Hooks to facilitate managing custom index values in Zope 2/CMF applications
This package provides primitives to help delegate ZCatalog indexing operations to adapters. It doesn’t do very much on its own, but can be used by catalog implementations that want to allow individual index values to be provided not by the object itself, but by separate adapters.
An indexer is a named adapter that adapts the type of an object and provides a value to be indexed when the catalog attempts to index the attribute with that name.
For example, let’s say we have two types, page and news item:
>>> from zope.interface import implements, Interface >>> from zope import schema>>> class IPage(Interface): ... text = schema.Text(title=u"Body text") >>> class Page(object): ... implements(IPage) ... def __init__(self, text): ... self.text = text>>> class INewsItem(Interface): ... summary = schema.TextLine(title=u"Short summary") ... story = schema.Text(title=u"Body text") ... audience = schema.TextLine(title=u"Audience")>>> class NewsItem(object): ... implements(INewsItem) ... def __init__(self, summary, story, audience): ... self.summary = summary ... self.story = story ... self.audience = audience
Now, pretend that our catalog had an index ‘description’, which for a page should contain the first 10 characters from the body text, and for a news item should contain the contents of the ‘summary’ field. Furthermore, there is an index ‘audience’ that should contain the value of the corresponding field for news items, in all uppercase. It should do nothing for pages.
We could write indexers for all of these like this
>>> from plone.indexer import indexer>>> @indexer(IPage) ... def page_description(object): ... return object.text[:10]>>> @indexer(INewsItem) ... def newsitem_description(object): ... return object.summary>>> @indexer(INewsItem) ... def newsitem_audience(object): ... return object.audience.upper()
These need to be registered as named adapters, where the name corresponds to the index name. In ZCML, that may be:
<adapter name="description" factory=".indexers.page_description" /> <adapter name="description" factory=".indexers.newsitem_description" /> <adapter name="audience" factory=".indexers.newsitem_audience" />
We can omit the ‘for’ attribute because we passed this to the @indexer decorator, and we can omit the ‘provides’ attribute because the thing returned by the decorator is actually a class providing the required IIndexer interface.
For the purposes of the ensuing tests, we’ll register these directly.
>>> from zope.component import provideAdapter >>> provideAdapter(page_description, name='description') >>> provideAdapter(newsitem_description, name='description') >>> provideAdapter(newsitem_audience, name='audience')
Testing your indexers (or calling them directly)
If you are writing tests for your indexers (as you should!), then you should be aware of the following:
When the @indexer decorator returns, it turns your function into an instance of type DelegatingIndexerFactory. This is an adapter factory that can create a DelegatingIndexer, which in turn will call your function when asked to perform indexing operations.
This means that you can’t just call your function to test the indexer. Instead, you need to instantiate the adapter and then call the delegating indexer with the portal root as the first argument. For example:
>>> test_page = Page(text=u"My page with some text") >>> page_description(test_page)() u'My page wi'
This will suffice in most cases. Note that there is actually a second parameter, catalog, which defaults to None. If you need to write an indexer that acts on catalog, you’ll need to register a conventional adapter, as described in the next section.
Other means of registering indexers
At the end of the day, an indexer is just a named multi-adapter from the indexable object (e.g. INewsItem or IPage above) and the catalog (usually portal_catalog in a CMF application) to IIndexer, where the name is the name of the indexed attribute in the catalog. Thus, you could register your indexers as more conventional adapters:
>>> from plone.indexer.interfaces import IIndexer >>> from Products.ZCatalog.interfaces import IZCatalog>>> from zope.interface import implements >>> from zope.component import adapts >>> class LengthIndexer(object): ... """Index the length of the body text ... """ ... implements(IIndexer) ... adapts(IPage, IZCatalog) ... ... def __init__(self, context, catalog): ... self.context = context ... self.catalog = catalog ... ... def __call__(self): ... return len(self.context.text)
We normally just use IZCatalog for the catalog adaptation, to apply to any catalog. However, if you want different indexers for different types of catalogs, there is an example later in this test.
You’d register this with ZCML like so:
<adapter factory=".indexers.LengthIndexer" name="length" />
Or in a test:
>>> provideAdapter(LengthIndexer, name="length")
If you’re only curious about how to write indexers, you can probably stop here. If you want to know more about how they work and how they are wired into a framework, read on.
Hooking up indexers to the framework
Here is a mock implementation of a ZCatalog.catalog_object() override, based on the one in Plone. We’ll use this for testing. We won’t bother with the full ZCatalog interface, only catalog_object(), and we’ll stub out a few things. This really is for illustration purposes only, to show the intended usage pattern.
In CMF 2.2, there is an IIndexableObject marker interface defined in Products.CMFCore.interfaces. We have a compatibility alias in this package for use with CMF 2.1.
>>> from OFS.interfaces import IItem >>> from Products.ZCatalog.interfaces import IZCatalog >>> from plone.indexer.interfaces import IIndexableObject >>> from zope.component import queryMultiAdapter>>> class FauxCatalog(object): ... implements(IZCatalog, IItem) ... ... def catalog_object(self, object, uid, idxs=): ... """Pretend to index 'object' under the key 'uid'. We'll ... print the results of the indexing operation to the screen . ... """ ... ... if not IIndexableObject.providedBy(object): ... wrapper = queryMultiAdapter((object, self,), IIndexableObject) ... if wrapper is not None: ... object = wrapper ... ... # Perform the actual indexing of attributes in the idxs list ... for idx in idxs: ... try: ... indexed_value = getattr(object, idx) ... if callable(indexed_value): ... indexed_value = indexed_value() ... print idx, "=", indexed_value ... except (AttributeError, TypeError,): ... pass
The important things here are:
- We attempt to obtain an IIndexableObject for the object to be indexed. This is just a way to get hold of an implementation of this interface (we’ll register one in a moment) and allow some coarse-grained overrides.
