You are viewing a plain text version of this content. The canonical link for it is here.
Posted to notifications@asterixdb.apache.org by "Yingyi Bu (JIRA)" <ji...@apache.org> on 2016/09/09 21:01:20 UTC
[jira] [Updated] (ASTERIXDB-1637) Incorrect plan generated by left
outer index join rewriting
[ https://issues.apache.org/jira/browse/ASTERIXDB-1637?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]
Yingyi Bu updated ASTERIXDB-1637:
---------------------------------
Description:
For optimizer test asterixdb/asterix-app/src/test/resources/optimizerts/queries/inverted-index-join/issue741.aql, the optimized plan is not right.
{noformat}
for $t in dataset('TweetMessages')
where $t.send_time >= datetime('2011-06-18T14:10:17')
and
$t.send_time < datetime('2011-06-18T15:10:17')
return {
"tweet": $t.tweetid,
"similar-tweets": for $t2 in dataset('TweetMessages')
let $sim := similarity-jaccard-check($t.referred_topics, $t2.referred_topics, 0.6f)
where $sim[0] and
$t2.tweetid != $t.tweetid
return $t2.tweetid
}
{noformat}
{noformat}
distribute result [%0->$$11]
-- DISTRIBUTE_RESULT |PARTITIONED|
exchange
-- ONE_TO_ONE_EXCHANGE |PARTITIONED|
project ([$$11])
-- STREAM_PROJECT |PARTITIONED|
assign [$$11] <- [function-call: asterix:closed-record-constructor, Args:[AString: {tweet}, %0->$$33, AString: {similar-tweets}, %0->$$23]]
-- ASSIGN |PARTITIONED|
exchange
-- ONE_TO_ONE_EXCHANGE |PARTITIONED|
group by ([$$33 := %0->$$25]) decor ([]) {
aggregate [$$23] <- [function-call: asterix:listify, Args:[%0->$$26]]
-- AGGREGATE |LOCAL|
select (function-call: algebricks:not, Args:[function-call: algebricks:is-missing, Args:[%0->$$26]])
-- STREAM_SELECT |LOCAL|
nested tuple source
-- NESTED_TUPLE_SOURCE |LOCAL|
}
-- PRE_CLUSTERED_GROUP_BY[$$25] |PARTITIONED|
exchange
-- ONE_TO_ONE_EXCHANGE |PARTITIONED|
order (ASC, %0->$$25)
-- STABLE_SORT [$$25(ASC)] |PARTITIONED|
exchange
-- HASH_PARTITION_EXCHANGE [$$25] |PARTITIONED|
project ([$$25, $$26])
-- STREAM_PROJECT |PARTITIONED|
exchange
-- ONE_TO_ONE_EXCHANGE |PARTITIONED|
join (function-call: algebricks:eq, Args:[%0->$$36, %0->$$25])
-- HYBRID_HASH_JOIN [$$36][$$25] |PARTITIONED|
exchange
-- ONE_TO_ONE_EXCHANGE |PARTITIONED|
project ([$$36])
-- STREAM_PROJECT |PARTITIONED|
select (function-call: algebricks:and, Args:[function-call: algebricks:ge, Args:[%0->$$24, ADateTime: { 2011-06-18T14:10:17.000Z }], function-call: algebricks:lt, Args:[%0->$$24, ADateTime: { 2011-06-18T15:10:17.000Z }]])
-- STREAM_SELECT |PARTITIONED|
project ([$$36, $$24])
-- STREAM_PROJECT |PARTITIONED|
assign [$$24] <- [function-call: asterix:field-access-by-index, Args:[%0->$$0, AInt32: {3}]]
-- ASSIGN |PARTITIONED|
exchange
-- ONE_TO_ONE_EXCHANGE |PARTITIONED|
data-scan []<-[$$36, $$0] <- test:TweetMessages
