Traversing

Description

Plone content is organized as a tree. Traversing means looking up content from this tree by path. When HTTP request hits a Plone server, Plone will traverse the corresponding content item and its view function by URI.

Introduction

In Plone, all content is mapped to a single tree: content objects, user objects, templates, etc. Even most object methods are directly mapped to HTTP-accessible URIs.

Each object has a path depending on its location. Traversal is a method of getting a handle on a persistent object in the ZODB object graph from its path.

Traversal can happen in two places:

  • When an HTTP request hits the server, the method on the object which will generate the HTTP response is looked up using traversal.
  • You can manually traverse the ZODB tree in your code to locate objects by their path.

When an HTTP request is being published the traversing happens in ZPublisher.BaseRequest.traverse

… but Zope includes other traversers, like unrestrictedTraverse() in the OFS module. Different traversing methods behave differently and may fire different events.

Object ids

Each content object has an id string which identifies the object in the parent container. The id string is visible in the browser address bar when you view the object. Ids are also visible in the Management Interface.

Besides id strings, the content objects have Unique Identifiers, or UID, which do not change even if the object is moved or renamed.

Though it’s technically possible for ids to contain spaces or slashes, this is seldom a good idea, as it complicates working with ids in various situations.

Path

The Zope path is the location of the object in the object graph. It is a sequence of id components from the parent node(s) to the child separated by slashes.

Note

A path need not always be a sequence of object ids. During traversal, an object may consume subsequent path elements, interpreting them however it likes.

Example:

documentation/howTos/myHowTo

Exploring Zope application server

You can use the Management Interface to explore the content of your Zope application server:

  • Sites
  • Folders within the sites
  • …and so on

The Management Interface does not expose individual attributes. It only exposes traversable content objects.

Attribute traversing

Zope exposes child objects as attributes.

Example:

# you have obtained the plone.org portal root object somehow and it's
# stored in local variable "portal"

documentation = portal.documentation
howTos = getattr(portal, "how-to") # note that we need use getattr because dash is invalid in syntax
myHowTo = getattr(howTos, "manipulating-plone-objects-programmatically")

Container traversing

Zope exposes child objects as container accessor.

Example:

# you have obtained the plone.org portal root object somehow and it's
# stored in a local variable "portal"

documentation = portal["documentation"]
howTos = documentation["how-to"]
myHowTo = howTos["manipulating-plone-objects-programmatically"]

Traversing by full path

Any content object provides the methods restrictedTraverse() and unrestrictedTraverse(). See Traversable.

Security warning: restrictedTraverse() executes with the privileges of the currently logged-in user. An Unauthorized exception is raised if the code tries to access an object for which the user lacks the Access contents information and View permissions.

Example:

myHowTo = portal.restrictedTraverse("documentation/howTos/myHowTo")

# Bypass security
myHowTo = portal.unrestrictedTraverse("documentation/howTos/myHowTo")

Warning

restrictedTraverse()/unrestrictedTraverse() does not honor IPublishTraverse adapters. Read more about the issue in this discussion.

Getting the object path

An object has two paths:

  • The physical path is the absolute location in the current ZODB object graph. This includes the site instance name as part of it.
  • The virtual path is the object location relative to the Plone site root.

Path mangling warning: Always store paths as virtual paths, or persistently stored paths will corrupt if you rename your site instance.

See Traversable.

Getting physical path

Use getPhysicalPath(). Example:

path = portal.getPhysicalPath() # returns "plone"

Getting virtual path

For content items you can use absolute_url_path() from OFS.Traversable:

path = context.absolute_url_path()

Map physical path to virtual path using HTTP request object physicalPathToVirtualPath(). Example:

request = self.request # HTTPRequest object

path = portal.document.getPhysicalPath()

virtual_path = request.physicalPathToVirtualPath(path) # returns "document"

Note

The virtual path is not necessarily the path relative to the site root, depending on the virtual host configuration.

Getting item path relative to the site root

There is no a direct, easy way to accomplish this.

Example:

from zope.component import getMultiAdapter

def getSiteRootRelativePath(context, request):
    """ Get site root relative path to an item

    @param context: Content item which path is resolved

    @param request: HTTP request object

    @return: Path to the context object, relative to site root, prefixed with a slash.
    """

    portal_state = getMultiAdapter((context, request), name=u'plone_portal_state')
    site = portal_state.portal()

    # Both of these are tuples
    site_path = site.getPhysicalPath();
    context_path = context.getPhysicalPath()

    relative_path = context_path[len(site_path):]

    return "/" + "/".join(relative_path)

Getting canonical object (breadcrumbs, visual path)

The visual path is presented in the breadcrumbs. It is how the site visitor sees the object path.

