> Date: Wed, 16 Oct 2002 14:53:09 +0100
> From: Robert Sanderson <[log in to unmask]>
>
> Query:
> exact=or and dc.title any "unix windows"
>
> Last time I brought this up, you wanted this to be:
> (default.exact = "or" and dc.title any "unix windows")
>
> Query:
> dc.title any "unix windows" and exact=or
>
> In my monster query, you agreed that exact=or at the end of the query was
> unambiguously a single term:
> (dc.title any "unix windows" and srw.serverChoice any "exact=or")
>
> You can't have it both ways.
I see. Thanks for that much clearer explanation.
Yes, you're right -- you've caught me with my metaphorical trousers
down. I've been inconsistent here, and we need to just choose one of
these interpretations and stick with it, in that if-there-are-
ambiguities-we-just-resolve-them-with-rules way.
I think the lexing rules are nicer if we make "=" _always_ a special
character that breaks words, so that whatever the context (whether an
index name or search term is expected), the input "xyz=abc" is always
lexed as three separate terms: "xyz", "=" and "abc".
That means that of the two interpretations you outlined above,
"exact=or" would be a search for the word "or" in the index called
"exact", rather than a search for the word "exact=or" in the index of
the server's choice.
So that's my suggestion.
_/|_ _______________________________________________________________
/o ) \/ Mike Taylor <[log in to unmask]> www.miketaylor.org.uk
)_v__/\ "The supplied fortune databases have been [hacked] to reduce
redundancy and repetition and redundancy" -- Linux manual
for /usr/games/fortune
|