> b& why empty-string-and-discard rather than just empty sequence?
[paraphrased]

Because I didnbt think of it! Thanks for the improvement p
________________________________

Are there significant advantages to your suggestions over the following?

(
  ( for $f in $fieldnames return if ( count( $fieldnames[ . eq $f ] ) gt 1 )
then $f else '' )
  => distinct-values()
)[. ne '']

I admit it is not simple, but strikes me as simpler. (Note that instead of the
null string, one could use any string that is guaranteed not to be a value. I
often like bbb (U+241C) or bbb (U+2205) for this kind of thing.
One could also probably use a test on that bifb that only looked at
preceding or following values of $fieldnames, which some folks might find more
readable.)


I would hope that the map:merge is ensured to be more efficient than that
count($fieldnames[. eq $f]) gt 1 but I am not good at expression complexity in
O.. terms. And in the end with advanced XQuery engines it is all thrown at a
query optimizer/rewriter that might optimize the expression in a way not
comparable to the originable one anyway.


I don't get the

  else ''

part to later have to get rid of those with [. ne ''], why can't you simply
use

  else ()

to return the empty sequence?