5596 2008-04-01 13:26:10 +0000 [FS] Missing rule in 7.2.5 The fn:collection? 2008-04-30 06:31:37 +0000 1 1 1 Unclassified XPath / XQuery / XSLT Formal Semantics 1.0 Recommendation PC Windows NT CLOSED WONTFIX consider for 1.1 P2 enhancement --- 1 tim jmdyck public-qt-comments oldest_to_newest 19676 0 tim 2008-04-01 13:26:10 +0000 In F&O 15.5.6 (fn:collection) says "If $arg is the empty sequence, the function behaves as if it had been called without an argument" The static typing rules in FS 7.2.5 should reflect this. I suspect it should be something along the lines of: statEnv |- QName of func expands to (FN-URI,"collection") statEnv |- Implementation-defined default sequence has type Type statEnv |- Expr : empty-sequence() ------------------------------------------------------------------ statEnv |- QName(Expr) : Type 19949 1 jmdyck 2008-04-26 01:35:24 +0000 As I understand it, the rules of FS 7.2.5 are complete, and consistent with F+O 15.5.6, they just aren't as precise as they could be in the case where the argument's static type is empty. So unless that lack of precision causes serious problems, I don't think I can process this as an erratum. (Note that an implementation is free to infer the more precise type under FS 6.1.1 Static Typing Extensions.) 19952 2 tim 2008-04-28 07:21:28 +0000 That's a shame. We do of course implement this rule. Might it be considered as a feature request for XQuery 1.1 FS? It seems unfortunate that fn:collection() and fn:collection( () ) should be typed differently when the behaviour is meant to be the same. 19953 3 jmdyck 2008-04-28 07:44:59 +0000 (In reply to comment #2) > Might it be considered as a feature request for XQuery 1.1 FS? Yes, I've marked it as "consider for 1.1" in the Status Whiteboard. > It seems unfortunate that fn:collection() and fn:collection( () ) should be > typed differently when the behaviour is meant to be the same. I agree, but I don't think "unfortunate" warrants an erratum. 19954 4 tim 2008-04-28 07:57:14 +0000 Thanks. The only argument in favour of adding it as an erratum would be: "If $arg is the empty sequence, the function behaves as if it had been called without an argument." but I don't think I'd even try to argue that the static type is a part of the behaviour. 19993 5 jmdyck 2008-04-29 22:24:31 +0000 At their meeting today, the WGs approved the response in comment #1, so I am changing this issue's Severity to "enhancement" and marking it as resolved-WONTFIX (and keeping "consider for 1.1" in the Status Whiteboard). If you accept this resolution, please mark the issue CLOSED. 19995 6 tim 2008-04-30 06:31:37 +0000 Thanks. I'd like to request that the static typing rules of fn:trace also be considered for 1.1 (Bug 3784).