- Cataloging involves looking up attributes on the indexable object wrapper matching the names of indexes (in the real ZCatalog, this is actually decoupled, but let’s not get carried away). If they are callable, they should be called. This is just mimicking what ZCatalog’s implementation does.
This package comes with an implementation of an IIndexableObject adapter that knows how to delegate to an IIndexer. Let’s now register that as the default IIndexableObject wrapper adapter so that the code above will find it.
>>> from plone.indexer.wrapper import IndexableObjectWrapper >>> from plone.indexer.interfaces import IIndexableObject >>> provideAdapter(factory=IndexableObjectWrapper, adapts=(Interface, IZCatalog,), provides=IIndexableObject)
Seeing it in action
Now for the testing. First, we need a faux catalog:
>>> catalog = FauxCatalog()
Finally, let’s create some objects to index.
>>> page = Page(u"The page body text here") >>> news = NewsItem(u"News summary", u"News body text", u"Audience")
First of all, let’s demonstrate that our indexers work and apply only to the types for which they are registered.
>>> catalog.catalog_object(page, 'p1', idxs=['description', 'audience', 'length']) description = The page b length = 23>>> catalog.catalog_object(news, 'n1', idxs=['description', 'audience', 'length']) description = News summary audience = AUDIENCE
Our custom indexable object wrapper is capable of looking up workflow variables if the portal_workflow tool is available. For testing purposes, we’ll create a fake minimal workflow tool and stash it onto the fake catalog so that it can be found by getToolByName. In real life, it would of course be acquirable as normal.
>>> class FauxWorkflowTool(object): ... implements(IItem) ... def getCatalogVariablesFor(self, object): ... return dict(review_state='published', audience='Somebody') >>> catalog.portal_workflow = FauxWorkflowTool()
If we now index ‘review_state’, it will be obtained from the workflow variables. However, a custom indexer still overrides workflow variables.
>>> catalog.catalog_object(news, 'n1', idxs=['description', 'audience', 'review_state']) description = News summary audience = AUDIENCE review_state = published
Finally, if not adapter can be found, we fall back on getattr() on the object.
>>> catalog.catalog_object(page, 'p3', idxs=['description', 'text']) description = The page b text = The page body text here
Customising indexers based on the catalog type
It is possible to provide a custom indexer for a different type of catalog. To test that, let’s create a secondary catalog and mark it with a marker interface.
>>> from zope.interface import Interface >>> class IAlternateCatalog(Interface): ... pass >>> from zope.interface import alsoProvides >>> catalog2 = FauxCatalog() >>> alsoProvides(catalog2, IAlternateCatalog)
Let’s say that we did not want the news item audience uppercased here. We could provide a custom indexer for just this catalog:
>>> @indexer(INewsItem, IAlternateCatalog) ... def alternate_newsitem_audience(object): ... return object.audience.lower() >>> provideAdapter(alternate_newsitem_audience, name='audience')
This does not affect the first catalog:
>>> catalog.catalog_object(news, 'n1', idxs=['description', 'audience', 'length']) description = News summary audience = AUDIENCE
However, the second catalog gets the audience in lowercase.
>>> catalog2.catalog_object(news, 'n1', idxs=['description', 'audience', 'length']) description = News summary audience = audience
Interfaces provided by the wrapper
The indexable object wrapper has one particular feature: instances of the wrapper will provide the same interfaces as instances of the wrapped object. For example:
>>> from plone.indexer.interfaces import IIndexableObjectWrapper, IIndexableObject >>> wrapper = IndexableObjectWrapper(page, catalog) >>> IIndexableObjectWrapper.providedBy(wrapper) True >>> IIndexableObject.providedBy(wrapper) True >>> IPage.providedBy(wrapper) True >>> INewsItem.providedBy(wrapper) False>>> wrapper = IndexableObjectWrapper(news, catalog) >>> IIndexableObjectWrapper.providedBy(wrapper) True >>> IPage.providedBy(wrapper) False >>> INewsItem.providedBy(wrapper) True
It is possible to obtain the wrapped object from the wrapper:
>>> wrapper = IndexableObjectWrapper(page, catalog) >>> wrapper._getWrappedObject() is page True
- Changed the @indexer decorator to maintain the information about the wrapped function (__doc__, __module__, __name__, etc). [dokai]
- Relicense under modified BSD license; per Plone Foundation board approval on 2012-05-31. See: http://plone.org/foundation/materials/foundation-resolutions/plone-framework-components-relicensing-policy [supton]
- Add MANIFEST.in. [WouterVH]
1.0 - 2010-07-18
- Fixed reSt markup in the changelog. [hannosch]
- Update license to GPL version 2 only. [hannosch]
1.0rc2 - 2009-04-05
- Added _getWrappedObject() method to get hold of the underlying object. Note that this means you can’t have an index/metadata column with this name. [optilude]
- Corrected IZCatalog import location to point to the interfaces module. [hannosch]
1.0rc1 - 2009-03-26
- Updated the interface to match the developments of similar functionality on CMF trunk. This means that indexers are now multi-adapters on (object, catalog), and the keyword arguments (including the implicit ‘portal’ parameter) are gone. [optilude]
1.0a1 - 2009-03-05
- Initial release