Skip to content
项目
群组
代码片段
帮助
正在加载...
帮助
为 GitLab 提交贡献
登录/注册
切换导航
H
h2database
项目
项目
详情
活动
周期分析
仓库
仓库
文件
提交
分支
标签
贡献者
分枝图
比较
统计图
议题
0
议题
0
列表
看板
标记
里程碑
合并请求
0
合并请求
0
CI / CD
CI / CD
流水线
作业
计划
统计图
Wiki
Wiki
代码片段
代码片段
成员
成员
折叠边栏
关闭边栏
活动
分枝图
统计图
创建新议题
作业
提交
议题看板
打开侧边栏
Administrator
h2database
Commits
992f5c7f
Unverified
提交
992f5c7f
authored
2月 01, 2019
作者:
Evgenij Ryazanov
提交者:
GitHub
2月 01, 2019
浏览文件
操作
浏览文件
下载
差异文件
Merge pull request #1711 from katzyn/qualify
Add QUALIFY clause to SELECT command
上级
5f8b6139
f77a7550
隐藏空白字符变更
内嵌
并排
正在显示
11 个修改的文件
包含
181 行增加
和
51 行删除
+181
-51
help.csv
h2/src/docsrc/help/help.csv
+23
-12
advanced.html
h2/src/docsrc/html/advanced.html
+3
-3
changelog.html
h2/src/docsrc/html/changelog.html
+2
-0
Parser.java
h2/src/main/org/h2/command/Parser.java
+8
-0
Select.java
h2/src/main/org/h2/command/dml/Select.java
+69
-29
JdbcDatabaseMetaData.java
h2/src/main/org/h2/jdbc/JdbcDatabaseMetaData.java
+5
-5
ParserUtil.java
h2/src/main/org/h2/util/ParserUtil.java
+11
-1
TestMetaData.java
h2/src/test/org/h2/test/jdbc/TestMetaData.java
+1
-1
select.sql
h2/src/test/org/h2/test/scripts/dml/select.sql
+16
-0
array-agg.sql
...est/org/h2/test/scripts/functions/aggregate/array-agg.sql
+16
-0
row_number.sql
.../test/org/h2/test/scripts/functions/window/row_number.sql
+27
-0
没有找到文件。
h2/src/docsrc/help/help.csv
浏览文件 @
992f5c7f
...
...
@@ -6,10 +6,12 @@
SELECT [ TOP term [ PERCENT ] [ WITH TIES ] ]
[ DISTINCT [ ON ( expression [,...] ) ] | ALL ]
selectExpression [,...]
FROM tableExpression [,...] [ WHERE expression ]
[ FROM tableExpression [,...] ]
[ WHERE expression ]
[ GROUP BY expression [,...] ] [ HAVING expression ]
[ QUALIFY expression ]
[ WINDOW { { windowName AS windowSpecification } [,...] } ]
[ { UNION [ ALL ] |
MINUS | EXCEPT
| INTERSECT } select ]
[ { UNION [ ALL ] |
EXCEPT | MINUS
| INTERSECT } select ]
[ ORDER BY order [,...] ]
[ LIMIT expression [ OFFSET expression ] [ SAMPLE_SIZE rowCountInt ] ]
[ [ OFFSET expression { ROW | ROWS } ]
...
...
@@ -18,26 +20,35 @@ FROM tableExpression [,...] [ WHERE expression ]
[ FOR UPDATE ]
","
Selects data from a table or multiple tables.
WHERE filters rows before other operations.
GROUP BY groups the result by the given expression(s).
HAVING filters rows after grouping.
If GROUP BY clause is not specified, but non-window aggregate functions are used or HAVING is specified
all rows are grouped together.
Aggregate functions are evaluated, SAMPLE_SIZE limits the number of rows read.
HAVING filters rows after grouping and evaluation of aggregate functions.
Window functions are evaluated.
QUALIFY filters rows after evaluation of window functions.
DISTINCT removes duplicates.
UNION, EXCEPT (MINUS), and INTERSECT combine the result of this query with the results of another query.
Multiple set operators (UNION, INTERSECT, MINUS, EXCEPT) are evaluated from left to right.
For compatibility with other databases and future versions of H2 please use parentheses.