It may differ from the physical path:

  • The default content item is not shown in the visual path.
  • The default view is not shown in the visual path.

The canonical object is the context object which the user sees from the request URL:

Example:

context_helper = getMultiAdapter((context, self.request), name="plone_context_state")
canonical = context_helper.canonical_object()

Getting object URL

Use absolute_url(). See Traversable.

URL mangling warning: absolute_url() is sensitive to virtual host URL mappings. absolute_url() will return different results depending on if you access your site from URLs http://yourhost/ or http://yourhost:8080/Plone. Do not persistently store the result of absolute_url().

Example:

url = portal.absolute_url() # http://nohost/plone in unit tests

Getting the parent

The object parent is accessible is acquisition chain for the object is set.

Use aq_parent:

parent = object.aq_parent

The parent is defined as __parent__ attribute of the object instance:

object.__parent__ = object.aq_parent

__parent__ is set when object’s __of__() method is called:

view = MyBrowserView(context, request)
view = view.__of__(context) # Inserts view into acquisition chain and acquisition functions become available

Getting all parents

Example:

def getAcquisitionChain(object):
    """
    @return: List of objects from context, its parents to the portal root

    Example::

        chain = getAcquisitionChain(self.context)
        print "I will look up objects:" + str(list(chain))

    @param object: Any content object
    @return: Iterable of all parents from the direct parent to the site root
    """

    # It is important to use inner to bootstrap the traverse,
    # or otherwise we might get surprising parents
    # E.g. the context of the view has the view as the parent
    # unless inner is used
    inner = object.aq_inner

    iter = inner

    while iter is not None:
        yield iter

        if ISiteRoot.providedBy(iter):
           break

        if not hasattr(iter, "aq_parent"):
            raise RuntimeError("Parent traversing interrupted by object: " + str(parent))

        iter = iter.aq_parent

Getting the site root

You can resolve the site root if you have the handle to any context object.

Using portal_url tool

Example:

from Products.CMFCore.utils import getToolByName

# you know some object which is referred as "context"
portal_url = getToolByName(context, "portal_url")
portal = portal_url.getPortalObject()

You can also do shortcut using acquisition:

portal = context.portal_url.getPortalObject()

Note

Application code should use the getToolByName method, rather than simply acquiring the tool by name, to ease forward migration (e.g., to Zope3).

Using getSite()

Site is also stored as a thread-local variable. In Zope each request is processed in its own thread. Site thread local is set when the request processing starts.

You can use this method even if you do not have the context object available, assuming that your code is called after Zope has traversed the context object once.

Example:

from zope.component.hooks import getSite

site = getSite() # returns portal root from thread local storage

Note

Due to the fact that Plone does not show the default content item as a separate object, the page you are viewing in the browser from the site root URL is not necessary the root item itself. For example, in the default Plone installation this URL internally maps to Page whose id is front-page and you can still query the actual parent object which is the site root.

If you need to traverse using user visible breadcrumbs, see how breadcrumbs viewlet code does it.

Traversing back to the site root

Sometimes getSite() or portal_url are not available, but you still have the acquisition chain intact. In these cases you can simply traverse parent objects back to the site root by iterating over the aquisition-chain or using the aq_parent accessor:

from Products.CMFCore.interfaces import ISiteRoot

def getSite(context):

    if not ISiteRoot.providedBy(context):
        return context
    else:
        for item in context.aq_chain:
            if ISiteRoot.providedBy(item):
                return item

Checking for the site root

You can check if the current context object is Plone the site root:

from Products.CMFCore.interfaces import ISiteRoot

if ISiteRoot.providedBy(context):
    # Special case
else:
    # Subfolder or on a page

Getting Zope application server handle

You can also access other sites within the same application server from your code.

Example:

app = context.restrictedTraverse('/') # Zope application server root
site = app["plone"] # your plone instance
site2 = app["mysiteid"] # another site

Acquisition effect

Sometimes traversal can give you attributes which actually do not exist on the object, but are inherited from the parent objects in the persistent object graph. See acquisition.

Default content item

Default content item or view sets some challenges for the traversing, as the object published path and internal path differ.

