On Querying Data and Metadata in Multiversion Data Warehouse - Data Warehousing Design and Advanced Engineering Applications - page 212

Querying Real and
Alternative DW Versions

Case 4: Different attribute names and
domains. Differences in the names of attributes
existing in a table in a parent and a child DW
version are registered in a dedicated dictionary
table. Corresponding attributes of two tables in two
adjacent DW versions may differ also with respect
to their domains. In order to compare the values of
such attributes, their values have to be convertable
to each other. In the MVDW , conversion methods
are used for this purpose. As an example we may
consider changing prices of products from Polish
Zloty to Euro. Appropriately defined conversion
methods from PLN to EUR and vice versa will
allow to compare sales of products expressed in
these two currencies.

With the support of the MVDWQL a user can
query either a single DW version or multiple DW
versions, real and alternative ones. A query that
addresses a single DW version will be further
called a singleversion query (SVQ), whereas a
query that addresses multiple DW versions will
be further called a multiversion query (MVQ).
The set of real DW versions addressed in a
MVQ is indicated by a user either by specifying
a time interval, represented by version begin and
version end validity times or by specifying the set
of version identifiers VID 1 , ..., VID n . In the first
case, the SELECT command uses the VERSION
FROM start_date TO end_date clause. In the sec-
ond case, the SELECT command uses VERSION
IN (VID 1 , ..., VID n ) clause. If the version selection
criteria point to one DW version, then a MVQ
becomes a SVQ. If clauses VERSION FROM
and VERSION IN are omitted in a MVQ, then the
query will address all real DW versions.

QUERYING THE CONTENT OF
MULTIVERSION DATA WAREHOUSE

In the MVDW , data of user interest are usually
distributed among several DW versions and a
user may not be aware of the location of the par-
ticular set of data. Moreover, DW versions being
addressed in queries may differ with respect to
their schemas and the structure of dimension in-
stances. Such DW versions will further be called
heterogeneous . Querying heterogeneous DW ver-
sions is challenging and requires intensive usage
of metadata. In our approach, we developed the
MVDWQL query language, capable of:

Processing Multiversion Queries

Parsing and executing a MVQ is performed in the
five following steps (Wrembel & Bębel, 2007).

1. Constructing the set of DW versions: The

set S V ={V 1 , V 2 , ..., V n } of DW versions (either
real or alternative ones) that is addressed in
a MVQ is constructed taking into account
version begin and end validity times or
explicitly provided version identifiers.

• Querying the content of multiple heteroge-

neous DW versions (Morzy & Wrembel,
2004; Wrembel & Bębel, 2007; Wrembel
& Morzy, 2006), such queries will be fur-
ther called content queries ;
Querying metadata describing DW ver-

2. Decomposing MVQ: Next, for every DW

version V i belonging to S V , the system con-
structs an appropriate singleversion query
SVQ i . In this process, the differences in
schema versions are taken into consideration.
If some tables and attributes changed their
names from one version to another, then their
names are found in an appropriate dictionary
table and are used for constructing SVQ i .

•

sions, such queries will be further called
metadata queries .

In this section we focus on content queries
whereas in the next section we focus on metadata
queries.