ORDER BY sorts the result by the given column(s) or expression(s).
UNION combines the result of this query with the results of another query.
Number of rows in output can be limited either with standard OFFSET / FETCH,
with non-standard LIMIT / OFFSET, or with non-standard TOP clauses.
Different clauses cannot be used together.
OFFSET specifies how many rows to skip.
Please note that queries with high offset values can be slow.
FETCH FIRST/NEXT, LIMIT or TOP limits the number of rows returned by the query (no limit if null or smaller than zero).
If PERCENT is specified number of rows is specified as a percent of the total number of rows
and should be an integer value between 0 and 100 inclusive.
WITH TIES can be used only together with ORDER BY and means that all additional rows that have the same sorting position
as the last row will be also returned.
OFFSET specifies how many rows to skip.
Please note that queries with high offset values can be slow.
SAMPLE_SIZE limits the number of rows read for aggregate queries.
Multiple set operators (UNION, INTERSECT, MINUS, EXCEPT) are evaluated
from left to right. For compatibility with other databases and future versions
of H2 please use parentheses.
If FOR UPDATE is specified, the tables or rows are locked for writing.
This clause is not allowed in DISTINCT queries and in queries with non-window aggregates, GROUP BY, or HAVING clauses.
...
...
@@ -246,7 +257,7 @@ Selects data from a table.
This command is an equivalent to SELECT * FROM tableName.
FROM, WHERE, GROUP BY, HAVING, and WINDOW clauses from the SELECT command are not allowed.
FROM, WHERE, GROUP BY, HAVING,
QUALIFY,
and WINDOW clauses from the SELECT command are not allowed.
","
TABLE TEST;
TABLE TEST ORDER BY ID FETCH FIRST ROW ONLY;
...
...
h2/src/docsrc/html/advanced.html
浏览文件 @
992f5c7f
...
...
@@ -480,9 +480,9 @@ CURRENT_TIME, CURRENT_TIMESTAMP, CURRENT_USER, DISTINCT, EXCEPT,
EXISTS, FALSE, FETCH, FOR, FOREIGN, FROM, FULL, GROUP, HAVING,
IF, INNER, INTERSECT, INTERSECTS, INTERVAL, IS, JOIN, LIKE,
LIMIT, LOCALTIME, LOCALTIMESTAMP, MINUS, NATURAL, NOT, NULL,
OFFSET, ON, ORDER, PRIMARY,
ROW, ROWNUM, SELECT, SYSDATE
,
SYS
TIME, SYSTIMESTAMP, TABLE, TODAY, TOP, TRUE, UNION, UNIQUE
,
VALUES, WHERE, WINDOW, WITH
OFFSET, ON, ORDER, PRIMARY,
QUALIFY, ROW, ROWNUM, SELECT
,
SYS
DATE, SYSTIME, SYSTIMESTAMP, TABLE, TODAY, TOP, TRUE, UNION
,
UNIQUE,
VALUES, WHERE, WINDOW, WITH
</code>
</p><p>
Certain words of this list are keywords because they are functions that can be used without '()',
...
...
h2/src/docsrc/html/changelog.html
浏览文件 @
992f5c7f
...
...
@@ -21,6 +21,8 @@ Change Log
<h2>
Next Version (unreleased)
</h2>
<ul>
<li>
PR #1711: Add QUALIFY clause to SELECT command
</li>
<li>
Issue #1708: CREATE TABLE AS doesn't support column lists without data types
</li>
<li>
PR #1707: Fix sort order and ENUM data type in external results
...
...
h2/src/main/org/h2/command/Parser.java
浏览文件 @
992f5c7f
...
...
@@ -49,6 +49,7 @@ import static org.h2.util.ParserUtil.OFFSET;
import
static
org
.
h2
.
util
.
ParserUtil
.
ON
;
import
static
org
.
h2
.
util
.
ParserUtil
.
ORDER
;
import
static
org
.
h2
.
util
.
ParserUtil
.
PRIMARY
;
import
static
org
.
h2
.
util
.
ParserUtil
.
QUALIFY
;
import
static
org
.
h2
.
util
.
ParserUtil
.
ROW
;
import
static
org
.
h2
.
util
.
