Date:         Thu, 18 Dec 2003 14:15:55 +1100
Reply-To:     "Z39.50 Next-Generation Initiative" <[log in to unmask]>
Sender:       "Z39.50 Next-Generation Initiative" <[log in to unmask]>
From:         Alan Kent <[log in to unmask]>
Subject:      Re: Attribute lists for queries - what are people doing?
Comments: To: "Z39.50 Next-Generation Initiative" <[log in to unmask]>
In-Reply-To:  <[log in to unmask]>
Content-Type: text/plain; charset=us-ascii

On Tue, Dec 09, 2003 at 01:27:40PM +0000, Robert Sanderson wrote:
> > Finally, exact, any, and all are interesting cases. There is no 'any'
> > in Bib-1 (unless I have an old printout). 'all' I guess is 'word list'.
>
> For any, I turn the clause into a tree of OR clauses in bib1, AND clauses
> for all. Yes, this isn't quite correct, as the client will likely have
> different 'word' extraction to the server.  Exact I use 2=3,5=100,6=3.

While debatable, there are advantages. If dealing with an existing Bib-1
system, it won't support the new pattern match form at present.  It also
wont support 'any' and 'all'. So it sounds like the best practical solution
available at present is:

For Bib-1:

    If a word based term (using 'all' and 'any' not '=' or 'exact')
        Split terms based on whitespace
        Add first-in-field if '^' at start
        Add left/right/both truncation if '*' at start/end/both
        Error on anything more complex (could try a regexp format)
    else
        Leave term as single string and do the above '^' etc rules
    end-if

For AA (if trying not to do it better than Bib-1 above):

    Wait for any/all to be fixed in the standard
    Wait for new CQL regexp to be added (if not added already)
    Add CQL regexp attribute if query contains any of '^', '*', '?', and '\'

For Bib-1, there are a couple of regexp formats for Bib-1 that could be
tried, but cross translation of patterns from one format to another can
be tricky to get right and not necessarily portable.


I almost wonder if CQL should say for 'all' and 'any' that terms are
white space separated. If you say 'book-case' and the system indexes it
as two words, then treat as a phase query (adjacent words). This seems
to be what all the existing implementations do, so I wonder if the spec
should be changed to make it compatible.

Alan