Databases Reference
In-Depth Information
<Source>
<Company>
<Name>IBM</Title>
<Location>NY</Address>
</Company>
<Company>
<Name>MS</Title>
<Location>WA</Address>
</Company>
</Source>
Source
Company
Target
Organization
Name
Location
Title
Address
Fig. 9.4
A simple mapping scenario and the source schema instance
<Target>
<Organization>
<Title>IBM</Title>
<Address>NY</Address>
</Organization>
</Target>
<Target>
<Organization>
<Title>IBM</Title>
<Address>NY</Address>
</Organization>
<Organization>
<Title>IBM</Title>
<Address>NY</Address>
</Organization>
</Target>
<Target>
<Organization>
<Title>IBM</Title>
<Title>MS</Title>
<Address>NY</Address>
<Address>WA</Address>
</Organization>
</Target>
(a)
(b)
(c)
Fig. 9.5
Three different target instances generated by different tools
matches, there is a large variety of specification options as a recent classification of
mapping tools illustrates [
Legler and Naumann 2007
].
The input problem goes even further. Some mapping tools allow the designer to
edit the generated mappings or transformation scripts to correct or enhance them.
In that way, the generated output is restricted only by the expressive power of the
mapping language or of the transformation script. Under such circumstances, a sce-
nario should be extended to include, apart from the two schemas and the intended
mapping specification, the modifications/corrections that the designer does on the
generated output. However, allowing the designer to edit the output makes unfair
any comparison to mapping tools that operate under the principle that the designer
can only use the high-level graphical input language [
Altova 2008
].
Another issue of inconsistency across different matching and mapping tools is
the lack of a standardized output. Some matching tools generate only 1-1 identity
function matches, i.e., simple interschema correspondences, while others generate
more complex relationships. Furthermore, some mapping tools generate mappings
as interschema dependencies only, while others produce also the transformation
scripts. The problem is becoming more crucial due to the fact that there is no
unique way of generating a target instance. Two different mapping tools may pro-
duce completely different transformation scripts and yet generate the same target
instance.
Deciding the metrics with which success is measured is another challenging task.
Since a general goal of a mapping tool is to reduce the required programming effort,