ParserUtil
.
ROWNUM
;
import
static
org
.
h2
.
util
.
ParserUtil
.
SELECT
;
...
...
@@ -507,6 +508,8 @@ public class Parser {
"ORDER"
,
// PRIMARY
"PRIMARY"
,
// QUALIFY
"QUALIFY"
,
// ROW
"ROW"
,
// ROWNUM
...
...
@@ -2748,6 +2751,11 @@ public class Parser {
Expression
condition
=
readExpression
();
command
.
setHaving
(
condition
);
}
if
(
readIf
(
QUALIFY
))
{
command
.
setWindowQuery
();
Expression
condition
=
readExpression
();
command
.
setQualify
(
condition
);
}
if
(
readIf
(
WINDOW
))
{
do
{
int
index
=
parseIndex
;
...
...
h2/src/main/org/h2/command/dml/Select.java
浏览文件 @
992f5c7f
...
...
@@ -77,9 +77,21 @@ public class Select extends Query {
private
final
ArrayList
<
TableFilter
>
filters
=
Utils
.
newSmallArrayList
();
private
final
ArrayList
<
TableFilter
>
topFilters
=
Utils
.
newSmallArrayList
();
private
Expression
having
;
/**
* WHERE condition.
*/
private
Expression
condition
;
/**
* HAVING condition.
*/
private
Expression
having
;
/**
* QUALIFY condition.
*/
private
Expression
qualify
;
/**
* The visible columns (the ones required in the result).
*/
...
...
@@ -112,6 +124,8 @@ public class Select extends Query {
private
int
havingIndex
;
private
int
qualifyIndex
;
private
int
[]
groupByCopies
;
/**
...
...
@@ -413,7 +427,7 @@ public class Select extends Query {
initGroupData
(
columnCount
);
try
{
gatherGroup
(
columnCount
,
DataAnalysisOperation
.
STAGE_WINDOW
);
processGroupResult
(
columnCount
,
result
,
offset
,
quickOffset
);
processGroupResult
(
columnCount
,
result
,
offset
,
quickOffset
,
false
);
}
finally
{
groupData
.
reset
();
}
...
...
@@ -433,7 +447,7 @@ public class Select extends Query {
}
}
groupData
.
done
();
processGroupResult
(
columnCount
,
result
,
offset
,
quickOffset
);
processGroupResult
(
columnCount
,
result
,
offset
,
quickOffset
,
/* Having was performed earlier */
false
);
}
finally
{
isGroupWindowStage2
=
false
;
}
...
...
@@ -446,7 +460,7 @@ public class Select extends Query {
initGroupData
(
columnCount
);
try
{
gatherGroup
(
columnCount
,
DataAnalysisOperation
.
STAGE_GROUP
);
processGroupResult
(
columnCount
,
result
,
offset
,
quickOffset
);
processGroupResult
(
columnCount
,
result
,
offset
,
quickOffset
,
true
);
}
finally
{
groupData
.
reset
();
}
...
...
@@ -498,7 +512,8 @@ public class Select extends Query {
}
}
private
void
processGroupResult
(
int
columnCount
,
LocalResult
result
,
long
offset
,
boolean
quickOffset
)
{
private
void
processGroupResult
(
int
columnCount
,
LocalResult
result
,
long
offset
,
boolean
quickOffset
,
boolean
withHaving
)
{
for
(
ValueRow
currentGroupsKey
;
(
currentGroupsKey
=
groupData
.
next
())
!=
null
;)
{
Value
[]
keyValues
=
currentGroupsKey
.
getList
();
Value
[]
row
=
new
Value
[
columnCount
];
...
...
@@ -519,7 +534,10 @@ public class Select extends Query {
Expression
expr
=
expressions
.
get
(
j
);
row
[
j
]
=
expr
.
getValue
(
session
);
}
if
(
isHavingNullOrFalse
(
row
))
{
if
(
withHaving
&&
isHavingNullOrFalse
(
row
))
{
continue
;
}
if
(
qualifyIndex
>=
0
&&
!
row
[
qualifyIndex
].
getBoolean
())
{
continue
;
}
if
(
quickOffset
&&
offset
>
0
)
{
...
...
