Mostafa Mokhtar created HIVE-7991:
-------------------------------------
Summary: 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: Bug
Components: Statistics
Affects Versions: 0.13.1
Reporter: Mostafa Mokhtar
Assignee: Prasanth J
Fix For: 0.14.0
For queries with rollup and cube the number of rows calculation in
GroupByStatsRule should be multiplied by number of elements in grouping set.
A side effect of this defect is that reducers will under estimate data size and
end up with small number of tasks which negatively affects query runtime.
{code}
// since we do not know if hash-aggregation will be enabled or
disabled
// at runtime we will assume that map-side group by does not do any
// reduction.hence no group by rule will be applied
// map-side grouping set present. if grouping set is present then
// multiply the number of rows by number of elements in grouping set
if (gop.getConf().isGroupingSetsPresent()) {
int multiplier = gop.getConf().getListGroupingSets().size();
// take into account the map-side parallelism as well, default is
1
multiplier *= mapSideParallelism;
newNumRows = multiplier * stats.getNumRows();
long dataSize = multiplier * stats.getDataSize();
stats.setNumRows(newNumRows);
stats.setDataSize(dataSize);
for (ColStatistics cs : colStats) {
if (cs != null) {
long oldNumNulls = cs.getNumNulls();
long newNumNulls = multiplier * oldNumNulls;
cs.setNumNulls(newNumNulls);
}
}
} else {
// map side no grouping set
newNumRows = stats.getNumRows() * mapSideParallelism;
updateStats(stats, newNumRows, true);
}
{code}
Query
{code}
select *
from (select i_category
,i_class
,i_brand
,i_product_name
,d_year
,d_qoy
,d_moy
,s_store_id
,sumsales
,rank() over (partition by i_category order by sumsales desc) rk
from (select i_category
,i_class
,i_brand
,i_product_name
,d_year
,d_qoy
,d_moy
,s_store_id
,sum(coalesce(ss_sales_price*ss_quantity,0)) sumsales
from store_sales
,date_dim
,store
,item
where store_sales.ss_sold_date_sk=date_dim.d_date_sk
and store_sales.ss_item_sk=item.i_item_sk
and store_sales.ss_store_sk = store.s_store_sk
and d_month_seq between 1193 and 1193+11
and ss_sold_date between '1999-06-01' and '2000-05-31'
group by i_category, i_class, i_brand, i_product_name, d_year, d_qoy,
d_moy,s_store_id with rollup)dw1) dw2
where rk <= 100
order by i_category
,i_class
,i_brand
,i_product_name
,d_year
,d_qoy
,d_moy
,s_store_id
,sumsales
,rk
limit 100
{code}
Plan generated , note the data size for Map 1
{code}
STAGE DEPENDENCIES:
Stage-1 is a root stage
Stage-0 depends on stages: Stage-1
STAGE PLANS:
Stage: Stage-1
Tez
Edges:
Map 1 <- Map 5 (BROADCAST_EDGE), Map 6 (BROADCAST_EDGE), Map 7
(BROADCAST_EDGE)
Reducer 2 <- Map 1 (SIMPLE_EDGE)
Reducer 3 <- Reducer 2 (SIMPLE_EDGE)
Reducer 4 <- Reducer 3 (SIMPLE_EDGE)
DagName: mmokhtar_20140903154848_7cf1519f-e95c-47ab-9f10-6d2130cd5734:1
Vertices:
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: NONE
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: 13792392 Data size: 602181687 Basic
stats: COMPLETE Column stats: NONE
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: 15171632 Data size: 662399872 Basic
stats: COMPLETE Column stats: NONE
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: 16688796 Data size: 728639872
Basic stats: COMPLETE Column stats: NONE
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: 18357676 Data size: 801503872
Basic stats: COMPLETE Column stats: NONE
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: 573677 Data size: 25046979
Basic stats: COMPLETE Column stats: NONE
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: 573677 Data size: 25046979
Basic stats: COMPLETE Column stats: NONE
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: 573677 Data size:
25046979 Basic stats: COMPLETE Column stats: NONE
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: 573677 Data size:
25046979 Basic stats: COMPLETE Column stats: NONE
value expressions: _col9 (type: double)
Map 5
Map Operator Tree:
TableScan
alias: store
filterExpr: s_store_sk is not null (type: boolean)
Statistics: Num rows: 212 Data size: 405680 Basic stats:
COMPLETE Column stats: COMPLETE
Filter Operator
predicate: s_store_sk is not null (type: boolean)
Statistics: Num rows: 212 Data size: 22048 Basic stats:
COMPLETE Column stats: COMPLETE
Reduce Output Operator
key expressions: s_store_sk (type: int)
sort order: +
Map-reduce partition columns: s_store_sk (type: int)
Statistics: Num rows: 212 Data size: 22048 Basic stats:
COMPLETE Column stats: COMPLETE
value expressions: s_store_id (type: string)
Execution mode: vectorized
Map 6
Map Operator Tree:
