RE: [search-ws] CQL discussion on SRU list

["Ray Denenberg, Library of Congress" <rden@loc.gov>]

Some of this is coming back to me, tell me if this rings a bell.   

Bear with me.

If we allow:  title=(cat or dog)

then we are introducing the concept of a structured term, that is, where a
search clause is 'index relation term', normally 'term' is, as they say,
"terminal". But now, it can be structured.

But fundamental to CQL is the ability to default the index and relation, so
that the query

                                                  cat

is a valid cql query, where it is expanded to 

                      <default index> <default relation>  cat


But with the structured term, this means that 


                                       cat or dog

which we previously would have understood to be term boolean term

might be a  term that needs to be expanded to 

                      <default index> <default relation>  cat or dog

i.e.  TOTAL AMBIGUITY!




-----Original Message-----
From: Matthew Dovey [mailto:m.dovey@jisc.ac.uk] 
Sent: Thursday, June 03, 2010 5:16 PM
To: Denenberg, Ray; 'LeVan,Ralph'; 'OASIS SWS TC'
Subject: RE: [search-ws] CQL discussion on SRU list

> 2.  title = (cat or dog)  is completely equivalent to title = cat or 
> title = dog, which is a sane and easy way to express it, so why do we 
> need these ridiculous histrionics.

This was the compelling reason (Occams Razor, KISS etc.). We were trying to
limit unnecessary variants and only adding to the BNF structures which were
essential to express new queries rather than adding new ways of expressing
existing queries.

CQL is not really human readable

title = smith or jones

Although this has a single meaning according to the BNF, it is already
ambiguous to the casual reader who might have meant it to be

title = "smith or jones"

or 

(title = smith)  or (any jones)

Add in Mike's variant and it could also have been intended to be

title = smith or title = jones

Moreover if we were to introduce short cuts of this nature, where would we
stop. Do we also allow

author or title = smith

In Mathematical terms, Mike seems to be proposing to make the comparison
operators distributive over the logical operators, I know of no formal logic
which allows that (I suspect for very good reasons).

Matthew