Table of contents
In order to pass
xapian.Document objects to
or from remote databases, Xapian includes support for serialising these objects
to binary strings, and then converting these strings back into objects. This
support may be accessed directly, and used for storing persistent
representations of such objects. The representations used are not architecture
dependent, so you can successfully unserialise an object on a machine with a
different word size or endianness to the machine it was serialised on.
Be aware that the serialised representation may change between release series, so if you’re using serialised objects for long term storage you will need a strategy for dealing with this. Changes to the representation will be clearly noted in the release notes.
If you have a
xapian.Document object which you want to serialise,
you can call the
xapian.Document.serialise() method on it, which
returns a string.
Documents are often lazily fetched from databases: this method will first force the full document contents to be fetched from the database, in order to serialise them. The serialised document will have identical contents (data, terms, positions, values) to the original document.
To get a document from a serialised form, call the static
xapian.Document.unserialise() method, passing it the string returned
serialise(), which will give you a new
Serialisation of queries is very similar to serialisation of documents: there
xapian.Query.serialise() method to produce a serialised Query,
and a corresponding
xapian.Query.unserialise() method to produce a
xapian.Query object from a serialised representation.
However, there is a wrinkle. A query can contain arbitrary user-defined
xapian.PostingSource subqueries. In order to serialise and
unserialise such queries, all the
xapian.PostingSource subclasses used
in the query must implement the
methods (see Posting sources for details).
In addition, a special form of unserialise must be used which takes a
xapian.Registry object as an additional parameter, which must know
all the custom posting sources used in the query. You need to call
xapian.Registry.register_posting_source() to register each such
xapian.Registry objects always know about built-in posting
sources (such as
xapian.ValueWeightPostingSource), so you don’t need
register_posting_source() for them.