[FO] fn:intersperse

[qt4cg]

[FO] fn:intersperse

[benibela]

fn:intersperse($div/span, ',')

Is that really better than tail($div/span!(',', .)) ?

[ChristianGruen]

In the past, our users were asking for a function that works similar to fn:string-join, but can be applied to sequences. Similar to other sequence functions (including fn:head, fn:tail, fn:subsequence, fn:remove, and others that can be realized with filter expressions), the new function could make common tasks more convenient and accessible by a single function (i.e., a function in which you have to supply the input sequence only once).

An equivalent expression for the proposed function would be:

head($sequence), tail($sequence) ! ($separator, .)

[michaelhkay]

The separator attribute on xsl:value-of has certainly been a great invention, and I've proposed adding a separator attribute to xsl:for-each, so the use case certainly exists.

The typical user today writes something like $sequence ! (., (if (position() != last()) then $separator else ())) which is very ugly, and very expensive unless optimized, and makes streamed execution difficult.

So I agree this function would be useful.

[ChristianGruen]

Alternative names: fn:item-join, fn:sequence-join.
What are your favorites?

[michaelhkay]

I like "intersperse". "Join" is wrong, nothing is being joined. "Inject" would also work.

[joewiz]

+1 to "intersperse" - as it suggests that something is being inserted between the items. Possibly, "interpose" or "delimit"?

[ChristianGruen]

In the meeting, we discussed an inverse function for fn:intersperse. Maybe https://qt4cg.org/specifications/xpath-functions-40/Overview-diff.html#func-array-partition already fills the gap?

It works slightly different at the moment, as it includes the separating item in the result, but that behavior could possibly be controlled with an additional option.

[michaelhkay]

I found the discussion rather confusing.

If the separators are single items, the problem is trivial, just remove all the items at even-numbered positions.

Otherwise, the problem is only soluble if the separator sequence can be distinguished from the other items; you're then looking to remove all subsequences that "match" some supplied sequence. Which I guess gets us back to functions for comparing sequences (for which one approach might be a generalised deep-equal).

But could someone tell me the use case?

[michaelhkay]

The change has been accepted and implemented. Therefore closing the issue.