[HIVE-12661] StatsSetupConst.COLUMN_STATS_ACCURATE is not used correctly

• Type: Bug
• Status: Closed
• Priority: Major
• Resolution: Fixed
• Affects Version/s: 1.0.0, 1.2.1
• Fix Version/s: 2.0.0
• Assignee: Pengcheng Xiong
• Reporter: Pengcheng Xiong
• Created: 12/Dec/15 02:14
• Updated: 02/Jun/16 10:17
• Resolved: 17/Jan/16 07:01

Description

PROBLEM:
Hive stats are autogathered properly till an 'analyze table [tablename] compute statistics for columns' is run. Then it does not auto-update the stats till the command is run again. repo:

set hive.stats.autogather=true; 
set hive.stats.atomic=false ; 
set hive.stats.collect.rawdatasize=true ; 
set hive.stats.collect.scancols=false ; 
set hive.stats.collect.tablekeys=false ; 
set hive.stats.fetch.column.stats=true; 
set hive.stats.fetch.partition.stats=true ; 
set hive.stats.reliable=false ; 
set hive.compute.query.using.stats=true; 

CREATE TABLE `default`.`calendar` (`year` int) ROW FORMAT SERDE 'org.apache.hadoop.hive.ql.io.orc.OrcSerde' STORED AS INPUTFORMAT 'org.apache.hadoop.hive.ql.io.orc.OrcInputFormat' OUTPUTFORMAT 'org.apache.hadoop.hive.ql.io.orc.OrcOutputFormat' TBLPROPERTIES ( 'orc.compress'='NONE') ; 

insert into calendar values (2010), (2011), (2012); 
select * from calendar; 
+----------------+--+ 
| calendar.year | 
+----------------+--+ 
| 2010 | 
| 2011 | 
| 2012 | 
+----------------+--+ 

select max(year) from calendar; 
| 2012 | 

insert into calendar values (2013); 
select * from calendar; 
+----------------+--+ 
| calendar.year | 
+----------------+--+ 
| 2010 | 
| 2011 | 
| 2012 | 
| 2013 | 
+----------------+--+ 

select max(year) from calendar; 
| 2013 | 

insert into calendar values (2014); 
select max(year) from calendar; 
| 2014 |

analyze table calendar compute statistics for columns;

insert into calendar values (2015);
select max(year) from calendar;
| 2014 |

insert into calendar values (2016), (2017), (2018);
select max(year) from calendar;
| 2014  |

analyze table calendar compute statistics for columns;
select max(year) from calendar;
| 2018  |

Issue Links

• duplicates: HIVE-13147 COLUMN_STATS_ACCURATE is not accurate
• is related to: HIVE-3917 Support noscan operation for analyze command
• links to: RB

Sub-Tasks

• Truncated table stats should return 0 as datasize: Resolved
• rawDataSize is always 0 for some query with concurrency support: Open
• Automatically choose to use noscan for stats collection: Open
• Support basic stats and column stats in table properties in HBaseStore: Closed