impala-reviews mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Dimitris Tsirogiannis (Code Review)" <ger...@cloudera.org>
Subject [Impala-ASF-CR] IMPALA-3719: Simplify CREATE TABLE statements with Kudu tables
Date Mon, 10 Oct 2016 20:41:55 GMT
Dimitris Tsirogiannis has posted comments on this change.

Change subject: IMPALA-3719: Simplify CREATE TABLE statements with Kudu tables
......................................................................


Patch Set 3:

(1 comment)

http://gerrit.cloudera.org:8080/#/c/4414/3/common/thrift/CatalogObjects.thrift
File common/thrift/CatalogObjects.thrift:

PS3, Line 350: // Parameters needed for hash distribution
             : struct TDistributeByHashParam {
             :   1: required list<string> columns
             :   2: required i32 num_buckets
             : }
             : 
             : struct TRangeLiteral {
             :   1: optional i64 int_literal
             :   2: optional string string_literal
             : }
             : 
             : struct TRangeLiteralList {
             :   1: required list<TRangeLiteral> values
             : }
             : 
             : // A range distribution is identified by a list of columns and a series of
split rows.
             : struct TDistributeByRangeParam {
             :   1: required list<string> columns
             :   2: optional list<TRangeLiteralList> split_rows;
             : }
             : 
             : // Parameters for the DISTRIBUTE BY clause. The actual distribution is identified
by
             : // the type parameter.
             : struct TDistributeParam {
             :   // Set if type is set to HASH
             :   1: optional TDistributeByHashParam by_hash_param;
             : 
             :   // Set if type is set to RANGE
             :   2: optional TDistributeByRangeParam by_range_param;
             : }
> I don't see why this needs to be in the TKuduTable though. I think we just 
I consider the distribution scheme to be part of table metadata that are loaded in the catalog.
Hence, they live in the KuduTable object and need to be serialized before they get sent to
all the impalad nodes. Currently, the distribution schemes are only used for the SHOW CREATE
TABLE statement. They should always be set in a TKuduTable. The part that is not set if the
list of split values because they are not available from the Kudu API.


-- 
To view, visit http://gerrit.cloudera.org:8080/4414
To unsubscribe, visit http://gerrit.cloudera.org:8080/settings

Gerrit-MessageType: comment
Gerrit-Change-Id: I7b9d51b2720ab57649abdb7d5c710ea04ff50dc1
Gerrit-PatchSet: 3
Gerrit-Project: Impala-ASF
Gerrit-Branch: master
Gerrit-Owner: Dimitris Tsirogiannis <dtsirogiannis@cloudera.com>
Gerrit-Reviewer: Alex Behm <alex.behm@cloudera.com>
Gerrit-Reviewer: Dimitris Tsirogiannis <dtsirogiannis@cloudera.com>
Gerrit-Reviewer: Matthew Jacobs <mj@cloudera.com>
Gerrit-HasComments: Yes

Mime
View raw message