-- DATASOURCE_SCAN |PARTITIONED|
exchange
-- ONE_TO_ONE_EXCHANGE |PARTITIONED|
empty-tuple-source
-- EMPTY_TUPLE_SOURCE |PARTITIONED|
exchange
-- HASH_PARTITION_EXCHANGE [$$25] |PARTITIONED|
project ([$$25, $$26])
-- STREAM_PROJECT |PARTITIONED|
select (function-call: algebricks:and, Args:[function-call: algebricks:neq, Args:[%0->$$26, %0->$$25], function-call: asterix:get-item, Args:[function-call: asterix:similarity-jaccard-check, Args:[%0->$$29, function-call: asterix:field-access-by-index, Args:[%0->$$1, AInt32: {4}], AFloat: {0.6}], AInt64: {0}]])
-- STREAM_SELECT |PARTITIONED|
project ([$$1, $$25, $$26, $$29])
-- STREAM_PROJECT |PARTITIONED|
exchange
-- ONE_TO_ONE_EXCHANGE |PARTITIONED|
left-outer-unnest-map [$$26, $$1] <- function-call: asterix:index-search, Args:[AString: {TweetMessages}, AInt32: {0}, AString: {test}, AString: {TweetMessages}, ABoolean: {true}, ABoolean: {false}, AInt32: {1}, %0->$$39, AInt32: {1}, %0->$$39, TRUE, TRUE, TRUE]
-- BTREE_SEARCH |PARTITIONED|
exchange
-- ONE_TO_ONE_EXCHANGE |PARTITIONED|
order (ASC, %0->$$39)
-- STABLE_SORT [$$39(ASC)] |PARTITIONED|
exchange
-- ONE_TO_ONE_EXCHANGE |PARTITIONED|
left-outer-unnest-map [$$39] <- function-call: asterix:index-search, Args:[AString: {topicIIx}, AInt32: {4}, AString: {test}, AString: {TweetMessages}, ABoolean: {true}, ABoolean: {true}, AInt32: {1}, AFloat: {0.6}, AInt32: {22}, AInt32: {1}, %0->$$29]
-- LENGTH_PARTITIONED_INVERTED_INDEX_SEARCH |PARTITIONED|
exchange
-- BROADCAST_EXCHANGE |PARTITIONED|
project ([$$25, $$29])
-- STREAM_PROJECT |PARTITIONED|
select (function-call: algebricks:and, Args:[function-call: algebricks:ge, Args:[%0->$$37, ADateTime: { 2011-06-18T14:10:17.000Z }], function-call: algebricks:lt, Args:[%0->$$37, ADateTime: { 2011-06-18T15:10:17.000Z }]])
-- STREAM_SELECT |PARTITIONED|
project ([$$37, $$25, $$29])
-- STREAM_PROJECT |PARTITIONED|
assign [$$29, $$37] <- [function-call: asterix:field-access-by-index, Args:[%0->$$38, AInt32: {4}], function-call: asterix:field-access-by-index, Args:[%0->$$38, AInt32: {3}]]
-- ASSIGN |PARTITIONED|
exchange
-- ONE_TO_ONE_EXCHANGE |PARTITIONED|
data-scan []<-[$$25, $$38] <- test:TweetMessages
-- DATASOURCE_SCAN |PARTITIONED|
exchange
-- ONE_TO_ONE_EXCHANGE |PARTITIONED|
empty-tuple-source
-- EMPTY_TUPLE_SOURCE |PARTITIONED|
{noformat}
There are several issues here:
1. The filtering condition on $t.send_time gets pushed to both input branches. It should *only* be on the outer branch, rather than the inner branch.
2. The left_outer_unnest_maps in the plan should be unnest_map.
3. The join in the plan should be a left outer join instead of an inner join.
was:
For optimizer test asterixdb/asterix-app/src/test/resources/optimizerts/queries/inverted-index-join/issue741.aql, the optimized plan is not right.
