b:

>>> ti.a = 11
>>> errors = zope.schema.getValidationErrors(ITwoInts, ti)
>>> errors.sort()
>>> errors
[('a', TooBig(11, 10)),
 ('b', SchemaNotFullyImplemented(...AttributeError...))]
>>> errors[1][1].args[0].args
("'TwoInts' object has no attribute 'b'",)
>>> errors[0][1].doc()
u'Value is too big'

After setting a valid value for a there is only the error for the missing b left:

>>> ti.a = 8
>>> r = zope.schema.getValidationErrors(ITwoInts, ti)
>>> r
[('b', SchemaNotFullyImplemented(...AttributeError...))]
>>> r[0][1].args[0].args
("'TwoInts' object has no attribute 'b'",)

After setting valid value for b the schema is valid so the invariants are checked. As b>a the invariant fails:

>>> ti.b = 10
>>> errors = zope.schema.getValidationErrors(ITwoInts, ti)
Checking if a > b
>>> errors
[(None, <zope.interface.exceptions.Invalid instance at 0x...>)]

When using getSchemaValidationErrors we do not get an error any more:

>>> zope.schema.getSchemaValidationErrors(ITwoInts, ti)

Set b=5 so everything is fine:

>>> ti.b = 5
>>> zope.schema.getValidationErrors(ITwoInts, ti)
Checking if a > b

Compare ValidationError

There was an issue with compare validation error with somthing else then an exceptions. Let’s test if we can compare ValidationErrors with different things

>>> from zope.schema._bootstrapinterfaces import ValidationError
>>> v1 = ValidationError('one')
>>> v2 = ValidationError('one')
>>> v3 = ValidationError('another one')

A ValidationError with the same arguments compares:

>>> v1 == v2

but not with an error with different arguments:

>>> v1 == v3

We can also compare validation erros with other things then errors. This was running into an AttributeError in previous versions of zope.schema. e.g. AttributeError: ‘NoneType’ object has no attribute ‘args’

>>> v1 == None
>>> v1 == object()
>>> v1 == False
>>> v1 == True
>>> v1 == 0
>>> v1 == 1
>>> v1 == int

If we compare a ValidationError with another validation error based class, we will get the following result:

>>> from zope.schema._bootstrapinterfaces import RequiredMissing
>>> r1 = RequiredMissing('one')
>>> v1 == r1


3.7.1 (2010-12-25)

  • The validation token, used in the validation of schema with Object Field to avoid infinite recursion, has been renamed. __schema_being_validated became _v_schema_being_validated, a volatile attribute, to avoid persistency and therefore, read/write conflicts.
  • Don’t allow “[]^`” in DottedName.

3.7.0 (2010-09-12)

  • Improve error messages when term tokens or values are duplicates.
  • Fix the buildout so the tests run.

3.6.4 (2010-06-08)

  • fix validation of schema with Object Field that specify Interface schema.

3.6.3 (2010-04-30)

  • Prefer the standard libraries doctest module to the one from zope.testing.

3.6.2 (2010-04-30)

  • Avoid maximum recursion when validating Object field that points to cycles
  • Made the dependency on zope.i18nmessageid optional.

3.6.1 (2010-01-05)

  • Allow “ test” to run at least a subset of the tests runnable via bin/test (227 for test vs. 258. for bin/test)
  • Make zope.schema._bootstrapfields.ValidatedProperty descriptor work under Jython.
  • Make “ test” tests pass on Jython.

3.6.0 (2009-12-22)

  • Prefer zope.testing.doctest over doctestunit.
  • Extend validation error to hold the field name.
  • Add FieldProperty class that uses Field.get and Field.set methods instead of storing directly on the instance __dict__.

3.5.4 (2009-03-25)

  • Don’t fail trying to validate default value for Choice fields with IContextSourceBinder object given as a source. See
  • Add an interface for DottedName field.
  • Add vocabularyName attribute to the IChoice interface, change “vocabulary” attribute description to be more sensible, making it zope.schema.Field instead of plain zope.interface.Attribute.
  • Make IBool interface of Bool more important than IFromUnicode so adapters registered for IBool take precendence over adapters registered for IFromUnicode.

