1688 2005-07-17 21:40:53 +0000 simplifying the definition of AtomizeAtomic 2005-09-06 13:05:33 +0000 1 1 1 Unclassified XPath / XQuery / XSLT Formal Semantics 1.0 Last Call drafts PC Windows 2000 CLOSED DUPLICATE 1576 P2 normal --- 1 fred.zemke simeon public-qt-comments oldest_to_newest 4774 0 fred.zemke 2005-07-17 21:40:53 +0000 4.1.5 Function calls Comments on the definition of [Expr]AtomizeAtomic(SequenceType): a) In the first case, where the result if fn:data(Expr), you have the judgment Expr:Type. I think you mean statEnv |- Expr:Type, since type inferencing always occurs in the context of the static environment group. b) The exclusion "and not (Type=empty)" is unnecessary since there is no harm in invoking fn:data(()). The parameter type of fn:data() is item()* so it is already defined on the empty sequence. c) The variable Type is a free variable with no quantification. You cannot mean universal quantification, since (for all Type)(statEnv |- Expr:Type) can never be true. You must mean (there exists Type)[Expr:Type and not (Type=empty)]. d) Instead of "Expr : Type and not (Type = empty)", you could just write "not (Expr : empty)", which has the advantage of eliminating the variable Type. e) The judgment SequenceType <: xdt:anyAtomicType* is not defined because the <: relationship holds between two Types. You need to first transform the SequenceType into its equivalent Type and then you can test this relationship. Thus the condition is [SequenceType]sequencetype <: xdt:anyAtomicType* f) The definition will read better if you place the "if" clause around the rule rather than nested within it. The summary of this analysis is that the rule is better expressed: If [SequenceType]sequencetype <: xdt:anyAtomicType* then fn:data(Expr) else Expr 5082 1 simeon 2005-07-21 18:47:11 +0000 (a), (b), (c), and (d) will be better handled, by removing "and Expr : Type and not(Type=empty)" in the first case of the definition of [Expr]AtomizeAtomic(SequenceType). See also bug 1576. (e) is correct and should be fixed as suggested. (f) is editorial. I think we should avoid using: If [SequenceType]sequencetype <: xdt:anyAtomicType* then fn:data(Expr) else Expr which looks like XQuery, but isn't XQuery. We cannot actually write this as a proper normalization rule since [SequenceType]sequencetype <: xdt:anyAtomicType* is not an acceptable expression. - Jerome 5141 2 simeon 2005-07-22 16:12:20 +0000 *** This bug has been marked as a duplicate of 1576 ***