Hypatia

------------------

- Add ``check_query`` method to text index to allow for checking if a search
term is parseable.

------------------

- Added support for Python 3.2 / 3.3.

- Fix signature of TextIndex.sort (it would fail when e.g. raise_unsortable was
passed).

- Add the a ``sort_type`` keyword argument to ``IIndexSort.sort`` and
``IResultSet.sort`` methods. This value can be passed by calling code to
control the type of sorting used.

- Add two constants: ``hypatia.interfaces.STABLE`` and
``hypatia.interfaces.OPTIMAL``. These can be used as explicit arguments to
the ``IIndexSort.sort`` and ``IResultSet.sort`` ``sort_type`` parameter to
control the stability of sorting.

- The constructor of ``IResultSet`` now accepts a ``sort_type`` keyword
argument.

- The ResultSet constructed by ``IResultSet.sort`` will be passed the value
``hypatia.interfaces.STABLE`` in its constructor to ensure that the second
and subsequent sorts of the result set will be done as a stable sort, unless
an explicit ``sort_type`` value is passed to that second sort.

------------------

- Add IResultSet interface definition.

- Normalize keyword argument ordering of IIndexSort.sort and IResultSet.sort.

- Add an argument ``raise_unsortable`` to IIndexSort.sort and IResultSet.sort
methods. By default this is ``True``. It means that iterating over the
results returned by one of these methods *may* raise a
``hypatia.exc.Unsortable`` exception when a member of the docids passed in
cannot be sorted by the index used to do the sort (e.g. a value for the docid
is not present in the index). It defaults to ``True``, which changes the
default behavior of indexes. To get the old default behavior back, pass
``False`` for this value. Alternately, write code like this::

from hypatia.exc import Unsortable

ids = []
results = resultset.sort(someindex)
try:
for id in results:
ids.append(id)
except Unsortable as e:
unsorted = e.docids
ids.extend(unsorted)

------------------

- Optimize ``index_doc`` implementations of field and keyword index in cases
where the discriminator returns the default.

- Remove code from ``hypatia.path``. This package no longer supports
PathIndex.

- Remove ``interfaces.IIndexQuery`` interface. It was never relevant, as
indices cannot be expected to implement all of its methods, only the ones
which apply to each index.

- ``BaseIndexMixin`` no longer supplies default implementation of applyFoo
methods which raise NotImplementedError. Each index is now responsible for
implementing all of its own applyFoo methods. This is in the interest of
fidelity with new query methods such as ``eq``, which are similarly not
implemented in the base.

- Indexes are now compelled to implement a ``qname`` method for use by
queries.

- ``DoesNotContain`` query renamed to ``NotContains`` for symmetry with other
negated query names.

- New index methods: ``eq``, ``noteq``, ``ge``, ``le``, ``lt``, ``gt``,
``any``, ``notany``, ``all``, ``notall``, ``inrange``, ``notinrange``,
``contains``, ``notcontains``. These methods return query objects. Ex::

catalog['flavors'].eq('peach')

- Query objects refactored internally to deal in index objects rather than
index names.

- The ``query.parse_query`` function now requires a ``catalog`` argument.

- Query objects now supply an .execute method which returns a ResultSet.

- ResultSet objects are returned from .execute. They represent a set of
docids; they are iterable and have various methods for obtaining single
objects (like ``one``, ``first``) and sorting (``sort``).

- All Query objects now have a ``flush`` method which accepts arbitrary
positional and keyword arguments. Calling the ``flush`` method of a query
object will cause the ``flush`` method of all indexes participating in the
query with the value passed to Query.flush with the same positional and
keyword arguments. This is to support Substance D upstream, which may
require indexes to be flushed before a query happens.

- Add a ``document_repr`` method to all indexes which accepts a docid and
returns a string represnting the index's knowledge about that docid.

------------------

- This version of the code is incompatible with indexes produced by 0.1a1.
There is no upgrade script. Shame on you for using software with a 0.1a1
version number and expecting backwards compatibility.

- Add `hypatia.catalog.CatalogQuery.sort` API for sorting external sets
of docids based on index values.

- Add ``IIndexEnumeration`` interface, which all indexes must support.
This implied the following backwards incompatibilities:

- New interface methods: docids, docids_count, indexed, indexed_count,
not_indexed and not_indexed_count.

- documentCount method renamed to indexed_count.

- wordCount method renamed to word_count.

- Remove unused INBest interface.

- IIndexInjection interface ``clear`` method renamed to ``reset`` to prevent
confusion with dictionary ``clear`` (catalog is often dictionarylike).
Catalog ``clear_indexes`` method replaced with ``reset``.

-----

- Initial release: fork of repoze.catalog and zope.index, combined.