[
https://issues.apache.org/jira/browse/DRILL-3942?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Victoria Markman updated DRILL-3942:
------------------------------------
Description:
It seems to me that we should be able to do that, x is a grouping column:
{code}
0: jdbc:drill:schema=dfs> explain plan for select x, y, z from ( select
ss_sold_date_sk, ss_customer_sk, avg(ss_quantity) from store_sales group by
ss_sold_date_sk, ss_customer_sk ) as sq(x, y, z) where x is not null;
+------+------+
| text | json |
+------+------+
| 00-00 Screen
00-01 Project(x=[$0], y=[$1], z=[$2])
00-02 UnionExchange
01-01 Project(x=[$0], y=[$1], z=[$2])
01-02 Project(ss_sold_date_sk=[$0], ss_customer_sk=[$1],
EXPR$2=[CAST(/(CastHigh(CASE(=($3, 0), null, $2)), $3)):ANY NOT NULL])
01-03 SelectionVectorRemover
01-04 Filter(condition=[IS NOT NULL($0)])
01-05 HashAgg(group=[{0, 1}], agg#0=[$SUM0($2)],
agg#1=[$SUM0($3)])
01-06 Project(ss_sold_date_sk=[$0], ss_customer_sk=[$1],
$f2=[$2], $f3=[$3])
01-07 HashToRandomExchange(dist0=[[$0]], dist1=[[$1]])
02-01 UnorderedMuxExchange
03-01 Project(ss_sold_date_sk=[$0],
ss_customer_sk=[$1], $f2=[$2], $f3=[$3],
E_X_P_R_H_A_S_H_F_I_E_L_D=[castInt(hash64AsDouble($1, hash64AsDouble($0)))])
03-02 HashAgg(group=[{0, 1}], agg#0=[$SUM0($2)],
agg#1=[COUNT($2)])
03-03 Project(ss_sold_date_sk=[$2],
ss_customer_sk=[$1], ss_quantity=[$0])
03-04 Scan(groupscan=[ParquetGroupScan
[entries=[ReadEntryWithPath [path=maprfs:///tpcds1/parquet/store_sales]],
selectionRoot=maprfs:/tpcds1/parquet/store_sales, numFiles=1,
usedMetadataFile=false, columns=[`ss_sold_date_sk`, `ss_customer_sk`,
`ss_quantity`]]])
{code}
If I add another not null filter, it is pushed down:
{code}
0: jdbc:drill:schema=dfs> explain plan for select x, y, z from ( select
ss_sold_date_sk, ss_customer_sk, avg(ss_quantity) from store_sales group by
ss_sold_date_sk, ss_customer_sk ) as sq(x, y, z) where x is not null and y is
not null;
+------+------+
| text | json |
+------+------+
| 00-00 Screen
00-01 Project(x=[$0], y=[$1], z=[$2])
00-02 UnionExchange
01-01 Project(x=[$0], y=[$1], z=[$2])
01-02 Project(ss_sold_date_sk=[$0], ss_customer_sk=[$1],
EXPR$2=[CAST(/(CastHigh(CASE(=($3, 0), null, $2)), $3)):ANY NOT NULL])
01-03 HashAgg(group=[{0, 1}], agg#0=[$SUM0($2)], agg#1=[$SUM0($3)])
01-04 Project(ss_sold_date_sk=[$0], ss_customer_sk=[$1],
$f2=[$2], $f3=[$3])
01-05 HashToRandomExchange(dist0=[[$0]], dist1=[[$1]])
02-01 UnorderedMuxExchange
03-01 Project(ss_sold_date_sk=[$0], ss_customer_sk=[$1],
$f2=[$2], $f3=[$3], E_X_P_R_H_A_S_H_F_I_E_L_D=[castInt(hash64AsDouble($1,
hash64AsDouble($0)))])
03-02 HashAgg(group=[{0, 1}], agg#0=[$SUM0($2)],
agg#1=[COUNT($2)])
03-03 SelectionVectorRemover
03-04 Filter(condition=[AND(IS NOT NULL($0), IS NOT
NULL($1))])
03-05 Project(ss_sold_date_sk=[$2],
ss_customer_sk=[$1], ss_quantity=[$0])
03-06 Scan(groupscan=[ParquetGroupScan