{noformat}
for $t in dataset('TweetMessages')
where $t.send_time >= datetime('2011-06-18T14:10:17')
and
$t.send_time < datetime('2011-06-18T15:10:17')
return {
"tweet": $t.tweetid,
"similar-tweets": for $t2 in dataset('TweetMessages')
let $sim := similarity-jaccard-check($t.referred_topics, $t2.referred_topics, 0.6f)
where $sim[0] and
$t2.tweetid != $t.tweetid
return $t2.tweetid
}
{noformat}
{noformat}
distribute result [%0->$$11]
-- DISTRIBUTE_RESULT |PARTITIONED|
exchange
-- ONE_TO_ONE_EXCHANGE |PARTITIONED|
project ([$$11])
-- STREAM_PROJECT |PARTITIONED|
assign [$$11] <- [function-call: asterix:closed-record-constructor, Args:[AString: {tweet}, %0->$$33, AString: {similar-tweets}, %0->$$23]]
-- ASSIGN |PARTITIONED|
exchange
-- ONE_TO_ONE_EXCHANGE |PARTITIONED|
group by ([$$33 := %0->$$25]) decor ([]) {
aggregate [$$23] <- [function-call: asterix:listify, Args:[%0->$$26]]
-- AGGREGATE |LOCAL|
select (function-call: algebricks:not, Args:[function-call: algebricks:is-missing, Args:[%0->$$26]])
-- STREAM_SELECT |LOCAL|
nested tuple source
-- NESTED_TUPLE_SOURCE |LOCAL|
}
-- PRE_CLUSTERED_GROUP_BY[$$25] |PARTITIONED|
exchange
-- ONE_TO_ONE_EXCHANGE |PARTITIONED|
order (ASC, %0->$$25)
-- STABLE_SORT [$$25(ASC)] |PARTITIONED|
exchange
-- HASH_PARTITION_EXCHANGE [$$25] |PARTITIONED|
project ([$$25, $$26])
-- STREAM_PROJECT |PARTITIONED|
exchange
-- ONE_TO_ONE_EXCHANGE |PARTITIONED|
join (function-call: algebricks:eq, Args:[%0->$$36, %0->$$25])
-- HYBRID_HASH_JOIN [$$36][$$25] |PARTITIONED|
exchange
-- ONE_TO_ONE_EXCHANGE |PARTITIONED|
project ([$$36])
-- STREAM_PROJECT |PARTITIONED|
select (function-call: algebricks:and, Args:[function-call: algebricks:ge, Args:[%0->$$24, ADateTime: { 2011-06-18T14:10:17.000Z }], function-call: algebricks:lt, Args:[%0->$$24, ADateTime: { 2011-06-18T15:10:17.000Z }]])
-- STREAM_SELECT |PARTITIONED|
project ([$$36, $$24])
-- STREAM_PROJECT |PARTITIONED|
assign [$$24] <- [function-call: asterix:field-access-by-index, Args:[%0->$$0, AInt32: {3}]]
-- ASSIGN |PARTITIONED|
exchange
-- ONE_TO_ONE_EXCHANGE |PARTITIONED|
data-scan []<-[$$36, $$0] <- test:TweetMessages
-- DATASOURCE_SCAN |PARTITIONED|
exchange
-- ONE_TO_ONE_EXCHANGE |PARTITIONED|
empty-tuple-source
-- EMPTY_TUPLE_SOURCE |PARTITIONED|
exchange
-- HASH_PARTITION_EXCHANGE [$$25] |PARTITIONED|
project ([$$25, $$26])
-- STREAM_PROJECT |PARTITIONED|
select (function-call: algebricks:and, Args:[function-call: algebricks:neq, Args:[%0->$$26, %0->$$25], function-call: asterix:get-item, Args:[function-call: asterix:similarity-jaccard-check, Args:[%0->$$29, function-call: asterix:field-access-by-index, Args:[%0->$$1, AInt32: {4}], AFloat: {0.6}], AInt64: {0}]])
-- STREAM_SELECT |PARTITIONED|
project ([$$1, $$25, $$26, $$29])
-- STREAM_PROJECT |PARTITIONED|
exchange
-- ONE_TO_ONE_EXCHANGE |PARTITIONED|
left-outer-unnest-map [$$26, $$1] <- function-call: asterix:index-search, Args:[AString: {TweetMessages}, AInt32: {0}, AString: {test}, AString: {TweetMessages}, ABoolean: {true}, ABoolean: {false}, AInt32: {1}, %0->$$39, AInt32: {1}, %0->$$39, TRUE, TRUE, TRUE]
-- BTREE_SEARCH |PARTITIONED|
exchange
-- ONE_TO_ONE_EXCHANGE |PARTITIONED|
order (ASC, %0->$$39)
-- STABLE_SORT [$$39(ASC)] |PARTITIONED|
exchange
-- ONE_TO_ONE_EXCHANGE |PARTITIONED|
left-outer-unnest-map [$$39] <- function-call: asterix:index-search, Args:[AString: {topicIIx}, AInt32: {4}, AString: {test}, AString: {TweetMessages}, ABoolean: {true}, ABoolean: {true}, AInt32: {1}, AFloat: {0.6}, AInt32: {22}, AInt32: {1}, %0->$$29]
-- LENGTH_PARTITIONED_INVERTED_INDEX_SEARCH |PARTITIONED|
exchange
-- BROADCAST_EXCHANGE |PARTITIONED|
project ([$$25, $$29])
-- STREAM_PROJECT |PARTITIONED|
select (function-call: algebricks:and, Args:[function-call: algebricks:ge, Args:[%0->$$37, ADateTime: { 2011-06-18T14:10:17.000Z }], function-call: algebricks:lt, Args:[%0->$$37, ADateTime: { 2011-06-18T15:10:17.000Z }]])
-- STREAM_SELECT |PARTITIONED|
project ([$$37, $$25, $$29])
-- STREAM_PROJECT |PARTITIONED|
assign [$$29, $$37] <- [function-call: asterix:field-access-by-index, Args:[%0->$$38, AInt32: {4}], function-call: asterix:field-access-by-index, Args:[%0->$$38, AInt32: {3}]]
-- ASSIGN |PARTITIONED|
exchange
-- ONE_TO_ONE_EXCHANGE |PARTITIONED|
data-scan []<-[$$25, $$38] <- test:TweetMessages
-- DATASOURCE_SCAN |PARTITIONED|
exchange
-- ONE_TO_ONE_EXCHANGE |PARTITIONED|
empty-tuple-source
-- EMPTY_TUPLE_SOURCE |PARTITIONED|
{noformat}
There are several issues here:
1. The filtering condition on $t.send_time gets pushed to a wrong input branch. It should be on the outer branch, rather than the inner branch.
