At 12:54 PM +0200 8/23/00, Quim Sanmarti wrote:
>1. I was particularly interested in how the boolean queries are parsed. The
>first thing I tried was:
Well, first off I can say that boolean queries are still stranger
than even you've seen. Try mixing in a phrase into a boolean query!
Whoops...
>Hmm. Does this imply that the parser generates *binary* children when
>trying boolean? Trying only with 'and' operators, does much the same...
Yes, the boolean parser doesn't try to do more than n=2 right now.
Even that's causing some problems that are hard to test as you've
seen. Of course you can also see that it's doing binary matching in
the code.
>Hmm. No precedence is defined between 'or' and 'and'.
I left precedence for both to be the same, based on left to right parsing.
>It seems to be ignoring the parens.
Yes, right now the HtWordToken function used for splitting queries
tosses out parens. I wrote a new tokenizer which will be needed
anyway to keep field:word together.
>Parsing as a boolean query FAILED
>[silence, infinite loop]
The loop is from AndParseTree parsing. I fixed that particular bug,
but it uncovered others--I need to gobble up whitespace between
tokens sometimes, but the String class doesn't offer a useful method
(yet).
I hope to commit a pile of cleanups tonight.
-Geoff
------------------------------------
To unsubscribe from the htdig3-dev mailing list, send a message to
[EMAIL PROTECTED]
You will receive a message to confirm this.
