Re: [xml-dev] XML-enabled databases, XQuery APIs

["Ken North" <kennorth@sbcglobal.net>]

>
> I would have expected that in a system using shredding ...

This paper discusses shredding and query optimization. The example data is from
the Internet Movie Database and there is discussion of alternative mappings.

"Adaptive XML Shredding: Architecture, Implementation, and Challenges"
http://www.cse.ogi.edu/~juliana/pub/eextt2002.pdf

The design component does a statistical analysis of the document. Meyer and
Klettke proposed that approach several years ago for developing the optimal
design for object-relational databases mapped from XML DTDs. This paper
discusses LegoDB but the authors say the query optimization algorithm and
optimization techniques apply to a variety of SQL query optimizers.

"The main task of the design component is to generate a target relational schema
to store the input XML document. In adaptive shredding systems such as LegoDB,
this component is rather complex. It takes into account information about the
target application to both generate a space of possible mappings and evaluate
the effectiveness of the derived mappings."
...
The subsequent experiments reveal that the application overhead of extracting
relevant data from the XML document can be kept rather small. For the simplest
queries, the XQuery overhead was less than the parser overhead. For more complex
queries, the experiments revealed a clear shift in overhead from parsing to the
XQuery application, as indicated by the decrease in speedups between expat and
the binary encoding parsers.

LegoDB (XML to SQL Mapping Engine)
http://www-db-out.bell-labs.com/legodb/