- From: Michael Kay <mhk@mhk.me.uk>
- Date: Fri, 13 May 2005 09:32:56 +0100
- To: "'Daniel Engovatov'" <dengovatov@bea.com>, <public-qt-comments@w3.org>, <w3c-xml-query-wg@w3.org>
- Message-ID: <E1DWVbX-00047W-Ty@lisa.w3.org>
It would be better to raise comments on Bugzilla! I have some sympathy with this requirement but I think this kind of thing is best handled using pragmas (or option declarations). For example, in my implementation I'm not so much concerned about whether the results are deterministic as about whether calling the function has side-effects. I think we'll find it very hard to define a set of properties with semantics that we can specify rigorously and agree on, so it's best to leave it to vendors. Michael Kay _____ From: w3c-xml-query-wg-request@w3.org [mailto:w3c-xml-query-wg-request@w3.org] On Behalf Of Daniel Engovatov Sent: 13 May 2005 03:14 To: public-qt-comments@w3.org; w3c-xml-query-wg@w3.org Subject: [Norton AntiSpam] "deterministic" indicator for external functions. We propose to add an optional clause to the external function declaration to indicate whether the external function always returns the same results for given argument values ("deterministic"). An example use case would be to allow users to indicate that two external functions embedded into xquery: f:random() and f:verySlowDeterministicExternalWebService() should be treated differently by the query processor. That can be implemented by changing FunctionDecl to [26] FunctionDecl ::= "declare" "function" QName "(" ParamList? (")" | (")" "as" SequenceType)) (EnclosedExpr | ("external" (<"not" "deterministic"> | "deterministic")?)) Daniel.
Received on Friday, 13 May 2005 08:33:11 UTC