@@ -1063,6 +1081,13 @@ public class Select extends Query {
}
else
{
havingIndex
=
-
1
;
}
if
(
qualify
!=
null
)
{
expressions
.
add
(
qualify
);
qualifyIndex
=
expressions
.
size
()
-
1
;
qualify
=
null
;
}
else
{
qualifyIndex
=
-
1
;
}
if
(
withTies
&&
!
hasOrder
())
{
throw
DbException
.
get
(
ErrorCode
.
WITH_TIES_WITHOUT_ORDER_BY
);
...
...
@@ -1130,12 +1155,17 @@ public class Select extends Query {
for
(
TableFilter
f
:
filters
)
{
mapColumns
(
f
,
0
);
}
if
(
havingIndex
>=
0
)
{
Expression
expr
=
expressions
.
get
(
havingIndex
);
mapCondition
(
havingIndex
);
mapCondition
(
qualifyIndex
);
checkInit
=
true
;
}
private
void
mapCondition
(
int
index
)
{
if
(
index
>=
0
)
{
Expression
expr
=
expressions
.
get
(
index
);
SelectListColumnResolver
res
=
new
SelectListColumnResolver
(
this
);
expr
.
mapColumns
(
res
,
0
,
Expression
.
MAP_INITIAL
);
}
checkInit
=
true
;
}
private
int
mergeGroupByExpressions
(
Database
db
,
int
index
,
ArrayList
<
String
>
expressionSQL
,
boolean
scanPrevious
)
...
...
@@ -1221,14 +1251,9 @@ public class Select extends Query {
}
}
}
if
(
isGroupQuery
&&
groupIndex
==
null
&&
havingIndex
<
0
&&
filters
.
size
()
==
1
)
{
if
(
condition
==
null
)
{
Table
t
=
filters
.
get
(
0
).
getTable
();
ExpressionVisitor
optimizable
=
ExpressionVisitor
.
getOptimizableVisitor
(
t
);
isQuickAggregateQuery
=
isEverything
(
optimizable
);
}
if
(
isGroupQuery
&&
groupIndex
==
null
&&
havingIndex
<
0
&&
qualifyIndex
<
0
&&
condition
==
null
&&
filters
.
size
()
==
1
)
{
isQuickAggregateQuery
=
isEverything
(
ExpressionVisitor
.
getOptimizableVisitor
(
filters
.
get
(
0
).
getTable
()));
}
cost
=
preparePlan
(
session
.
isParsingCreateView
());
if
(
distinct
&&
session
.
getDatabase
().
getSettings
().
optimizeDistinct
&&
...
...
@@ -1504,18 +1529,8 @@ public class Select extends Query {
g
.
getUnenclosedSQL
(
buff
.
builder
());
}
}
if
(
having
!=
null
)
{
// could be set in addGlobalCondition
// in this case the query is not run directly, just getPlanSQL is
// called
Expression
h
=
having
;
buff
.
append
(
"\nHAVING "
);
h
.
getUnenclosedSQL
(
buff
.
builder
());
}
else
if
(
havingIndex
>=
0
)
{
Expression
h
=
exprList
[
havingIndex
];
buff
.
append
(
"\nHAVING "
);
h
.
getUnenclosedSQL
(
buff
.
builder
());
}
getFilterSQL
(
buff
,
"\nHAVING "
,
exprList
,
having
,
havingIndex
);
getFilterSQL
(
buff
,
"\nQUALIFY "
,
exprList
,
qualify
,
qualifyIndex
);
if
(
sort
!=
null
)
{
buff
.
append
(
"\nORDER BY "
).
append
(
sort
.
getSQL
(
exprList
,
visibleColumnCount
));
...
...
@@ -1554,6 +1569,17 @@ public class Select extends Query {
return
buff
.
toString
();
}
private
static
void
getFilterSQL
(
StatementBuilder
buff
,
String
sql
,
Expression
[]
exprList
,
Expression
condition
,
int
conditionIndex
)
{
if
(
condition
!=
null
)
{
buff
.
append
(
sql
);
condition
.
getUnenclosedSQL
(
buff
.
builder
());
}
else
if
(
conditionIndex
>=
0
)
{
buff
.
append
(
sql
);
exprList
[
conditionIndex
].
getUnenclosedSQL
(
buff
.
builder
());
}
}
public
void
setHaving
(
Expression
having
)
{
this
.
having
=
having
;
}
...
