Chopping and off the beginning would mean you couldn't search for any phrase or collection of words that starts with and. Editing the ccl.properties file and changing buildquery is a much more robust solution and doesn't actually result in reduced functionality. I would accept a patch for that, if anyone wants to send one, otherwise ill try to work on one next week. I don't think one that drops words out that the user is trying to search is one that would be accepted, at least not into 3.4. I can't speak for chris nighswonger the 3.2 maintainer, only myself as 3.4. If you want to work through making a patch you can find me, or others who can help you on irc (irc.katipo.co.nz #koha) Chris RM for 3.4 and totally over xmas shopping Bob Ewart <bob-ewart@bobsown.com> wrote:
Ona simple query, where you don't have ccl=, you could just chop 'and ' off the beginning and ' and' off the end of the query.
I thought I could do it after line 1032 in Search.pm, but that didn't seem to work.
Where does the debug output go? I'd like to see what Search.pm is doing.
On 12/17/2010 06:24 PM, Chris Cormack wrote:
You are absolutely right removing it would cause problems, but I don't suggest doing that, if you look at bug 5515 you will see I suggest we use something other than and for the directive perhaps +++ and then change build query to use that.
People could still use ccl from the browser if they type ccl= but all other searches would be constructed by buildquery and would look like ti=the hobbit +++ au=tolkien. Which would mean ti=and simple +++ au=bob would work too. We might want to do the same with or and not.
Hope this makes sense, typed on my phone while taking a break from christmas shopping. Merry christmas to your family Henri Damien, and to everyone else in the Koha family :)
Chris
On 18 Dec 2010 07:08, "LAURENT Henri-Damien" <henridamien.laurent@biblibre.com <mailto:henridamien.laurent@biblibre.com>> wrote:
Le 17/12/2010 18:17, Chris Cormack a écrit :
On 17 December 2010 23:46, LAURENT Henri-Damien <henridamien.laurent@biblibre.com <mailto:henridamien.laurent@biblibre.com>> wrote:
Le ...
mmm... removing and from ccl properties would break all the crossed searches... say I want Title=Stories and author=Poe and pb=Penguin Removing and from ccl.properties... would make it fail. Hope that helps. -- Henri-Damien LAURENT
-- Bob
_______________________________________________ Koha mailing list http://koha-community.org Koha@lists.katipo.co.nz http://lists.katipo.co.nz/mailman/listinfo/koha
It would mean that you could get more hits than with the and in place, but that's better than no hits. An alternate idea would be to put quotes, either kind, around the search phrase, if there is not ccl= or and = for that matter. That does work. I'm not disagreeing with your solution, just trying to come up with alternatives. On 12/17/2010 11:13 PM, Chris Cormack wrote:
Chopping and off the beginning would mean you couldn't search for any phrase or collection of words that starts with and. Editing the ccl.properties file and changing buildquery is a much more robust solution and doesn't actually result in reduced functionality.
I would accept a patch for that, if anyone wants to send one, otherwise ill try to work on one next week. I don't think one that drops words out that the user is trying to search is one that would be accepted, at least not into 3.4. I can't speak for chris nighswonger the 3.2 maintainer, only myself as 3.4.
If you want to work through making a patch you can find me, or others who can help you on irc (irc.katipo.co.nz #koha)
Chris RM for 3.4 and totally over xmas shopping
Bob Ewart <bob-ewart@bobsown.com> wrote:
Ona simple query, where you don't have ccl=, you could just chop 'and ' off the beginning and ' and' off the end of the query.
I thought I could do it after line 1032 in Search.pm, but that didn't seem to work.
Where does the debug output go? I'd like to see what Search.pm is doing.
On 12/17/2010 06:24 PM, Chris Cormack wrote:
You are absolutely right removing it would cause problems, but I don't suggest doing that, if you look at bug 5515 you will see I suggest we use something other than and for the directive perhaps +++ and then change build query to use that.
People could still use ccl from the browser if they type ccl= but all other searches would be constructed by buildquery and would look like ti=the hobbit +++ au=tolkien. Which would mean ti=and simple +++ au=bob would work too. We might want to do the same with or and not.
Hope this makes sense, typed on my phone while taking a break from christmas shopping. Merry christmas to your family Henri Damien, and to everyone else in the Koha family :)
Chris
On 18 Dec 2010 07:08, "LAURENT Henri-Damien" <henridamien.laurent@biblibre.com <mailto:henridamien.laurent@biblibre.com>> wrote:
Le 17/12/2010 18:17, Chris Cormack a écrit :
On 17 December 2010 23:46, LAURENT Henri-Damien <henridamien.laurent@biblibre.com <mailto:henridamien.laurent@biblibre.com>> wrote:
Le ...
mmm... removing and from ccl properties would break all the crossed searches... say I want Title=Stories and author=Poe and pb=Penguin Removing and from ccl.properties... would make it fail. Hope that helps. -- Henri-Damien LAURENT
-- Bob
-- Bob
On 19 December 2010 02:21, Bob Ewart <bob-ewart@bobsown.com> wrote:
It would mean that you could get more hits than with the and in place, but that's better than no hits.
Yep, but surely it's even better to get hits for what you are actually searching for :)
An alternate idea would be to put quotes, either kind, around the search phrase, if there is not ccl= or and = for that matter. That does work.
Yep, Henri-Damien suggested that, it becomes a phrase search, not a wordlist search at that point. Which means that the words have to be in that order. So it's now a different search.
I'm not disagreeing with your solution, just trying to come up with alternatives.
Sure, I'm just unsure why :) The problem is we are treating the word and as an operator, if we don't do that, the problem goes away. Chris
participants (3)
-
Bob Ewart -
Chris Cormack -
Chris Cormack