2543 2005-11-21 14:54:29 +0000 Unique constraint misinterpretation. 2007-01-12 16:02:54 +0000 1 1 1 Unclassified XML Schema Structures: XSD Part 1 1.1 only All All RESOLVED WONTFIX important, work, idc cluster editorial P4 normal --- 4201 4202 1 etishkin cmsmcq sandygao www-xml-schema-comments oldest_to_newest 7242 0 etishkin 2005-11-21 14:54:29 +0000 Hi, Proposed change for the unique constraint definition is: Change paragraph 3.11.4 by adding a new clause 4.1.1 to make it clear that not all fields must be present. 3.11.4 Identity-constraint Definition Validation Rules ... 4.1.1 The ·target node set· is a subset of the ·qualified node set·, that is, every member of the ·target node set· is also a member of the ·qualified node set· however not every member of ·taget node set· is a member of ·qualified node set·. ... Why this is important? Xerces parser has a bug 18405, which has not been resolved since 2003. The reason for not fixing is : " ... I (and some others I've asked) think that the spec seems open to interpretation here; so I'm marking this bug invalid for now. I'd encourage you to pose a question to the Schema working group via their comments list; if you get a response that demonstrates that your reading is correct, please reopen this bug with the appropriate text and I'll try and fix it when I get a chance." The below is the comment from Xerces parser BUG 18405 report: ================================================ "3 For each node in the ·target node set· all of the {fields}, with that node as the context node, evaluate to either an empty node-set or a node-set with exactly one member, which must have a simple type." The target node set is the set of nodes on which the selector is matched. Note that this condition must hold for *any* identity constraint; only in bullet 4 of the tableau are the differences between key and unique described. Therefore, it seems clear that, if a selector matches, then either all of the fields must match or none of them must match; even for xsd:unique, you can't have some fields matching. ========================================================= Thanks, Eugene 13480 1 sandygao 2007-01-12 16:02:54 +0000 The Schema WG discussed this issue at its 2007-01-05 telecon, and decided to close this bug without adopting the suggested changes. The 1.0 specification may be unclear with respect to whether "unique" values are allowed to be "partial". The updated rules in 1.1 is believed to have clarified this. It may still be helpful to have a note to make this explicit to avoid further mis-interpretation. Bug 4202 is opened to track this, which is purely editorial.