...
@@ -1562,6 +1588,14 @@ public class Select extends Query {
return
having
;
}
public
void
setQualify
(
Expression
qualify
)
{
this
.
qualify
=
qualify
;
}
public
Expression
getQualify
()
{
return
qualify
;
}
@Override
public
int
getColumnCount
()
{
return
visibleColumnCount
;
...
...
@@ -1696,6 +1730,9 @@ public class Select extends Query {
if
(
having
!=
null
)
{
having
.
updateAggregate
(
s
,
stage
);
}
if
(
qualify
!=
null
)
{
qualify
.
updateAggregate
(
s
,
stage
);
}
}
@Override
...
...
@@ -1747,6 +1784,9 @@ public class Select extends Query {
if
(
having
!=
null
&&
!
having
.
isEverything
(
v2
))
{
return
false
;
}
if
(
qualify
!=
null
&&
!
qualify
.
isEverything
(
v2
))
{
return
false
;
}
return
true
;
}
...
...
h2/src/main/org/h2/jdbc/JdbcDatabaseMetaData.java
浏览文件 @
992f5c7f
...
...
@@ -1542,7 +1542,7 @@ public class JdbcDatabaseMetaData extends TraceObject implements
* table/column/index name, in addition to the SQL-2003 keywords. The list
* returned is:
* <pre>
* INTERSECTS,LIMIT,MINUS,OFFSET,ROWNUM,SYSDATE,SYSTIME,SYSTIMESTAMP,TODAY,TOP
* INTERSECTS,LIMIT,MINUS,OFFSET,
QUALIFY,
ROWNUM,SYSDATE,SYSTIME,SYSTIMESTAMP,TODAY,TOP
* </pre>
* The complete list of keywords (including SQL-2003 keywords) is:
* <pre>
...
...
@@ -1551,9 +1551,9 @@ public class JdbcDatabaseMetaData extends TraceObject implements
* EXISTS, FALSE, FETCH, FOR, FOREIGN, FROM, FULL, GROUP, HAVING,
* IF, INNER, INTERSECT, INTERSECTS, INTERVAL, IS, JOIN, LIKE,
* LIMIT, LOCALTIME, LOCALTIMESTAMP, MINUS, NATURAL, NOT, NULL,
* OFFSET, ON, ORDER, PRIMARY,
ROW, ROWNUM, SELECT, SYSDATE
,
* SYS
TIME, SYSTIMESTAMP, TABLE, TODAY, TOP, TRUE, UNION, UNIQUE
,
* VALUES, WHERE, WINDOW, WITH
* OFFSET, ON, ORDER, PRIMARY,
QUALIFY, ROW, ROWNUM, SELECT
,
* SYS
DATE, SYSTIME, SYSTIMESTAMP, TABLE, TODAY, TOP, TRUE, UNION
,
*
UNIQUE,
VALUES, WHERE, WINDOW, WITH
* </pre>
*
* @return a list of additional the keywords
...
...
@@ -1561,7 +1561,7 @@ public class JdbcDatabaseMetaData extends TraceObject implements
@Override
public
String
getSQLKeywords
()
{
debugCodeCall
(
"getSQLKeywords"
);
return
"IF,INTERSECTS,LIMIT,MINUS,OFFSET,ROWNUM,SYSDATE,SYSTIME,SYSTIMESTAMP,TODAY,TOP"
;
return
"IF,INTERSECTS,LIMIT,MINUS,OFFSET,
QUALIFY,
ROWNUM,SYSDATE,SYSTIME,SYSTIMESTAMP,TODAY,TOP"
;
}
/**
...
...
h2/src/main/org/h2/util/ParserUtil.java
浏览文件 @
992f5c7f
...
...
@@ -217,10 +217,15 @@ public class ParserUtil {
*/
public
static
final
int
PRIMARY
=
ORDER
+
1
;
/**
* The token "QUALIFY".
*/
public
static
final
int
QUALIFY
=
PRIMARY
+
1
;
/**
* The token "ROW".
