[
https://issues.apache.org/jira/browse/HIVE-24203?focusedWorklogId=497313&page=com.atlassian.jira.plugin.system.issuetabpanels:worklog-tabpanel#worklog-497313
]
ASF GitHub Bot logged work on HIVE-24203:
-----------------------------------------
Author: ASF GitHub Bot
Created on: 08/Oct/20 12:56
Start Date: 08/Oct/20 12:56
Worklog Time Spent: 10m
Work Description: okumin commented on a change in pull request #1531:
URL: https://github.com/apache/hive/pull/1531#discussion_r501697579
##########
File path:
ql/src/java/org/apache/hadoop/hive/ql/optimizer/stats/annotation/StatsRulesProcFactory.java
##########
@@ -2921,6 +2920,97 @@ public Object process(Node nd, Stack<Node> stack,
NodeProcessorCtx procCtx,
}
}
+ /**
+ * LateralViewJoinOperator changes the data size and column level statistics.
+ *
+ * A diagram of LATERAL VIEW.
+ *
+ * [Lateral View Forward]
+ * / \
+ * [Select] [Select]
+ * | |
+ * | [UDTF]
+ * \ /
+ * [Lateral View Join]
+ *
+ * For each row of the source, the left branch just picks columns and the
right branch processes UDTF.
+ * And then LVJ joins a row from the left branch with rows from the right
branch.
+ * The join has one-to-many relationship since UDTF can generate multiple
rows.
+ *
+ * This rule multiplies the stats from the left branch by T(right) / T(left)
and sums up the both sides.
+ */
+ public static class LateralViewJoinStatsRule extends DefaultStatsRule
implements SemanticNodeProcessor {
+ @Override
+ public Object process(Node nd, Stack<Node> stack, NodeProcessorCtx procCtx,
+ Object... nodeOutputs) throws SemanticException {
+ final LateralViewJoinOperator lop = (LateralViewJoinOperator) nd;
+ final AnnotateStatsProcCtx aspCtx = (AnnotateStatsProcCtx) procCtx;
+ final HiveConf conf = aspCtx.getConf();
+
+ if (!isAllParentsContainStatistics(lop)) {
+ return null;
+ }
+
+ final List<Operator<? extends OperatorDesc>> parents =
lop.getParentOperators();
+ if (parents.size() != 2) {
+ LOG.warn("LateralViewJoinOperator should have just two parents but
actually has "
+ + parents.size() + " parents.");
+ return null;
+ }
+
+ final Statistics selectStats =
parents.get(LateralViewJoinOperator.SELECT_TAG).getStatistics();
+ final Statistics udtfStats =
parents.get(LateralViewJoinOperator.UDTF_TAG).getStatistics();
+
+ final double factor = (double) udtfStats.getNumRows() / (double)
selectStats.getNumRows();
+ final long selectDataSize =
StatsUtils.safeMult(selectStats.getDataSize(), factor);
+ final long dataSize = StatsUtils.safeAdd(selectDataSize,
udtfStats.getDataSize());
+ Statistics joinedStats = new Statistics(udtfStats.getNumRows(),
dataSize, 0, 0);
+
+ if (satisfyPrecondition(selectStats) && satisfyPrecondition(udtfStats)) {
+ final Map<String, ExprNodeDesc> columnExprMap = lop.getColumnExprMap();
+ final RowSchema schema = lop.getSchema();
+
+ joinedStats.updateColumnStatsState(selectStats.getColumnStatsState());
+ final List<ColStatistics> selectColStats = StatsUtils
+ .getColStatisticsFromExprMap(conf, selectStats, columnExprMap,
schema);
+ joinedStats.addToColumnStats(multiplyColStats(selectColStats, factor));
+
+ joinedStats.updateColumnStatsState(udtfStats.getColumnStatsState());
+ final List<ColStatistics> udtfColStats = StatsUtils
+ .getColStatisticsFromExprMap(conf, udtfStats, columnExprMap,
schema);
+ joinedStats.addToColumnStats(udtfColStats);
+
+ joinedStats = applyRuntimeStats(aspCtx.getParseContext().getContext(),
joinedStats, lop);
+ lop.setStatistics(joinedStats);
+
+ if (LOG.isDebugEnabled()) {
+ LOG.debug("[0] STATS-" + lop.toString() + ": " +
joinedStats.extendedToString());
+ }
+ } else {
+ joinedStats = applyRuntimeStats(aspCtx.getParseContext().getContext(),
joinedStats, lop);
+ lop.setStatistics(joinedStats);
+
+ if (LOG.isDebugEnabled()) {
+ LOG.debug("[1] STATS-" + lop.toString() + ": " +
joinedStats.extendedToString());
+ }
+ }
+ return null;
+ }
+
+ private List<ColStatistics> multiplyColStats(List<ColStatistics>
colStatistics, double factor) {
+ for (ColStatistics colStats : colStatistics) {
+ colStats.setNumFalses(StatsUtils.safeMult(colStats.getNumFalses(),
factor));
+ colStats.setNumTrues(StatsUtils.safeMult(colStats.getNumTrues(),
factor));
+ colStats.setNumNulls(StatsUtils.safeMult(colStats.getNumNulls(),
factor));
+ // When factor > 1, the same records are duplicated and countDistinct
never changes.
+ if (factor < 1.0) {
+
colStats.setCountDistint(StatsUtils.safeMult(colStats.getCountDistint(),
factor));
Review comment:
Ceiled. I moved this method since I'd like to reuse it for HIVE-24240.
https://github.com/apache/hive/commit/50396346eaed5d6bab4ff87dd079918a769a7ebd
----------------------------------------------------------------
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: 497313)
Time Spent: 2.5h (was: 2h 20m)
> Implement stats annotation rule for the LateralViewJoinOperator
> ---------------------------------------------------------------
>
> Key: HIVE-24203
> URL: https://issues.apache.org/jira/browse/HIVE-24203
> Project: Hive
> Issue Type: Improvement
> Components: Physical Optimizer
> Affects Versions: 4.0.0, 3.1.2, 2.3.7
> Reporter: okumin
> Assignee: okumin
> Priority: Major
> Labels: pull-request-available
> Time Spent: 2.5h
> Remaining Estimate: 0h
>
> StatsRulesProcFactory doesn't have any rules to handle a JOIN by LATERAL VIEW.
> This can cause an underestimation in case that UDTF in LATERAL VIEW generates
> multiple rows.
> HIVE-20262 has already added the rule for UDTF.
> This issue would add the rule for LateralViewJoinOperator.
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