TableScan
alias: date_dim
filterExpr: (d_date_sk is not null and d_month_seq BETWEEN
1193 AND 1204) (type: boolean)
Statistics: Num rows: 73049 Data size: 81741831 Basic stats:
COMPLETE Column stats: COMPLETE
Filter Operator
predicate: (d_date_sk is not null and d_month_seq BETWEEN
1193 AND 1204) (type: boolean)
Statistics: Num rows: 36524 Data size: 730480 Basic stats:
COMPLETE Column stats: COMPLETE
Reduce Output Operator
key expressions: d_date_sk (type: int)
sort order: +
Map-reduce partition columns: d_date_sk (type: int)
Statistics: Num rows: 36524 Data size: 730480 Basic
stats: COMPLETE Column stats: COMPLETE
value expressions: d_month_seq (type: int), d_year (type:
int), d_moy (type: int), d_qoy (type: int)
Execution mode: vectorized
Map 7
Map Operator Tree:
TableScan
alias: item
filterExpr: i_item_sk is not null (type: boolean)
Statistics: Num rows: 48000 Data size: 68732712 Basic stats:
COMPLETE Column stats: COMPLETE
Filter Operator
predicate: i_item_sk is not null (type: boolean)
Statistics: Num rows: 48000 Data size: 18672000 Basic
stats: COMPLETE Column stats: COMPLETE
Reduce Output Operator
key expressions: i_item_sk (type: int)
sort order: +
Map-reduce partition columns: i_item_sk (type: int)
Statistics: Num rows: 48000 Data size: 18672000 Basic
stats: COMPLETE Column stats: COMPLETE
value expressions: i_brand (type: string), i_class (type:
string), i_category (type: string), i_product_name (type: string)
Execution mode: vectorized
Reducer 2
Reduce Operator Tree:
Group By Operator
aggregations: sum(VALUE._col0)
keys: KEY._col0 (type: string), KEY._col1 (type: string),
KEY._col2 (type: string), KEY._col3 (type: string), KEY._col4 (type: int),
KEY._col5 (type: int), KEY._col6 (type: int), KEY._col7 (type: string),
KEY._col8 (type: string)
mode: mergepartial
outputColumnNames: _col0, _col1, _col2, _col3, _col4, _col5,
_col6, _col7, _col8, _col9
Statistics: Num rows: 286838 Data size: 12523467 Basic stats:
COMPLETE Column stats: NONE
Select Operator
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), _col9 (type: double)
outputColumnNames: _col0, _col1, _col2, _col3, _col4, _col5,
_col6, _col7, _col8
Statistics: Num rows: 286838 Data size: 12523467 Basic stats:
COMPLETE Column stats: NONE
Reduce Output Operator
key expressions: _col0 (type: string), _col8 (type: double)
sort order: +-
Map-reduce partition columns: _col0 (type: string)
Statistics: Num rows: 286838 Data size: 12523467 Basic
stats: COMPLETE Column stats: NONE
TopN Hash Memory Usage: 0.04
value 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: double)
Reducer 3
Reduce Operator Tree:
Extract
Statistics: Num rows: 286838 Data size: 12523467 Basic stats:
COMPLETE Column stats: NONE
PTF Operator
Statistics: Num rows: 286838 Data size: 12523467 Basic stats:
COMPLETE Column stats: NONE
Filter Operator
predicate: (_wcol0 <= 100) (type: boolean)
Statistics: Num rows: 95612 Data size: 4174459 Basic stats:
COMPLETE Column stats: NONE
Select Operator
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: double), _wcol0
(type: int)
outputColumnNames: _col0, _col1, _col2, _col3, _col4,
_col5, _col6, _col7, _col8, _col9
Statistics: Num rows: 95612 Data size: 4174459 Basic
stats: COMPLETE Column stats: NONE
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: double),
_col9 (type: int)
sort order: ++++++++++
Statistics: Num rows: 95612 Data size: 4174459 Basic
stats: COMPLETE Column stats: NONE
TopN Hash Memory Usage: 0.04
Reducer 4
Reduce Operator Tree:
Select Operator
expressions: KEY.reducesinkkey0 (type: string),
KEY.reducesinkkey1 (type: string), KEY.reducesinkkey2 (type: string),
KEY.reducesinkkey3 (type: string), KEY.reducesinkkey4 (type: int),
KEY.reducesinkkey5 (type: int), KEY.reducesinkkey6 (type: int),
KEY.reducesinkkey7 (type: string), KEY.reducesinkkey8 (type: double),
KEY.reducesinkkey9 (type: int)
outputColumnNames: _col0, _col1, _col2, _col3, _col4, _col5,
_col6, _col7, _col8, _col9
Statistics: Num rows: 95612 Data size: 4174459 Basic stats:
COMPLETE Column stats: NONE
Limit
Number of rows: 100
Statistics: Num rows: 100 Data size: 4300 Basic stats:
COMPLETE Column stats: NONE
File Output Operator
compressed: false
Statistics: Num rows: 100 Data size: 4300 Basic stats:
COMPLETE Column stats: NONE
table:
input format: org.apache.hadoop.mapred.TextInputFormat
output format:
org.apache.hadoop.hive.ql.io.HiveIgnoreKeyTextOutputFormat
serde:
org.apache.hadoop.hive.serde2.lazy.LazySimpleSerDe
Execution mode: vectorized
Stage: Stage-0
Fetch Operator
limit: 100
Processor Tree:
ListSink
{code}
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
