[
https://issues.apache.org/jira/browse/HIVE-22239?focusedWorklogId=325331&page=com.atlassian.jira.plugin.system.issuetabpanels:worklog-tabpanel#worklog-325331
]
ASF GitHub Bot logged work on HIVE-22239:
-----------------------------------------
Author: ASF GitHub Bot
Created on: 08/Oct/19 21:12
Start Date: 08/Oct/19 21:12
Worklog Time Spent: 10m
Work Description: jcamachor commented on pull request #787: HIVE-22239
URL: https://github.com/apache/hive/pull/787#discussion_r332736102
##########
File path: ql/src/java/org/apache/hadoop/hive/ql/stats/StatsUtils.java
##########
@@ -935,8 +942,11 @@ else
if(colTypeLowerCase.equals(serdeConstants.SMALLINT_TYPE_NAME)){
cs.setNumTrues(Math.max(1, numRows/2));
cs.setNumFalses(Math.max(1, numRows/2));
cs.setAvgColLen(JavaDataModel.get().primitive1());
- } else if (colTypeLowerCase.equals(serdeConstants.TIMESTAMP_TYPE_NAME) ||
- colTypeLowerCase.equals(serdeConstants.TIMESTAMPLOCALTZ_TYPE_NAME)) {
+ } else if (colTypeLowerCase.equals(serdeConstants.TIMESTAMP_TYPE_NAME)) {
+ cs.setAvgColLen(JavaDataModel.get().lengthOfTimestamp());
+ // epoch, seconds since epoch
+ cs.setRange(0, 2177452799L);
Review comment:
I do not think this is critical, but I explored a bit and this seems to be a
poor choice for a heuristic, since in most cases it will lead to
underestimation of the data size (since most users will not have dates starting
from 1970).
I will take as lower limit `01-01-2015` (epoch for ORC) and as upper limit
`12-31-2025`. Let me know if you see any cons with this approach.
----------------------------------------------------------------
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
For queries about this service, please contact Infrastructure at:
us...@infra.apache.org
Issue Time Tracking
-------------------
Worklog Id: (was: 325331)
Time Spent: 2h 10m (was: 2h)
> Scale data size using column value ranges
> -----------------------------------------
>
> Key: HIVE-22239
> URL: https://issues.apache.org/jira/browse/HIVE-22239
> Project: Hive
> Issue Type: Improvement
> Components: Physical Optimizer
> Reporter: Jesus Camacho Rodriguez
> Assignee: Jesus Camacho Rodriguez
> Priority: Major
> Labels: pull-request-available
> Attachments: HIVE-22239.01.patch, HIVE-22239.02.patch,
> HIVE-22239.patch
>
> Time Spent: 2h 10m
> Remaining Estimate: 0h
>
> Currently, min/max values for columns are only used to determine whether a
> certain range filter falls out of range and thus filters all rows or none at
> all. If it does not, we just use a heuristic that the condition will filter
> 1/3 of the input rows. Instead of using that heuristic, we can use another
> one that assumes that data will be uniformly distributed across that range,
> and calculate the selectivity for the condition accordingly.
> This patch also includes the propagation of min/max column values from
> statistics to the optimizer for timestamp type.
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
