34. Relations

You can model relationships between content items by placing them in a hierarchy (a folder speakers containing the (folderish) speakers and within each speaker the talks) or by linking them to each other in Richtext-Fields. But where would you store a talk that two speakers give together together?

Relations allow developers to model relationships between objects without a links or a hierarchy. The behavior plone.app.relationfield.behavior.IRelatedItems provides the field Related Items in the tab Categorization. That field simply says a is somehow related to b.

By using custom relations you can model your data in a much more meaningful way.

Creating relations in a schema

Relate to one item only.

from plone.app.vocabularies.catalog import CatalogSource
from z3c.relationfield.schema import RelationChoice
from z3c.relationfield.schema import RelationList

evil_mastermind = RelationChoice(
    title=_(u'The Evil Masterimind'),

Relate to multiple items.

from z3c.relationfield.schema import RelationChoice
from z3c.relationfield.schema import RelationList

minions = RelationList(

We can see that the code for the behavior IRelatedItems does exactly the same.

Instead of using a named vocabulary we can also use source:

from plone.app.vocabularies.catalog import CatalogSource
from z3c.relationfield.schema import RelationChoice
from z3c.relationfield.schema import RelationList

minions = RelationList(
    title=_(u'Talks by this speaker'),
        source=CatalogSource(portal_type=['one_eyed_minion', 'minion'])),

To CatalogSource you can pass the same argument that you use for catalog-queries. This makes it very flexible to limit relateable items by type, path, date etc.

For even more flexibility you can create your own dynamic vocabularies.

Creating Relationfields through the web

It is surprisingly easy to create RelationFields through the web

  • In the dexterity schema-editor add a new field and select Relation List or Relation Choice, depending on wether you want to relate to multiple items or not.
  • When configuring the field you can even select the content-type the relation should be limited to.

When you click on Edit xml field model you will see the fields in the xml-schema:


<field name="boss" type="z3c.relationfield.schema.RelationChoice">


<field name="underlings" type="z3c.relationfield.schema.RelationList">
  <value_type type="z3c.relationfield.schema.RelationChoice">
    <title i18n:translate="">Relation Choice</title>
      <element>News Item</element>

The stack

Relations are based on zc.relation. This package allows to store transitive and intransitive relationships. It allows for complex relationships and searches along them. Because of this functionality, the package is a bit complicated.

The package zc.relation provides its own catalog, a relation catalog. This is a storage optimized for the queries needed. zc.relation is sort of an outlier with regards to zope documentation. It has extensive documentation, with a good level of doctests for explaining things.

You can use zc.relation to store the objects and its relations directly into the catalog. But the additional packages that make up the relation functionality don’t use the catalog this way.

We want to work with schemas to get auto generated forms. The logic for this is provided by the package z3c.relationfield. This package contains the RelationValue object and everything needed to define a relation schema, and all the code that is necessary to automatically update the catalog.

A RelationValue Object does not reference all objects directly. For the target, it uses an id it gets from the IntId Utility. This id allows direct recovery of the object. The source object stores it directly.

Widgets are provided by plone.app.z3cform and some converters are provided by plone.app.relationfield. The widget that Plone uses can also store objects directly. Because of this, the following happens when saving a relation via a form:

  1. The html shows some nice representation of selectable objects.
  2. When the user submits the form, selected items are submitted by their UUIDs.
  3. The Widget retrieves the original object with the UUID.
  4. Some datamanager gets another unique ID from an IntID Tool.
  5. The same datamanager creates a RelationValue from this id, and stores this relation value on the source object.
  6. Some Event handlers update the catalogs.

You could delete a Relation like this delattr(rel.from_object, rel.from_attribute)

This is a terrible idea by the way, because when you define in your schema that one can store multiple RelationValues, your Relation is stored in a list on this attribute.

Relations depend on a lot of infrastructure to work. This infrastructure in turn depends a lot on event handlers being thrown properly. When this is not the case things can break. Because of this, there is a method isBroken which you can use to check if the target is available.

There are alternatives to using Relations. You could instead just store the UUID of an object. But using real relations and the catalog allows for very powerful things. The simplest concrete advantage is the possibility to see what links to your object.

The builtin linkintegrity-feature of Plone 5 is also built using relations.


RelationValue objects have a fairly complete API. For both target and source, you can receive the IntId, the object and the path. On a RelationValue, the terms source and target aren’t used. Instead, they are from and to. So the API for getting the target is:

  • to_id
  • to_path
  • to_object

In addition, the relation value knows under which attribute it has been stored as from_attribute. It is usually the name of the field with which the relation is created. But it can also be the name of a relation that is created by code, e.g. linkintegrity-relations (isReferencing) or the relation between a working copy and the original (iterate-working-copy).

Accessing relations and backrelations from code

If you want to find out what objects are related to each other, you use the relation catalog. Here is a convenience-method that allows you to find all kinds of relations.

from zc.relation.interfaces import ICatalog
from zope.component import getUtility
from zope.intid.interfaces import IIntIds
from plone.app.linkintegrity.handlers import referencedRelationship

def example_get_backlinks(obj):
    backlinks = []
    for rel in get_backrelations(attribute=referencedRelationship):
        if rel.isBroken():
                                  title='broken reference',
            obj = rel.from_object
    return backlinks

def get_relations(obj, attribute=None, backrefs=False):
    """Get any kind of references and backreferences"""
    int_id = get_intid(obj)
    if not int_id:
        return retval

    relation_catalog = getUtility(ICatalog)
    if not relation_catalog:
        return retval

    query = {}
    if attribute:
        # Constrain the search for certain relation-types.
        query['from_attribute'] = attribute

    if backrefs:
        query['to_id'] = int_id
        query['from_id'] = int_id

    return relation_catalog.findRelations(query)

def get_backrelations(obj, attribute=None):
    return get_relations(obj, attribute=attribute, backrefs=True)

def get_intid(obj):
    """Return the intid of an object from the intid-catalog"""
    intids = component.queryUtility(IIntIds)
    if intids is None:
    # check that the object has an intid, otherwise there's nothing to be done
        return intids.getId(obj)
    except KeyError:
        # The object has not been added to the ZODB yet