The data(.) function is required by the XQuery spec for atomization.
Look at: http://www.w3.org/TR/xquery/#id-atomization
When a node is passed to anything that expects an atomic type, the typed
value is extracted. Since at the time of translation we have no idea
what the input is, we always throw in the data(.) call. The optimizer
will eliminate it later if it was not required (like in these cases).
Vinayak
On 6/6/12 10:46 PM, Michael Carey wrote:
And the data(.) function is needed for these expressions because...? :-)
(I'm enjoying following his explorations - great way to learn!)
On 6/6/12 7:54 PM, Vinayak Borkar wrote:
My questions:
- in 1. The plan use the function promote. Is this to ensure the type
is correct for the addition function?
- in 1b. Why does the type here show anyAtomicType instead of integer?
The add function you use in the query is defined as an operator in the
builtin-operators.xml file. Remember that this operator is
polymorphic. It can add integers, doubles, duration to dates etc. So
the static type of the operator's argument specify that they should be
the supertype of all acceptable argument types and that is
anyAtomicType in this case.
You are correct in that the promote will ensure that the input data is
compliant with the expected argument type of the addition function.
At runtime, the function needs to make sure that the inputs are
correct and throw a dynamic error otherwise.
Vinayak