[entries=[ReadEntryWithPath [path=maprfs:///tpcds1/parquet/store_sales]],
selectionRoot=maprfs:/tpcds1/parquet/store_sales, numFiles=1,
usedMetadataFile=false, columns=[`ss_sold_date_sk`, `ss_customer_sk`,
`ss_quantity`]]])
{code}
IS NULL filter is pushed down:
{code}
0: jdbc:drill:schema=dfs> explain plan for select x, y, z from ( select
ss_sold_date_sk, ss_customer_sk, avg(ss_quantity) from store_sales group by
ss_sold_date_sk, ss_customer_sk ) as sq(x, y, z) where x is null;
+------+------+
| text | json |
+------+------+
| 00-00 Screen
00-01 Project(x=[$0], y=[$1], z=[$2])
00-02 Project(x=[$0], y=[$1], z=[$2])
00-03 Project(ss_sold_date_sk=[$0], ss_customer_sk=[$1],
EXPR$2=[CAST(/(CastHigh(CASE(=($3, 0), null, $2)), $3)):ANY NOT NULL])
00-04 HashAgg(group=[{0, 1}], agg#0=[$SUM0($2)], agg#1=[$SUM0($3)])
00-05 HashAgg(group=[{0, 1}], agg#0=[$SUM0($2)], agg#1=[COUNT($2)])
00-06 SelectionVectorRemover
00-07 Filter(condition=[IS NULL($0)])
00-08 Project(ss_sold_date_sk=[$2], ss_customer_sk=[$1],
ss_quantity=[$0])
00-09 Scan(groupscan=[ParquetGroupScan
[entries=[ReadEntryWithPath [path=maprfs:///tpcds1/parquet/store_sales]],
selectionRoot=maprfs:/tpcds1/parquet/store_sales, numFiles=1,
usedMetadataFile=false, columns=[`ss_sold_date_sk`, `ss_customer_sk`,
`ss_quantity`]]])
{code}
was:
It seems to me that we should be able to do that, x is a grouping column:
{code}
0: jdbc:drill:schema=dfs> explain plan for select x, y, z from ( select
ss_sold_date_sk, ss_customer_sk, avg(ss_quantity) from store_sales group by
ss_sold_date_sk, ss_customer_sk ) as sq(x, y, z) where x is not null;
+------+------+
| text | json |
+------+------+
| 00-00 Screen
00-01 Project(x=[$0], y=[$1], z=[$2])
00-02 UnionExchange
01-01 Project(x=[$0], y=[$1], z=[$2])
01-02 Project(ss_sold_date_sk=[$0], ss_customer_sk=[$1],
EXPR$2=[CAST(/(CastHigh(CASE(=($3, 0), null, $2)), $3)):ANY NOT NULL])
01-03 SelectionVectorRemover
01-04 Filter(condition=[IS NOT NULL($0)])
01-05 HashAgg(group=[{0, 1}], agg#0=[$SUM0($2)],
agg#1=[$SUM0($3)])
01-06 Project(ss_sold_date_sk=[$0], ss_customer_sk=[$1],
$f2=[$2], $f3=[$3])
01-07 HashToRandomExchange(dist0=[[$0]], dist1=[[$1]])
02-01 UnorderedMuxExchange
03-01 Project(ss_sold_date_sk=[$0],
ss_customer_sk=[$1], $f2=[$2], $f3=[$3],
E_X_P_R_H_A_S_H_F_I_E_L_D=[castInt(hash64AsDouble($1, hash64AsDouble($0)))])
03-02 HashAgg(group=[{0, 1}], agg#0=[$SUM0($2)],
agg#1=[COUNT($2)])
03-03 Project(ss_sold_date_sk=[$2],
ss_customer_sk=[$1], ss_quantity=[$0])
03-04 Scan(groupscan=[ParquetGroupScan
[entries=[ReadEntryWithPath [path=maprfs:///tpcds1/parquet/store_sales]],
selectionRoot=maprfs:/tpcds1/parquet/store_sales, numFiles=1,
usedMetadataFile=false, columns=[`ss_sold_date_sk`, `ss_customer_sk`,
`ss_quantity`]]])
{code}
If I add another not null filter, it is pushed down:
{code}
0: jdbc:drill:schema=dfs> explain plan for select x, y, z from ( select
ss_sold_date_sk, ss_customer_sk, avg(ss_quantity) from store_sales group by