2. The left_outer_unnest_maps in the plan should be unnest_map.
3. The join in the plan should be a left outer join instead of an inner join.
> Incorrect plan generated by left outer index join rewriting
> -----------------------------------------------------------
>
> Key: ASTERIXDB-1637
> URL: https://issues.apache.org/jira/browse/ASTERIXDB-1637
> Project: Apache AsterixDB
> Issue Type: Bug
> Components: Optimizer
> Reporter: Yingyi Bu
> Assignee: Taewoo Kim
>
> For optimizer test asterixdb/asterix-app/src/test/resources/optimizerts/queries/inverted-index-join/issue741.aql, the optimized plan is not right.
> {noformat}
> for $t in dataset('TweetMessages')
> where $t.send_time >= datetime('2011-06-18T14:10:17')
> and
> $t.send_time < datetime('2011-06-18T15:10:17')
> return {
> "tweet": $t.tweetid,
> "similar-tweets": for $t2 in dataset('TweetMessages')
> let $sim := similarity-jaccard-check($t.referred_topics, $t2.referred_topics, 0.6f)
> where $sim[0] and
> $t2.tweetid != $t.tweetid
> return $t2.tweetid
> }
> {noformat}
> {noformat}
> distribute result [%0->$$11]
> -- DISTRIBUTE_RESULT |PARTITIONED|
> exchange
> -- ONE_TO_ONE_EXCHANGE |PARTITIONED|
> project ([$$11])
> -- STREAM_PROJECT |PARTITIONED|
> assign [$$11] <- [function-call: asterix:closed-record-constructor, Args:[AString: {tweet}, %0->$$33, AString: {similar-tweets}, %0->$$23]]
> -- ASSIGN |PARTITIONED|
> exchange
> -- ONE_TO_ONE_EXCHANGE |PARTITIONED|
> group by ([$$33 := %0->$$25]) decor ([]) {
> aggregate [$$23] <- [function-call: asterix:listify, Args:[%0->$$26]]
> -- AGGREGATE |LOCAL|
> select (function-call: algebricks:not, Args:[function-call: algebricks:is-missing, Args:[%0->$$26]])
> -- STREAM_SELECT |LOCAL|
> nested tuple source
> -- NESTED_TUPLE_SOURCE |LOCAL|
> }
> -- PRE_CLUSTERED_GROUP_BY[$$25] |PARTITIONED|
> exchange
> -- ONE_TO_ONE_EXCHANGE |PARTITIONED|
> order (ASC, %0->$$25)
> -- STABLE_SORT [$$25(ASC)] |PARTITIONED|
> exchange
> -- HASH_PARTITION_EXCHANGE [$$25] |PARTITIONED|
> project ([$$25, $$26])
> -- STREAM_PROJECT |PARTITIONED|
> exchange
> -- ONE_TO_ONE_EXCHANGE |PARTITIONED|
> join (function-call: algebricks:eq, Args:[%0->$$36, %0->$$25])
> -- HYBRID_HASH_JOIN [$$36][$$25] |PARTITIONED|
> exchange
> -- ONE_TO_ONE_EXCHANGE |PARTITIONED|
> project ([$$36])
> -- STREAM_PROJECT |PARTITIONED|
> select (function-call: algebricks:and, Args:[function-call: algebricks:ge, Args:[%0->$$24, ADateTime: { 2011-06-18T14:10:17.000Z }], function-call: algebricks:lt, Args:[%0->$$24, ADateTime: { 2011-06-18T15:10:17.000Z }]])
> -- STREAM_SELECT |PARTITIONED|
> project ([$$36, $$24])
> -- STREAM_PROJECT |PARTITIONED|
> assign [$$24] <- [function-call: asterix:field-access-by-index, Args:[%0->$$0, AInt32: {3}]]
> -- ASSIGN |PARTITIONED|
> exchange
> -- ONE_TO_ONE_EXCHANGE |PARTITIONED|
