jon-wei commented on a change in pull request #7133: 6088 - Time Ordering On
Scans
URL: https://github.com/apache/incubator-druid/pull/7133#discussion_r263639613
##########
File path:
processing/src/main/java/org/apache/druid/query/scan/ScanQueryRunnerFactory.java
##########
@@ -68,34 +81,121 @@ public ScanQueryRunnerFactory(
)
{
// in single thread and in jetty thread instead of processing thread
- return new QueryRunner<ScanResultValue>()
- {
- @Override
- public Sequence<ScanResultValue> run(
- final QueryPlus<ScanResultValue> queryPlus, final Map<String,
Object> responseContext
- )
- {
- // Note: this variable is effective only when queryContext has a
timeout.
- // See the comment of CTX_TIMEOUT_AT.
- final long timeoutAt = System.currentTimeMillis() +
QueryContexts.getTimeout(queryPlus.getQuery());
- responseContext.put(CTX_TIMEOUT_AT, timeoutAt);
+ return (queryPlus, responseContext) -> {
+ ScanQuery query = (ScanQuery) queryPlus.getQuery();
+ int numSegments = 0;
+ final Iterator<QueryRunner<ScanResultValue>> segmentIt =
queryRunners.iterator();
+ for (; segmentIt.hasNext(); numSegments++) {
+ segmentIt.next();
+ }
+ // Note: this variable is effective only when queryContext has a timeout.
+ // See the comment of CTX_TIMEOUT_AT.
+ final long timeoutAt = System.currentTimeMillis() +
QueryContexts.getTimeout(queryPlus.getQuery());
+ responseContext.put(CTX_TIMEOUT_AT, timeoutAt);
+ if (query.getTimeOrder().equals(ScanQuery.TimeOrder.NONE)) {
+ // Use normal strategy
return Sequences.concat(
Sequences.map(
Sequences.simple(queryRunners),
- new Function<QueryRunner<ScanResultValue>,
Sequence<ScanResultValue>>()
- {
- @Override
- public Sequence<ScanResultValue> apply(final
QueryRunner<ScanResultValue> input)
- {
- return input.run(queryPlus, responseContext);
- }
- }
+ input -> input.run(queryPlus, responseContext)
)
);
+ } else if (query.getLimit() <=
scanQueryConfig.getMaxRowsQueuedForTimeOrdering()) {
+ // Use priority queue strategy
+ return sortAndLimitScanResultValues(
+ Sequences.concat(Sequences.map(
+ Sequences.simple(queryRunners),
+ input -> input.run(queryPlus, responseContext)
+ )),
+ query
+ );
+ } else if (numSegments <=
scanQueryConfig.getMaxSegmentsTimeOrderedInMemory()) {
+ // Use n-way merge strategy
+ final Sequence<ScanResultValue> unbatched =
+ Sequences.map(
+ Sequences.simple(queryRunners),
+ (input) -> Sequences.concat(
+ Sequences.map(
+ input.run(queryPlus, responseContext),
+ srv ->
Sequences.simple(srv.toSingleEventScanResultValues())
+ )
+ )
+ ).flatMerge(
+ seq -> seq,
+ Ordering.from(new ScanResultValueTimestampComparator(
+ query
+ )).reverse() // This needs to be reversed because
+ ).limit(
+ Math.toIntExact(query.getLimit())
+ );
+
+ return unbatched;
+ } else if (query.getLimit() >
scanQueryConfig.getMaxRowsQueuedForTimeOrdering()) {
Review comment:
I think these UOE throwing blocks should be combined, since if you reach
this part both the row limit and segment limit were exceeded
----------------------------------------------------------------
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:
[email protected]
With regards,
Apache Git Services
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
