[ https://issues.apache.org/jira/browse/CASSANDRA-5437?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13625699#comment-13625699 ]
Brandon Williams commented on CASSANDRA-5437: --------------------------------------------- There is a third option, and it's the best: perform gets and parallelize them yourself. Here's why: if you perform a multiget on [x,y,z] keys and x and y are fine, but z times out, all you get is an exception. If you had done parallel gets yourself, you'd have gotten x and y, then retried z, but since you used multiget now you have to retry the entire query again. That said, I don't think we have any plans to change or improve the thrift interface at this point. > multiget specific key filtering > ------------------------------- > > Key: CASSANDRA-5437 > URL: https://issues.apache.org/jira/browse/CASSANDRA-5437 > Project: Cassandra > Issue Type: Improvement > Reporter: Brooke Bryan > Priority: Minor > > Currently, you are able to specify a list of columns, or a slice when pulling > back keys from a multiget, however, we find ourselves pulling back a lot more > data than we actually need when columns overlap. We currently have 2 options > (as far as im aware). > 1. perform multiple get calls, and get the required data back, but connect to > thrift more > 2. perform a multiget, and get more data due to column crossovers, but in a > single request. > Similar to a batch mutation, a batch read would solve this problem perfectly. > e.g. > read = [ > 'columnFamily' => [ > 'key' => ColumnPath, > 'key2' => SlicePredicate, > 'key3' => ColumnPath > ] > ]; > result = batch_read(read); > //result > 'columnFamily' => [ > 'key' => [1 => 'a',2 => 'b',3 => 'c'], > 'key2' => [4 => 'd',5 => 'e'], > 'key3' => [1 => 'a',3 => 'c'] > ] -- This message is automatically generated by JIRA. If you think it was sent incorrectly, please contact your JIRA administrators For more information on JIRA, see: http://www.atlassian.com/software/jira