Received: from redrose.tibcofinance.com (redrose.tibcofinance.com [216.101.119.54]) by hub.org (8.9.3/8.9.3) with ESMTP id IAA03861 for ; Tue, 15 Jun 1999 08:10:36 -0400 (EDT) (envelope-from boekhold@tibco.com) Received: from tekbspa.tibco.com ([216.101.119.76]) by redrose.tibcofinance.com (8.9.3/8.9.3) with SMTP id FAA18109; Tue, 15 Jun 1999 05:05:06 -0700 (PDT) Received: from ukmail.tibco.com (tekuk01) by tekbspa.tibco.com (4.1/SMI-4.1) id AA17055; Tue, 15 Jun 99 05:08:29 PDT Received: from tibco.com by ukmail.tibco.com (SMI-8.6/SMI-4.1) id NAA10327; Tue, 15 Jun 1999 13:09:04 +0100 Message-Id: <37663E80.4EBEC2B1@tibco.com> Date: Tue, 15 Jun 1999 13:52:32 +0200 From: Maarten Boekhold X-Mailer: Mozilla 4.5 [en] (WinNT; I) X-Accept-Language: en Mime-Version: 1.0 To: Alexander Schneider Cc: pgsql-sql@postgreSQL.org Subject: Re: [SQL] Full-Text-Indexing References: <1282947191-50950897@mail.isolution.de> Content-Type: text/plain; charset=us-ascii Content-Transfer-Encoding: 7bit > 1) Why should I use "varchar(n)" instead of "text" though the space for a > varchar-field is larger than text (varchar: 4byte+n, text: 2byte+n)? Is > it more efficient for queries? No reason, I think it the code would even work without changes if you use 'text' fields... > 2) I would like a table where I can insert/update/delete the stop-words > instead of an array inside the code. This would also help in lower the > words to search for in the query. Are there any grave performance > penaltys for this solution? It is much more flexible and I would trade a > small performance loss for this flexibility. [I cannot program in C so > maybe there is a volunteer who could write this little snippet for me? The problem here is that without resorting to some advanced stuff using shared memory or whatever that table would have to be read by every backend starting up, which might be some significant overhead. I'm still hoping somebody can rewrite this stuff to use some specific indexing method so we don't have to go the way of introducing this intermediate table (i.e. store the substring/oid pairs directly in a btree item and use some new index access method to take something like field ~= 'rolling NOT stone' to search this index). Maarten -- Maarten Boekhold, boekhold@tibco.com TIBCO Finance Technology Inc. The Atrium Strawinskylaan 3051 1077 ZX Amsterdam, The Netherlands tel: +31 20 3012158, fax: +31 20 3012358 http://www.tibco.com