[
https://issues.apache.org/jira/browse/HIVE-7991?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Prasanth J updated HIVE-7991:
-----------------------------
Labels: (was: performance)
> Incorrect calculation of number of rows in JoinStatsRule.process results in
> overflow
> ------------------------------------------------------------------------------------
>
> Key: HIVE-7991
> URL: https://issues.apache.org/jira/browse/HIVE-7991
> Project: Hive
> Issue Type: Sub-task
> Components: Statistics
> Affects Versions: 0.13.1
> Reporter: Mostafa Mokhtar
> Assignee: Prasanth J
>
> This loop results in adding the parent twice incase of a 3 way join of
> store_sales x date_dim x store
> {code}
> for (int pos = 0; pos < parents.size(); pos++) {
> ReduceSinkOperator parent = (ReduceSinkOperator)
> jop.getParentOperators().get(pos);
> Statistics parentStats = parent.getStatistics();
> List<ExprNodeDesc> keyExprs = parent.getConf().getKeyCols();
> // Parent RS may have column statistics from multiple parents.
> // Populate table alias to row count map, this will be used later
> to
> // scale down/up column statistics based on new row count
> // NOTE: JOIN with UNION as parent of RS will not have table alias
> // propagated properly. UNION operator does not propagate the
> table
> // alias of subqueries properly to expression nodes. Hence
> union20.q
> // will have wrong number of rows.
> Set<String> tableAliases =
> StatsUtils.getAllTableAlias(parent.getColumnExprMap());
> for (String tabAlias : tableAliases) {
> rowCountParents.put(tabAlias, parentStats.getNumRows());
> }
> {code}
> In the first join we have rowCountParents with {store_sales=120464862,
> date_dim=36524} which is correct.
> For the second join result rowCountParents ends up with {store=212,
> store_sales=120464862, date_dim=120464862} where it should be {store=212,
> store_sales=120464862, date_dim=36524}.
> The result of this is that computeNewRowCount ends up multiplying row count
> of store_sales x store_sales which makes the number of rows really high and
> eventually over flow.
> Plan snippet :
> {code}
> Map 1
> Map Operator Tree:
> TableScan
> alias: store_sales
> filterExpr: (((ss_sold_date_sk is not null and ss_store_sk
> is not null) and ss_item_sk is not null) and ss_sold_date BETWEEN
> '1999-06-01' AND '2000-05-31') (type: boolean)
> Statistics: Num rows: 110339135 Data size: 4817453454 Basic
> stats: COMPLETE Column stats: COMPLETE
> Filter Operator
> predicate: ((ss_sold_date_sk is not null and ss_store_sk
> is not null) and ss_item_sk is not null) (type: boolean)
> Statistics: Num rows: 107740258 Data size: 2124353556
> Basic stats: COMPLETE Column stats: COMPLETE
> Map Join Operator
> condition map:
> Inner Join 0 to 1
> condition expressions:
> 0 {ss_sold_date_sk} {ss_item_sk} {ss_store_sk}
> {ss_quantity} {ss_sales_price} {ss_sold_date}
> 1 {d_date_sk} {d_month_seq} {d_year} {d_moy} {d_qoy}
> keys:
> 0 ss_sold_date_sk (type: int)
> 1 d_date_sk (type: int)
> outputColumnNames: _col0, _col2, _col7, _col10, _col13,
> _col23, _col27, _col30, _col33, _col35, _col37
> input vertices:
> 1 Map 6
> Statistics: Num rows: 120464862 Data size: 26984129088
> Basic stats: COMPLETE Column stats: COMPLETE
> Map Join Operator
> condition map:
> Inner Join 0 to 1
> condition expressions:
> 0 {_col0} {_col2} {_col7} {_col10} {_col13}
> {_col23} {_col27} {_col30} {_col33} {_col35} {_col37}
> 1 {s_store_sk} {s_store_id}
> keys:
> 0 _col7 (type: int)
> 1 s_store_sk (type: int)
> outputColumnNames: _col0, _col2, _col7, _col10,
> _col13, _col23, _col27, _col30, _col33, _col35, _col37, _col58, _col59
> input vertices:
> 1 Map 5
> Statistics: Num rows: 17886616227069518 Data size:
> 5866810122478801920 Basic stats: COMPLETE Column stats: COMPLETE
> Map Join Operator
> condition map:
> Inner Join 0 to 1
> condition expressions:
> 0 {_col0} {_col2} {_col7} {_col10} {_col13}
> {_col23} {_col27} {_col30} {_col33} {_col35} {_col37} {_col58} {_col59}
> 1 {i_item_sk} {i_brand} {i_class} {i_category}
> {i_product_name}
> keys:
> 0 _col2 (type: int)
> 1 i_item_sk (type: int)
> outputColumnNames: _col0, _col2, _col7, _col10,
> _col13, _col23, _col27, _col30, _col33, _col35, _col37, _col58, _col59,
> _col90, _col98, _col100, _col102, _col111
> input vertices:
> 1 Map 7
> Statistics: Num rows: -9223372036854775808 Data
> size: 0 Basic stats: NONE Column stats: COMPLETE
> Filter Operator
> predicate: (((((_col0 = _col27) and (_col2 =
> _col90)) and (_col7 = _col58)) and _col30 BETWEEN 1193 AND 1204) and _col23
> BETWEEN '1999-06-01' AND '2000-05-31') (type: boolean)
> Statistics: Num rows: -9223372036854775808 Data
> size: 0 Basic stats: NONE Column stats: COMPLETE
> Select Operator
> expressions: _col102 (type: string), _col100
> (type: string), _col98 (type: string), _col111 (type: string), _col33 (type:
> int), _col37 (type: int), _col35 (type: int), _col59 (type: string), _col13
> (type: float), _col10 (type: int)
> outputColumnNames: _col102, _col100, _col98,
> _col111, _col33, _col37, _col35, _col59, _col13, _col10
> Statistics: Num rows: -9223372036854775808 Data
> size: 0 Basic stats: NONE Column stats: COMPLETE
> Group By Operator
> aggregations: sum(COALESCE((_col13 *
> _col10),0))
> keys: _col102 (type: string), _col100 (type:
> string), _col98 (type: string), _col111 (type: string), _col33 (type: int),
> _col37 (type: int), _col35 (type: int), _col59 (type: string), '0' (type:
> string)
> mode: hash
> outputColumnNames: _col0, _col1, _col2,
> _col3, _col4, _col5, _col6, _col7, _col8, _col9
> Statistics: Num rows: -9223372036854775808
> Data size: 0 Basic stats: NONE Column stats: COMPLETE
> Reduce Output Operator
> key expressions: _col0 (type: string),
> _col1 (type: string), _col2 (type: string), _col3 (type: string), _col4
> (type: int), _col5 (type: int), _col6 (type: int), _col7 (type: string),
> _col8 (type: string)
> sort order: +++++++++
> Map-reduce partition columns: _col0 (type:
> string), _col1 (type: string), _col2 (type: string), _col3 (type: string),
> _col4 (type: int), _col5 (type: int), _col6 (type: int), _col7 (type:
> string), _col8 (type: string)
> Statistics: Num rows: -9223372036854775808
> Data size: 0 Basic stats: NONE Column stats: COMPLETE
> value expressions: _col9 (type: double)
> {code}
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
