Re: draft XML query results format spec

On Tue, 28 Sep 2004 18:04:46 +0100, "Seaborne, Andy" <andy.seaborne@hp.com> wrote:

> Dave,
> 
> I don't mind which style of design - I would go for whatever the XML
> community see as most appropriate for processing with XSLT and XQuery.
> 
> Two points from the telecon:
> 
> 1/ Handling of undefined variables in individual <result>:
>    Either an explicit way to say "is undefined" or
>    omission of the var for that entry would work for me.

I've gone with omission in the revised version.

> [Note the spec does not work in terms of "nulls" but in terms of "not
> defined" and leaving it to the local API to decide what to do, whether
> to make it appear as whatever the programming language convention is]
> 
> 2/ Do XML literals go in as XML subtrees?
>    I assume so - so using the same tags as the wrapper may arise and a
>    poorly written XPath may (mis)match.  Not sure we can remove all
>    situations of this.  Sometimes may actually want it.

Can't do much about bad queries.  There are XPath forms for
selecting the immediate child (and all subchildren) but we should
encourage use of rooted XPaths into the XML so this doesn't happen.
/results/result[0]/name rather than */name.

>    Worse case: a query result as XML literal within an XML result set.
>    Example: querying a detailed server log file of queries and their
> results.
> 
> 
> FYI I checked and variable names are currently "NCNameStartChar
> (NCNameChar)*" so there are no restrictions there and if we have
> variables marked with ? or $, this can easily change to match the
> XMLresult format.  Currently ':' is not legal in variable names to help
> simple lexing but that is not fundamental.

NCName seems fine as a variable name restriction.  Without checking
I'd assume XQuery restricts similarly.

> Grammar-wise "?1" could be made legal but XML format 1 is a good enough
> reason not to have it as such.

Thinks like ?, $ seems a bit odd in the xml.

I never saw ? as part of the name (or $) in rdql, brql but syntax to indicate a
variable name was coming up.

Dave

Received on Wednesday, 29 September 2004 13:23:39 UTC