A specialized implementation of the
NodeCombiner interface
that constructs a union from two passed in node hierarchies.
The given source hierarchies are traversed, and their nodes are added to the
resulting structure. Under some circumstances two nodes can be combined
rather than adding both. This is the case if both nodes are single children
(no lists) of their parents and do not have values. The corresponding check
is implemented in the
findCombineNode() method.
Sometimes it is not possible for this combiner to detect whether two nodes
can be combined or not. Consider the following two node hierarchies:
Hierarchy 1:
Database
+--Tables
+--Table
+--name [users]
+--fields
+--field
| +--name [uid]
+--field
| +--name [usrname]
...
Hierarchy 2:
Database
+--Tables
+--Table
+--name [documents]
+--fields
+--field
| +--name [docid]
+--field
| +--name [docname]
...
Both hierarchies contain data about database tables. Each describes a single
table. If these hierarchies are to be combined, the result should probably
look like the following:
Database
+--Tables
+--Table
| +--name [users]
| +--fields
| +--field
| | +--name [uid]
| ...
+--Table
+--name [documents]
+--fields
+--field
| +--name [docid]
...
i.e. the
Tables nodes should be combined, while the
Table nodes should both be added to the resulting tree. From
the combiner's point of view there is no difference between the
Tables and the
Table nodes in the source trees,
so the developer has to help out and give a hint that the
Tablenodes belong to a list structure. This can be done using the
addListNode() method; this method expects the name of a node,
which should be treated as a list node. So if
addListNode("Table"); was called, the combiner knows that it
must not combine the
Table nodes, but add it both to the
resulting tree.
Another limitation is the handling of attributes: Attributes can only
have a single value. So if two nodes are to be combined which both have
an attribute with the same name, it is not possible to construct a
proper union attribute. In this case, the attribute value from the
first node is used.