okumin commented on a change in pull request #1531:
URL: https://github.com/apache/hive/pull/1531#discussion_r501101794
##########
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());
+ }
Review comment:
I wonder if we should switch `[0]` or `[1]` based on a condition. I can
see some rules use a different marker based on maybe the existence of column
stats.
----------------------------------------------------------------
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
---------------------------------------------------------------------
To unsubscribe, e-mail: gitbox-unsubscr...@hive.apache.org
For additional commands, e-mail: gitbox-h...@hive.apache.org
