[
https://issues.apache.org/jira/browse/OMID-78?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16268397#comment-16268397
]
ASF GitHub Bot commented on OMID-78:
------------------------------------
Github user yonigottesman commented on a diff in the pull request:
https://github.com/apache/incubator-omid/pull/16#discussion_r153421145
--- Diff:
hbase-client/src/main/java/org/apache/omid/transaction/TTable.java ---
@@ -92,16 +95,40 @@ public TTable(Configuration conf, String tableName)
throws IOException {
this(conf, Bytes.toBytes(tableName));
}
+ public TTable(Configuration conf, String tableName, CommitTable.Client
commitTableClient) throws IOException {
+ this(conf, Bytes.toBytes(tableName), commitTableClient);
+ }
+
public TTable(HTableInterface hTable) throws IOException {
table = hTable;
healerTable = new HTable(table.getConfiguration(),
table.getTableName());
+ serverSideFilter =
table.getConfiguration().getBoolean("omid.server.side.filter", false);
+ snapshotFilter = (serverSideFilter) ? new
AttributeSetSnapshotFilter(hTable) : new SnapshotFilterImpl(new
HTableAccessWrapper(hTable, healerTable));
+ }
+
+ public TTable(HTableInterface hTable, CommitTable.Client
commitTableClient) throws IOException {
+ table = hTable;
+ healerTable = new HTable(table.getConfiguration(),
table.getTableName());
+ serverSideFilter =
table.getConfiguration().getBoolean("omid.server.side.filter", false);
--- End diff --
serverSideFilter should be local
> Identify transaction snapshot at the server
> -------------------------------------------
>
> Key: OMID-78
> URL: https://issues.apache.org/jira/browse/OMID-78
> Project: Apache Omid
> Issue Type: New Feature
> Reporter: Ohad Shacham
> Assignee: Ohad Shacham
>
> To identify a transaction's snapshot in scans and gets, Omid reads and filter
> cells at the client side. This filtering requires checking the shadow cells
> and accessing the commit table if needed.
> Doing the filtering at the server side should be more efficient since less
> data is being transferred to the client and also, in some cases, RPCs are
> being saved.
> This feature requires implementing a coprocessor that will do the filtering
> at the server side and returns only the relevant snapshot to the client side.
> This feature is also needed to support the integration of Omid with Phoenix.
> This is required since Phoenix has coprocessors that change and condence the
> data returned to the client and therefore, filtering at the client side in
> this case is impossible.
--
This message was sent by Atlassian JIRA
(v6.4.14#64029)
