source: main/waeup.sirp/trunk/src/waeup/sirp/catalog.txt @ 8794

:mod:`waeup.sirp.catalog` -- Cataloging support SIRP
****************************************************

.. module:: waeup.sirp.catalog

Components that support cataloging and searching objects inside a
SIRP site.

.. :doctest:
.. :layer: waeup.sirp.testing.SIRPUnitTestLayer

.. contents::

Classes
=======

:class:`SIRPQuery`
------------------

.. class:: SIRPQuery()

   .. attribute:: grok.implements(hurry.query.interfaces.IQuery)

   A `hurry.query.query.Query` compatible query that also supports
   retrival of plain ``Bree`` result sets as used inside a catalog.

   Like `hurry.query.query.Query` objects, `SIRPQuery` is some kind
   of a meta query or 'compound query' that can give the cataloged
   objects (or their int ids) matching one or more 'subqueries'.

   This way you can search for objects (or their int ids) that match
   several criteria at the same time. See ``examples`` section below.

   A singleton instance of this class is also available as global
   utility.

   .. method:: searchResults(query)

     Get the cataloged objects determined by ``query``.

   .. method:: apply(query)

     Get the list of int ids (a `BTree` result set) for objects
     determined by ``query``.

     The list of int ids is less expensive to compute than the
     complete search results and sufficient, for instance, when you
     only need the number of objects that match a query and not the
     objects themselves.

Examples
========

Getting a general query object
------------------------------

We can get a SIRPQuery object by asking for an unnamed global utility
implementing `hurry.query.interfaces.IQuery`:

    >>> from hurry.query.interfaces import IQuery
    >>> from zope.component import getUtility
    >>> q = getUtility(IQuery)
    >>> q
    <waeup.sirp.catalog.SIRPQuery object at 0x...>

This query can get 'subqueries' and delivers the objects found or
their ids. To show this we have to setup a catalog with some entries.


Setting up a catalog and feeding it
-----------------------------------

    >>> from zope.catalog.interfaces import ICatalog
    >>> from zope.catalog.catalog import Catalog
    >>> mycat = Catalog()

We register this catalog with the component architechture as a utility
named 'mycatalog':

    >>> from zope.component import provideUtility
    >>> provideUtility(mycat, ICatalog, 'mycatalog')

We setup a special content type whose instances we will catalog later:

    >>> from zope.interface import Interface, Attribute, implements
    >>> from zope.container.contained import Contained
    >>> class IMammoth(Interface):
    ...   name = Attribute('name')
    ...   age = Attribute('age')

    >>> class Mammoth(Contained):
    ...   implements(IMammoth)
    ...   def __init__(self, name, age):
    ...     self.name = name
    ...     self.age = age
    ...   def __cmp__(self, other):
    ...     return cmp(self.name, other.name)

By including the __cmp__ method we make sure search results can be
stably sorted.

We also setup a `zope.intid.interfaces.IIntIds` utility. This is not
necessary for plain catalogs, but when we want to use SIRPQuery (or
`hurry.query.query.Query` objects), as to get a unique mapping from
objects (stored in ZODB) to integer numbers (stored in catalogs),
these query objects lookup a global IIntIds utiliy:

    >>> from zope import interface
    >>> import zope.intid.interfaces
    >>> class DummyIntId(object):
    ...     interface.implements(zope.intid.interfaces.IIntIds)
    ...     MARKER = '__dummy_int_id__'
    ...     def __init__(self):
    ...         self.counter = 0
    ...         self.data = {}
    ...     def register(self, obj):
    ...         intid = getattr(obj, self.MARKER, None)
    ...         if intid is None:
    ...             setattr(obj, self.MARKER, self.counter)
    ...             self.data[self.counter] = obj
    ...             intid = self.counter
    ...             self.counter += 1
    ...         return intid
    ...     def getObject(self, intid):
    ...         return self.data[intid]
    ...     def __iter__(self):
    ...         return iter(self.data)
    >>> intid = DummyIntId()
    >>> from zope.component import provideUtility
    >>> provideUtility(intid, zope.intid.interfaces.IIntIds)

Now we can catalog some mammoths. Here we create a herd and catalog
each item of it:

    >>> from zope.catalog.field import FieldIndex
    >>> mycat['mammoth_name'] = FieldIndex('name', IMammoth)
    >>> mycat['mammoth_age'] = FieldIndex('age', IMammoth)

    >>> herd = [
    ...   Mammoth(name='Fred', age=33),
    ...   Mammoth(name='Hank', age=30),
    ...   Mammoth(name='Wilma', age=28),
    ... ]

    >>> for mammoth in herd:
    ...   mycat.index_doc(intid.register(mammoth), mammoth)


Searching for result sets
-------------------------

Finally we can perform queries:

    >>> from hurry.query import Eq
    >>> from zope.component import getUtility
    >>> subquery1 = Eq(('mycatalog', 'mammoth_name'), 'Fred')

The latter means: search for objects whose name is ``'Fred'`` in the
``mammoth_name`` index of a catalog registered as a utility named
``mycatalog``.

    >>> from hurry.query import Between
    >>> subquery2 = Between(('mycatalog', 'mammoth_age'), 30, 33)

This means: ask for objects cataloged in an index named 'mammoth_age',
whose cataloged value is between 30 and 33 (including this values).

    >>> r1 = q.apply(subquery2)
    >>> r1
    IFSet([0, 1])

Using ``apply()`` above, we get a set of values stored in an
``IFBTree``:

    >>> type(r1)
    <type 'BTrees.IFBTree.IFSet'>

``IFBTree`` objects implement a rather efficient integer to float
mapping where also integers are allowed as values. For each object
found (i.e. mammoths whose age is between 30 and 33), we get the
number of its entry.

To get the real object, we can use intids here, because we setup an
appropriate IIntIds utility before:

    >>> [intid.getObject(x).name for x in r1]
    ['Fred', 'Hank']

We can (and should) also use the `searchResults()` method explained
below to do that.

Retrieving BTree sets can, however, make sense, if you want to know
only the number of results for a particular query or whether there are
results at all in a more efficient way:

    >>> len(r1)
    2

Searching for objects
---------------------

Very often we don't want to know the catalog-internal 'ids' of
searched objects but the objects themselves.

This can be done by using the ``searchResults`` method of
``SIRPQuery``:

    >>> r2 = q.searchResults(subquery1)
    >>> r2
    <zope.catalog.catalog.ResultSet instance at 0x...>

    >>> list(r2)
    [<Mammoth object at 0x...>]

We got one result item, we can immediately ask for further infos. To
access a result item by its index number, we have to turn the
ResultSet into an ordinary list before:

    >>> entry = list(r2)[0]
    >>> entry.name, entry.age
    ('Fred', 33)

We can also use ``subquery2`` as above:

    >>> r3 = q.searchResults(subquery2)
    >>> [(x.name, x.age) for x in r3]
    [('Fred', 33), ('Hank', 30)]

or use both queries at once:

    >>> r4 = q.searchResults(subquery1 & subquery2)
    >>> [(x.name, x.age) for x in r4]
    [('Fred', 33)]

which will give us, of course, the same result set as with subquery1.