Marshall Schor created UIMA-4433:
------------------------------------
Summary: UIMA Sets have contradictory aspect - document this
Key: UIMA-4433
URL: https://issues.apache.org/jira/browse/UIMA-4433
Project: UIMA
Issue Type: Bug
Components: Core Java Framework
Reporter: Marshall Schor
Priority: Minor
Fix For: 2.7.1SDK
While examining the CasObj implementation, I was surprised to notice that the
comparator for Set indexes *always* included a "types-are-the-same". I looked
at the test case for this, and saw some contradictory things.
1) You can ask an index to compare (using its defined comparator) two feature
structures.
1a) If you define an index to compare feature A, and you make two instances of
a Feature Structure with the same A value, but one is a subtype of the other,
the myIndex.compare(fs1, fs2) will return 0 (indicating they are "equal" for
the purposes of the comparator.
1b) If you define the comparator to compare feature A and use type priorities,
then the compare will not be 0 (because the types are different).
2) back to the first comparator - just on feature A. If you add-to-indexes the
two FSs, the 2nd one, even though it "compares" equal to the first using the
index's comparator, will still be added to the indexes; the size() of the index
will be 2, not 1.
This is long-standing behavior, so I'm reluctant to change it. It is caused by
a detail of the underlying implementation, where the index is actually kept as
"n" separate indexes, by type, so the Type and the SubType are in two separate
indexes.
My feeling at this point is to update the documentation to correspond to this
detail of the implementation. Other thoughts?
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
