Re: [xsl] index-of and empty stuff

[Michael Kay <mike@xxxxxxxxxxxx>]

On 21 Dec 2013, at 20:42, Ihe Onwuka <ihe.onwuka@xxxxxxxxx> wrote:

> Of course I was not in the room . There may be a very good reason - it
> is not at all apparent. If there is a good reason it's worth knowing
> (in fact it's also worth knowing if there isn't a good reason) so
> please bear with my question(s).
>

If you want a historical evidence-based answer to the question, why was it
designed like this, you won't get one. It's very rarely possible to find a
specific decision minuted in enough detail to reconstruct the decision
process.

However, the general rationale for XPath 2.0 functions was:

* arguments are classified into "input" and "control" arguments. Input data is
data likely to come from the source document, while control arguments define
how the input data is to be processed.

* input data arguments generally accept an empty sequence, and the function
generally either (a) treats an empty sequence as a zero-length string, or (b)
returns an empty sequence if an empty sequence is supplied

* control arguments generally do not accept an empty sequence.

* in the case of index-of(), the first argument is considered a data argument,
and the second a control argument.

But of course there are inconsistencies, if only because this rationale was
not retrospectively applied to 1.0 functions for compatibility reasons.

Michael Kay
Saxonica