On 2/26/07, Chris Hostetter <[EMAIL PROTECTED]> wrote:
: It probably would simplify some logic to assume unique keys are : required, but we should think a little on it first. Would it make it : harder to model existing random lucene indicies? yeah ... because then people would be force client side to generate an arbitrary unique key for each doc ... if we really wnat to go this route 9and i don't see any reason why we should) we should make sure we support an autoincrment counter as a default field value so people who don't care about document uniqueness don't have to do anything new to keep using Solr (or to migrate from Lucene to Solr)
The problem with SOLR-172 is when you send an add command with "overwrite*". If this happens, the update handler could: 1. quietly use allowDups() instead 2. send an error saying "missing id" In the applications i'm working on, 2 is the better option - that may not always be the case. I hate to suggest schema bloat, but maybe we should have: <uniqueKey required="true"> Also, if no <uniqueKey> is defined in schema.xml, the add command should always use allowDups() even if the command said overwrite. ryan
