I've changed DocumentImpl.java in three locations:
buildIndex() traverseParents() updatePathIndex()
This patch modifies traverseParents(). It changes pathsFoundForIndexing from a vector to a HashMap, and while it processes leaf node xpaths, relative xpaths, and absolute xpaths, it populates this HashMap with PathIndexEntry objects that will be inserted into the xml_path_index table. This method was incorrectly indexing nodes of type ATTRIBUTE, because attributes stored their node data in the same xml_nodes record, whereas nodes of type ELEMENT stored their data in a subsequent child TEXT record. This was causing offsets to occur in the indexing tables. I've added to variables in this method now, leafData and leafDataNumerical to fix this offset. See the patch on buildIndex().
As part of a patch fix for:
I've changed DocumentImpl.java in three locations:
buildIndex()
traverseParents()
updatePathIndex()
This patch modifies traverseParents(). It changes pathsFoundForIndexing from
a vector to a HashMap, and while it processes leaf node xpaths, relative xpaths,
and absolute xpaths, it populates this HashMap with PathIndexEntry objects
that will be inserted into the xml_path_index table. This method was incorrectly
indexing nodes of type ATTRIBUTE, because attributes stored their node data
in the same xml_nodes record, whereas nodes of type ELEMENT stored their data in
a subsequent child TEXT record. This was causing offsets to occur in the indexing
tables. I've added to variables in this method now, leafData and leafDataNumerical
to fix this offset. See the patch on buildIndex().