3.5.3 (2009-03-10)

  • Make Choice and Bool fields implement IFromUnicode interface, because they do provide the fromUnicode method.
  • Change package’s mailing list address to zope-dev at, as zope3-dev at is now retired.
  • Fix package’s documentation formatting. Change package’s description.
  • Add buildout part that builds Sphinx-generated documentation.
  • Remove zpkg-related file.

3.5.2 (2009-02-04)

  • Made validation tests compatible with Python 2.5 again (hopefully not breaking Python 2.4)
  • Added an __all__ package attribute to expose documentation.

3.5.1 (2009-01-31)

  • Stop using the old old set type.
  • Make tests compatible and silent with Python 2.4.
  • Fix __cmp__ method in ValidationError. Show some side effects based on the existing __cmp__ implementation. See validation.txt
  • Make ‘repr’ of the ValidationError and its subclasses more sensible. This may require you to adapt your doctests for the new style, but now it makes much more sense for debugging for developers.

3.5.0a2 (2008-12-11)

  • Move zope.testing to “test” extras_require, as it is not needed for zope.schema itself.
  • Change the order of classes in SET_TYPES tuple, introduced in previous release to one that was in 3.4 (SetType, set), because third-party code could be dependent on that order. The one example is z3c.form’s converter.

3.5.0a1 (2008-10-10)

  • Added the doctests to the long description.
  • Removed use of deprecated ‘sets’ module when running under Python 2.6.
  • Removed spurious doctest failure when running under Python 2.6.
  • Added support to bootstrap on Jython.
  • Added helper methods for schema validation: getValidationErrors and getSchemaValidationErrors.
  • zope.schema now works on Python2.5

3.4.0 (2007-09-28)

Added BeforeObjectAssignedEvent that is triggered before the object field sets a value.

3.3.0 (2007-03-15)

Corresponds to the version of the zope.schema package shipped as part of the Zope 3.3.0 release.

3.2.1 (2006-03-26)

Corresponds to the version of the zope.schema package shipped as part of the Zope 3.2.1 release.

Fixed missing import of ‘VocabularyRegistryError’. See .

3.2.0 (2006-01-05)

Corresponds to the version of the zope.schema package shipped as part of the Zope 3.2.0 release.

Added “iterable” sources to replace vocabularies, which are now deprecated and scheduled for removal in Zope 3.3.

3.1.0 (2005-10-03)

Corresponds to the version of the zope.schema package shipped as part of the Zope 3.1.0 release.

Allowed ‘Choice’ fields to take either a ‘vocabulary’ or a ‘source’ argument (sources are a simpler implementation).

Added ‘TimeDelta’ and ‘ASCIILine’ field types.

3.0.0 (2004-11-07)

Corresponds to the version of the zope.schema package shipped as part of the Zope X3.0.0 release.

Release history Release notifications

History Node


History Node


History Node


History Node


History Node


History Node


History Node


History Node


History Node


History Node


History Node


History Node


History Node


History Node


History Node


History Node


History Node


This version
History Node


History Node


History Node


History Node


History Node


History Node


History Node


History Node


History Node


History Node


History Node


History Node


History Node


History Node


History Node


Download files

Download the file for your platform. If you're not sure which to choose, learn more about installing packages.

Filename, size & hash SHA256 hash help File type Python version Upload date
zope.schema-3.7.1.tar.gz (65.8 kB) Copy SHA256 hash SHA256 Source None Dec 25, 2010

Supported by

Elastic Elastic Search Pingdom Pingdom Monitoring Google Google BigQuery Sentry Sentry Error logging CloudAMQP CloudAMQP RabbitMQ AWS AWS Cloud computing Fastly Fastly CDN DigiCert DigiCert EV certificate StatusPage StatusPage Status page