Jeni's Musings

Odd how appealing expressions like

for $p in (0 to string-length($string1)) return $p

are, even to experts! Fortunately, the optimizer quickly turns them into

(0 to string-length($string1))

It’s one of those things like writing (if ($x=true())…) - somehow we do it even though we know it’s grossly redundant.

I wonder if it’s worth inferring a type of “xs:integer+” for (0 to string-length($string1)) rather than the current “xs:integer*”? In this example we’re doing an unnecessary dynamic type check. No, probably not, special-casing based on single examples has to stop somewhere! Unless the example is in an important benchmark, of course.