*/
public
static
final
int
ROW
=
PRIMAR
Y
+
1
;
public
static
final
int
ROW
=
QUALIF
Y
+
1
;
/**
* The token "ROWNUM".
...
...
@@ -476,6 +481,11 @@ public class ParserUtil {
return
PRIMARY
;
}
return
IDENTIFIER
;
case
'Q'
:
if
(
eq
(
"QUALIFY"
,
s
,
ignoreCase
,
start
,
end
))
{
return
QUALIFY
;
}
return
IDENTIFIER
;
case
'R'
:
if
(
eq
(
"ROW"
,
s
,
ignoreCase
,
start
,
end
))
{
return
ROW
;
...
...
h2/src/test/org/h2/test/jdbc/TestMetaData.java
浏览文件 @
992f5c7f
...
...
@@ -463,7 +463,7 @@ public class TestMetaData extends TestDb {
assertEquals
(
"schema"
,
meta
.
getSchemaTerm
());
assertEquals
(
"\\"
,
meta
.
getSearchStringEscape
());
assertEquals
(
"IF,INTERSECTS,LIMIT,MINUS,OFFSET,ROWNUM,SYSDATE,SYSTIME,SYSTIMESTAMP,TODAY,TOP"
,
assertEquals
(
"IF,INTERSECTS,LIMIT,MINUS,OFFSET,
QUALIFY,
ROWNUM,SYSDATE,SYSTIME,SYSTIMESTAMP,TODAY,TOP"
,
meta
.
getSQLKeywords
());
assertTrue
(
meta
.
getURL
().
startsWith
(
"jdbc:h2:"
));
...
...
h2/src/test/org/h2/test/scripts/dml/select.sql
浏览文件 @
992f5c7f
...
...
@@ -609,3 +609,19 @@ SELECT * FROM TEST WHERE (A, B) IN ((1000000000000, 2), (3, 4));
DROP
TABLE
TEST
;
>
ok
CREATE
TABLE
TEST
(
I
)
AS
VALUES
1
,
2
,
3
;
>
ok
SELECT
COUNT
(
*
)
C
FROM
TEST
HAVING
C
<
1
;
>
C
>
-
>
rows
:
0
SELECT
COUNT
(
*
)
C
FROM
TEST
QUALIFY
C
<
1
;
>
C
>
-
>
rows
:
0
DROP
TABLE
TEST
;
>
ok
h2/src/test/org/h2/test/scripts/functions/aggregate/array-agg.sql
浏览文件 @
992f5c7f
...
...
@@ -168,6 +168,22 @@ SELECT ARRAY_AGG(ARRAY_AGG(ID ORDER BY ID)) OVER (PARTITION BY NAME), NAME FROM
>
[[
4
,
5
,
6
]]
c
>
rows
:
3
SELECT
ARRAY_AGG
(
ARRAY_AGG
(
ID
ORDER
BY
ID
))
OVER
(
PARTITION
BY
NAME
),
NAME
FROM
TEST
WHERE
ID
<>
5
GROUP
BY
NAME
HAVING
ARRAY_AGG
(
ID
ORDER
BY
ID
)[
1
]
>
1
QUALIFY
ARRAY_AGG
(
ARRAY_AGG
(
ID
ORDER
BY
ID
))
OVER
(
PARTITION
BY
NAME
)
<>
ARRAY
[
ARRAY
[
3
]];
>
ARRAY_AGG
(
ARRAY_AGG
(
ID
ORDER
BY
ID
))
OVER
(
PARTITION
BY
NAME
)
NAME
>
------------------------------------------------------------- ----
>
[[
4
,
6
]]
c
>
rows
:
1
EXPLAIN
SELECT
ARRAY_AGG
(
ARRAY_AGG
(
ID
ORDER
BY
ID
))
OVER
(
PARTITION
BY
NAME
),
NAME
FROM
TEST
WHERE
ID
<>
5
GROUP
BY
NAME
HAVING
ARRAY_AGG
(
ID
ORDER
BY
ID
)[
1
]
>
1
QUALIFY
ARRAY_AGG
(
ARRAY_AGG
(
ID
ORDER
BY
ID
))
OVER
(
PARTITION
BY
NAME
)
<>
ARRAY
[
ARRAY
[
3
]];
>>
SELECT
ARRAY_AGG
(
ARRAY_AGG
(
ID
ORDER
BY
ID
))
OVER
(
PARTITION
BY
NAME
),
NAME
FROM
PUBLIC
.
