On Sat, Jun 22, 2013, at 03:40 AM, Chris Hostetter wrote:
>
> : This could be a very useful feature. To do it properly, you'd want some
> : new update syntax, extending that of the atomic updates. That is, a new
> : custom request handler could do it, but might now be the best way.
>
> the biggest complexity to implementing this in a general way would be
> dealing with multi-node indexes in SolrCloud
>
> for single node indexes, the basic gist of what you'd want to do in a
> custom handler would be something like...
>
> // you don't care about the DocList
> DocSet docs = SolrIndexSearcher.execute(your query)
> foreach docId in docs {
> String docKey = fieldCache.lookup(docId)
> SolrInputDocument newDoc;
> newDoc.addField(uniqueKeyField, docKey)
> newDoc.addField( ... your atomic update operation ...)
> SolrCore.getUpdateProcessor().process(new AddDocCOmmand(newDoc));
> }
And what makes it complex across a multinode cluster? You'd need a
method to push that query out to all shards, and then each shard would
do the above. You'd need a rule that the unique key cannot be changed,
otherwise the doc might need to switch shards, but that's probably the
case already. Is there more to it than that?
Upayavira