ss_sold_date_sk, ss_customer_sk ) as sq(x, y, z) where x is not null and y is
not null;
+------+------+
| text | json |
+------+------+
| 00-00 Screen
00-01 Project(x=[$0], y=[$1], z=[$2])
00-02 UnionExchange
01-01 Project(x=[$0], y=[$1], z=[$2])
01-02 Project(ss_sold_date_sk=[$0], ss_customer_sk=[$1],
EXPR$2=[CAST(/(CastHigh(CASE(=($3, 0), null, $2)), $3)):ANY NOT NULL])
01-03 HashAgg(group=[{0, 1}], agg#0=[$SUM0($2)], agg#1=[$SUM0($3)])
01-04 Project(ss_sold_date_sk=[$0], ss_customer_sk=[$1],
$f2=[$2], $f3=[$3])
01-05 HashToRandomExchange(dist0=[[$0]], dist1=[[$1]])
02-01 UnorderedMuxExchange
03-01 Project(ss_sold_date_sk=[$0], ss_customer_sk=[$1],
$f2=[$2], $f3=[$3], E_X_P_R_H_A_S_H_F_I_E_L_D=[castInt(hash64AsDouble($1,
hash64AsDouble($0)))])
03-02 HashAgg(group=[{0, 1}], agg#0=[$SUM0($2)],
agg#1=[COUNT($2)])
03-03 SelectionVectorRemover
03-04 Filter(condition=[AND(IS NOT NULL($0), IS NOT
NULL($1))])
03-05 Project(ss_sold_date_sk=[$2],
ss_customer_sk=[$1], ss_quantity=[$0])
03-06 Scan(groupscan=[ParquetGroupScan
[entries=[ReadEntryWithPath [path=maprfs:///tpcds1/parquet/store_sales]],
selectionRoot=maprfs:/tpcds1/parquet/store_sales, numFiles=1,
usedMetadataFile=false, columns=[`ss_sold_date_sk`, `ss_customer_sk`,
`ss_quantity`]]])
{code}
> IS NOT NULL filter is not pushed pass aggregation
> -------------------------------------------------
>
> Key: DRILL-3942
> URL: https://issues.apache.org/jira/browse/DRILL-3942
> Project: Apache Drill
> Issue Type: Bug
> Components: Query Planning & Optimization
> Affects Versions: 1.2.0
> Reporter: Victoria Markman
>
> It seems to me that we should be able to do that, x is a grouping column:
> {code}
> 0: jdbc:drill:schema=dfs> explain plan for select x, y, z from ( select
> ss_sold_date_sk, ss_customer_sk, avg(ss_quantity) from store_sales group by
> ss_sold_date_sk, ss_customer_sk ) as sq(x, y, z) where x is not null;
> +------+------+
> | text | json |
> +------+------+
> | 00-00 Screen
> 00-01 Project(x=[$0], y=[$1], z=[$2])
> 00-02 UnionExchange
> 01-01 Project(x=[$0], y=[$1], z=[$2])
> 01-02 Project(ss_sold_date_sk=[$0], ss_customer_sk=[$1],
> EXPR$2=[CAST(/(CastHigh(CASE(=($3, 0), null, $2)), $3)):ANY NOT NULL])
> 01-03 SelectionVectorRemover
> 01-04 Filter(condition=[IS NOT NULL($0)])
> 01-05 HashAgg(group=[{0, 1}], agg#0=[$SUM0($2)],
> agg#1=[$SUM0($3)])
> 01-06 Project(ss_sold_date_sk=[$0], ss_customer_sk=[$1],
> $f2=[$2], $f3=[$3])
> 01-07 HashToRandomExchange(dist0=[[$0]], dist1=[[$1]])
> 02-01 UnorderedMuxExchange
> 03-01 Project(ss_sold_date_sk=[$0],
> ss_customer_sk=[$1], $f2=[$2], $f3=[$3],
> E_X_P_R_H_A_S_H_F_I_E_L_D=[castInt(hash64AsDouble($1, hash64AsDouble($0)))])
> 03-02 HashAgg(group=[{0, 1}], agg#0=[$SUM0($2)],
> agg#1=[COUNT($2)])
> 03-03 Project(ss_sold_date_sk=[$2],
> ss_customer_sk=[$1], ss_quantity=[$0])
> 03-04 Scan(groupscan=[ParquetGroupScan
> [entries=[ReadEntryWithPath [path=maprfs:///tpcds1/parquet/store_sales]],