TEST
/* PUBLIC.TEST.tableScan */
WHERE
ID
<>
5
GROUP
BY
NAME
HAVING
ARRAY_GET
(
ARRAY_AGG
(
ID
ORDER
BY
ID
),
1
)
>
1
QUALIFY
ARRAY_AGG
(
ARRAY_AGG
(
ID
ORDER
BY
ID
))
OVER
(
PARTITION
BY
NAME
)
<>
ARRAY
[
ARRAY
[
3
]]
SELECT
ARRAY_AGG
(
ARRAY_AGG
(
ID
ORDER
BY
ID
))
OVER
(
PARTITION
BY
NAME
),
NAME
FROM
TEST
GROUP
BY
NAME
ORDER
BY
NAME
OFFSET
1
ROW
;
>
ARRAY_AGG
(
ARRAY_AGG
(
ID
ORDER
BY
ID
))
OVER
(
PARTITION
BY
NAME
)
NAME
...
...
h2/src/test/org/h2/test/scripts/functions/window/row_number.sql
浏览文件 @
992f5c7f
...
...
@@ -83,6 +83,33 @@ SELECT *,
>
9
4
41
1
1
1
0
.
0
1
.
0
>
rows
:
9
SELECT
*
,
ROW_NUMBER
()
OVER
(
PARTITION
BY
CATEGORY
ORDER
BY
ID
)
RN
,
RANK
()
OVER
(
PARTITION
BY
CATEGORY
ORDER
BY
ID
)
RK
,
DENSE_RANK
()
OVER
(
PARTITION
BY
CATEGORY
ORDER
BY
ID
)
DR
,
ROUND
(
PERCENT_RANK
()
OVER
(
PARTITION
BY
CATEGORY
ORDER
BY
ID
),
2
)
PR
,
ROUND
(
CUME_DIST
()
OVER
(
PARTITION
BY
CATEGORY
ORDER
BY
ID
),
2
)
CD
FROM
TEST
QUALIFY
ROW_NUMBER
()
OVER
(
PARTITION
BY
CATEGORY
ORDER
BY
ID
)
=
2
;
>
ID
CATEGORY
VALUE
RN
RK
DR
PR
CD
>
-- -------- ----- -- -- -- --- ----
>
2
1
12
2
2
2
0
.
5
0
.
67
>
5
2
22
2
2
2
1
.
0
1
.
0
>
7
3
32
2
2
2
0
.
5
0
.
67
>
rows
:
3
SELECT
*
,
ROW_NUMBER
()
OVER
(
PARTITION
BY
CATEGORY
ORDER
BY
ID
)
RN
,
RANK
()
OVER
(
PARTITION
BY
CATEGORY
ORDER
BY
ID
)
RK
,
DENSE_RANK
()
OVER
(
PARTITION
BY
CATEGORY
ORDER
BY
ID
)
DR
,
ROUND
(
PERCENT_RANK
()
OVER
(
PARTITION
BY
CATEGORY
ORDER
BY
ID
),
2
)
PR
,
ROUND
(
CUME_DIST
()
OVER
(
PARTITION
BY
CATEGORY
ORDER
BY
ID
),
2
)
CD
FROM
TEST
QUALIFY
RN
=
3
;
>
ID
CATEGORY
VALUE
RN
RK
DR
PR
CD
>
-- -------- ----- -- -- -- --- ---
>
3
1
13
3
3
3
1
.
0
1
.
0
>
8
3
33
3
3
3
1
.
0
1
.
0
>
rows
:
2
SELECT
ROW_NUMBER
()
OVER
(
ORDER
BY
CATEGORY
)
RN
,
RANK
()
OVER
(
ORDER
BY
CATEGORY
)
RK
,
...
...
编写
预览
Markdown
格式
0%
重试
或
添加新文件
添加附件
取消
您添加了
0
人
到此讨论。请谨慎行事。
请先完成此评论的编辑!
取消
请
注册
或者
登录
后发表评论