> data-scan []<-[$$36, $$0] <- test:TweetMessages
> -- DATASOURCE_SCAN |PARTITIONED|
> exchange
> -- ONE_TO_ONE_EXCHANGE |PARTITIONED|
> empty-tuple-source
> -- EMPTY_TUPLE_SOURCE |PARTITIONED|
> exchange
> -- HASH_PARTITION_EXCHANGE [$$25] |PARTITIONED|
> project ([$$25, $$26])
> -- STREAM_PROJECT |PARTITIONED|
> select (function-call: algebricks:and, Args:[function-call: algebricks:neq, Args:[%0->$$26, %0->$$25], function-call: asterix:get-item, Args:[function-call: asterix:similarity-jaccard-check, Args:[%0->$$29, function-call: asterix:field-access-by-index, Args:[%0->$$1, AInt32: {4}], AFloat: {0.6}], AInt64: {0}]])
> -- STREAM_SELECT |PARTITIONED|
> project ([$$1, $$25, $$26, $$29])
> -- STREAM_PROJECT |PARTITIONED|
> exchange
> -- ONE_TO_ONE_EXCHANGE |PARTITIONED|
> left-outer-unnest-map [$$26, $$1] <- function-call: asterix:index-search, Args:[AString: {TweetMessages}, AInt32: {0}, AString: {test}, AString: {TweetMessages}, ABoolean: {true}, ABoolean: {false}, AInt32: {1}, %0->$$39, AInt32: {1}, %0->$$39, TRUE, TRUE, TRUE]
> -- BTREE_SEARCH |PARTITIONED|
> exchange
> -- ONE_TO_ONE_EXCHANGE |PARTITIONED|
> order (ASC, %0->$$39)
> -- STABLE_SORT [$$39(ASC)] |PARTITIONED|
> exchange
> -- ONE_TO_ONE_EXCHANGE |PARTITIONED|
> left-outer-unnest-map [$$39] <- function-call: asterix:index-search, Args:[AString: {topicIIx}, AInt32: {4}, AString: {test}, AString: {TweetMessages}, ABoolean: {true}, ABoolean: {true}, AInt32: {1}, AFloat: {0.6}, AInt32: {22}, AInt32: {1}, %0->$$29]
> -- LENGTH_PARTITIONED_INVERTED_INDEX_SEARCH |PARTITIONED|
> exchange
> -- BROADCAST_EXCHANGE |PARTITIONED|
> project ([$$25, $$29])
> -- STREAM_PROJECT |PARTITIONED|
> select (function-call: algebricks:and, Args:[function-call: algebricks:ge, Args:[%0->$$37, ADateTime: { 2011-06-18T14:10:17.000Z }], function-call: algebricks:lt, Args:[%0->$$37, ADateTime: { 2011-06-18T15:10:17.000Z }]])
> -- STREAM_SELECT |PARTITIONED|
> project ([$$37, $$25, $$29])
> -- STREAM_PROJECT |PARTITIONED|
> assign [$$29, $$37] <- [function-call: asterix:field-access-by-index, Args:[%0->$$38, AInt32: {4}], function-call: asterix:field-access-by-index, Args:[%0->$$38, AInt32: {3}]]
> -- ASSIGN |PARTITIONED|
> exchange
> -- ONE_TO_ONE_EXCHANGE |PARTITIONED|
> data-scan []<-[$$25, $$38] <- test:TweetMessages
> -- DATASOURCE_SCAN |PARTITIONED|
> exchange
> -- ONE_TO_ONE_EXCHANGE |PARTITIONED|
> empty-tuple-source
> -- EMPTY_TUPLE_SOURCE |PARTITIONED|
> {noformat}
> There are several issues here:
> 1. The filtering condition on $t.send_time gets pushed to both input branches. It should *only* be on the outer branch, rather than the inner branch.
> 2. The left_outer_unnest_maps in the plan should be unnest_map.
> 3. The join in the plan should be a left outer join instead of an inner join.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)