> selectionRoot=maprfs:/tpcds1/parquet/store_sales, numFiles=1,
> usedMetadataFile=false, columns=[`ss_sold_date_sk`, `ss_customer_sk`,
> `ss_quantity`]]])
> {code}
> If I add another not null filter, it is pushed down:
> {code}
> 0: jdbc:drill:schema=dfs> explain plan for select x, y, z from ( select
> ss_sold_date_sk, ss_customer_sk, avg(ss_quantity) from store_sales group by
> ss_sold_date_sk, ss_customer_sk ) as sq(x, y, z) where x is not null and y is
> not null;
> +------+------+
> | text | json |
> +------+------+
> | 00-00 Screen
> 00-01 Project(x=[$0], y=[$1], z=[$2])
> 00-02 UnionExchange
> 01-01 Project(x=[$0], y=[$1], z=[$2])
> 01-02 Project(ss_sold_date_sk=[$0], ss_customer_sk=[$1],
> EXPR$2=[CAST(/(CastHigh(CASE(=($3, 0), null, $2)), $3)):ANY NOT NULL])
> 01-03 HashAgg(group=[{0, 1}], agg#0=[$SUM0($2)],
> agg#1=[$SUM0($3)])
> 01-04 Project(ss_sold_date_sk=[$0], ss_customer_sk=[$1],
> $f2=[$2], $f3=[$3])
> 01-05 HashToRandomExchange(dist0=[[$0]], dist1=[[$1]])
> 02-01 UnorderedMuxExchange
> 03-01 Project(ss_sold_date_sk=[$0], ss_customer_sk=[$1],
> $f2=[$2], $f3=[$3], E_X_P_R_H_A_S_H_F_I_E_L_D=[castInt(hash64AsDouble($1,
> hash64AsDouble($0)))])
> 03-02 HashAgg(group=[{0, 1}], agg#0=[$SUM0($2)],
> agg#1=[COUNT($2)])
> 03-03 SelectionVectorRemover
> 03-04 Filter(condition=[AND(IS NOT NULL($0), IS
> NOT NULL($1))])
> 03-05 Project(ss_sold_date_sk=[$2],
> ss_customer_sk=[$1], ss_quantity=[$0])
> 03-06 Scan(groupscan=[ParquetGroupScan
> [entries=[ReadEntryWithPath [path=maprfs:///tpcds1/parquet/store_sales]],
> selectionRoot=maprfs:/tpcds1/parquet/store_sales, numFiles=1,
> usedMetadataFile=false, columns=[`ss_sold_date_sk`, `ss_customer_sk`,
> `ss_quantity`]]])
> {code}
> IS NULL filter is pushed down:
> {code}
> 0: jdbc:drill:schema=dfs> explain plan for select x, y, z from ( select
> ss_sold_date_sk, ss_customer_sk, avg(ss_quantity) from store_sales group by
> ss_sold_date_sk, ss_customer_sk ) as sq(x, y, z) where x is null;
> +------+------+
> | text | json |
> +------+------+
> | 00-00 Screen
> 00-01 Project(x=[$0], y=[$1], z=[$2])
> 00-02 Project(x=[$0], y=[$1], z=[$2])
> 00-03 Project(ss_sold_date_sk=[$0], ss_customer_sk=[$1],
> EXPR$2=[CAST(/(CastHigh(CASE(=($3, 0), null, $2)), $3)):ANY NOT NULL])
> 00-04 HashAgg(group=[{0, 1}], agg#0=[$SUM0($2)], agg#1=[$SUM0($3)])
> 00-05 HashAgg(group=[{0, 1}], agg#0=[$SUM0($2)],
> agg#1=[COUNT($2)])
> 00-06 SelectionVectorRemover
> 00-07 Filter(condition=[IS NULL($0)])
> 00-08 Project(ss_sold_date_sk=[$2], ss_customer_sk=[$1],
> ss_quantity=[$0])
> 00-09 Scan(groupscan=[ParquetGroupScan
> [entries=[ReadEntryWithPath [path=maprfs:///tpcds1/parquet/store_sales]],
> selectionRoot=maprfs:/tpcds1/parquet/store_sales, numFiles=1,
> usedMetadataFile=false, columns=[`ss_sold_date_sk`, `ss_customer_sk`,
> `ss_quantity`]]])
> {code}
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