Below is an example to get the folder of the published object (parent folder for the default item) in page templates:

<div tal:define="folder context/@@plone_context_state/canonical_object"
     tal:condition="python:hasattr(folder, 'carousel') and
                           hasattr(folder['carousel'],
                           'carouselText')">xxx</div>

Checking if an item is the site front page

Example:

from zope.component import getMultiAdapter

def is_front_page(self):
    """Check if we are in the context of a front page."""
    context_helper = getMultiAdapter((self.context, self.request), name='plone_context_state')
    return context_helper.is_portal_root()

Custom traversal

There exist many ways to make your objects traversable:

  • __getitem__() which makes your objects act like Python dictionary. This is the simplest method and recommended.
  • IPublishTraverse interface. There is an example below and works for making nice urls and path munging.
  • ITraversable interface. You can create your own traversing hooks. zope.traversing.interfaces.ITraversable provides an interface traversable objects must provider. You need to register ITraversable as adapter for your content types. This is only for publishing methods for HTTP requests.
  • __bobo_traverse__() which is an archaic method from the early 2000s.

Warning

Zope traversal is a minefield. There are different traversers. One is the ZPublisher traverser which does HTTP request looks. One is OFS.Traversable.unrestrictedTraverse() which is used when you call traverse from Python code. Then another case is zope.tales.expression.PathExpr which uses a really simple traverser.

Warning

If an AttributeError is risen inside a traverse() function bad things happen, as Zope publisher specially handles this and raises a NotFound exception which will mask the actual problem.

Example using __getitem__():

class Viewlets(BrowserView):
    """ Expose arbitrary viewlets to traversing by name.
    Exposes viewlets to templates by names.
    Example how to render plone.logo viewlet in arbitrary template
    code point::

        <div tal:content="context/@@viewlets/plone.logo" />

    """

    ...

    def __getitem__(self, name):
        """
        Allow travering intoviewlets by viewlet name.

        @return: Viewlet HTML output

        @raise: ViewletNotFoundException if viewlet is not found
        """
        viewlet = self.setupViewletByName(name)
        if viewlet is None:
            active_layers = [ str(x) for x in self.request.__provides__.__iro__ ]
            active_layers = tuple(active_layers)
            raise ViewletNotFoundException("Viewlet does not exist by"
                "name %s for the active theme layer set %s."
                "Probably theme interface not registered in "
                "plone.browserlayers. Try reinstalling the theme."
                % (name, str(active_layers)))

        viewlet.update()
        return viewlet.render()

Example using IPublishTraverse:

from Products.Five.browser import BrowserView
from zope.publisher.interfaces import IPublishTraverse
from zope.interface import implementer
from zope.component import getMultiAdapter
from AccessControl import getSecurityManager
from AccessControl import Unauthorized
from plone import api

@implementer(IPublishTraverse)
class MyUser(BrowserView):
    """Registered as a browser view at '/user', collect the username and
    view name from the url, check security, and display that page. For
    example, '/user/jjohns/log' will look up the log view for user
    'jjohns'
    """
    path = []

    def publishTraverse(self, request, name):
        # stop traversing, we have arrived
        request['TraversalRequestNameStack'] = []
        # return self so the publisher calls this view
            return self


    def __init__(self, context, request):
        """Once we get to __call__, the path is lost so we
        capture it here on initialization
        """
        super(MyUser, self).__init__(context, request)
        self.section = 'profile-latest' # default page
        if len(request.path) == 2:
            [self.section, profileid] = request.path
        elif len(self.request.path) == 1:
            self.section = request.path[0]

    def __call__(self):
        # do the permission check here, now that Zope has set
        # up the security context. It can't be checked in __init__
        # because the security manager isn't set up on traverse
        self.checkPermission()

        # XXX: still need to check the permission of the view
        try:
            view = api.content.get_view(self.section,
                                        self.context,
                                        self.request)
        except api.exc.InvalidParameterError:
            # just return the default view
            view = api.content.get_view('profile-latest',
                                        self.context,
                                        self.request)
        return view()

    def checkPermission(self):
        """You might want to do other stuff"""
        raise Unauthorized

More information:

Traverse events

Use zope.traversing.interfaces.IBeforeTraverseEvent for register a traversing hook for Plone site object or such.

Example:

from Products.CMFCore.interfaces import ISiteRoot
from zope.traversing.interfaces import IBeforeTraverseEvent
from five import grok

@grok.subscribe(ISiteRoot, IBeforeTraverseEvent)
def check_redirect(site, event):
    """
    """
    request = event.request

    # XXX: To something

Use ZPublisher.BeforeTraverse to register traverse hooks for any objects.

Todo

Example - not sure if before travese hooks are persistent or not

Advanced traversing with search conditions

All Plone content should exist in the portal_catalog. Catalog provides fast query access with various indexes to the Plone content.

Other resources

See object publishing.