| 1 |
SQL Language Extension: Improved aggregate handling |
| 2 |
|
| 3 |
Syntax |
| 4 |
======== |
| 5 |
|
| 6 |
|
| 7 |
GROUP BY |
| 8 |
-------- |
| 9 |
SELECT ... FROM .... [GROUP BY group_by_list] |
| 10 |
|
| 11 |
group_by_list : group_by_item [, group_by_list]; |
| 12 |
|
| 13 |
group_by_item : column_name |
| 14 |
| ordinal |
| 15 |
| udf |
| 16 |
| group_by_function; |
| 17 |
|
| 18 |
group_by_function : numeric_value_function |
| 19 |
| string_value_function |
| 20 |
| case_expression |
| 21 |
; |
| 22 |
|
| 23 |
numeric_value_function : EXTRACT '(' timestamp_part FROM value ')'; |
| 24 |
|
| 25 |
string_value_function : SUBSTRING '(' value FROM pos_short_integer ')' |
| 26 |
| SUBSTRING '(' value FROM pos_short_integer FOR nonneg_short_integer ')' |
| 27 |
| KW_UPPER '(' value ')' |
| 28 |
; |
| 29 |
|
| 30 |
(FB1.0 only allowed GROUP BY column_name and udf) |
| 31 |
Ordinal references to a n-th select-item from the select-list (same as ORDER BY). |
| 32 |
The group_by_item cannot reference to any aggregate-function (also not burried |
| 33 |
inside a expression) from the same context. |
| 34 |
|
| 35 |
|
| 36 |
HAVING |
| 37 |
-------- |
| 38 |
The having clause only allows aggregate functions or expressions |
| 39 |
that are part of the GROUP BY clause. Previously it was allowed |
| 40 |
to use columns that were not part of the GROUP BY clause. |
| 41 |
|
| 42 |
|
| 43 |
ORDER BY |
| 44 |
-------- |
| 45 |
When the context is an aggregate statement then the ORDER BY |
| 46 |
clause only allows valid expressions. That are aggregate |
| 47 |
functions or expression part of the GROUP BY clause. |
| 48 |
Previously it was allowed to use non-valid expressions. |
| 49 |
|
| 50 |
|
| 51 |
Aggregate functions inside sub-selects |
| 52 |
-------- |
| 53 |
It is now possible to use a aggregate function or expression |
| 54 |
contained in the GROUP BY clause inside a sub-select. |
| 55 |
(See Examples A) |
| 56 |
|
| 57 |
|
| 58 |
Mixing aggregate functions from different contexts |
| 59 |
-------- |
| 60 |
You can use aggregate functions from different contexts inside |
| 61 |
a expression. |
| 62 |
(See Example B) |
| 63 |
|
| 64 |
|
| 65 |
Sub-selects are supported inside a aggregate function |
| 66 |
-------- |
| 67 |
Using a singleton select expression inside a aggregate function is |
| 68 |
supported. |
| 69 |
(See Example C) |
| 70 |
|
| 71 |
|
| 72 |
Nested aggregate functions |
| 73 |
-------- |
| 74 |
Using a aggregate function inside a other aggregate function is |
| 75 |
possible if the aggregate function inside is from a lower context. |
| 76 |
(See Example C) |
| 77 |
|
| 78 |
|
| 79 |
Author: |
| 80 |
Arno Brinkman <[email protected]> |
| 81 |
|
| 82 |
|
| 83 |
N O T E S |
| 84 |
========= |
| 85 |
|
| 86 |
- Not all expressions are currently alowed inside the GROUP BY list. |
| 87 |
(concatenation for example) |
| 88 |
- ORDER BY clause only accepts valid expressions (this was not for FB1.0) |
| 89 |
- HAVING clause only accepts valid expressions (this was not for FB1.0) |
| 90 |
- Using ordinal 'copies' the expression from the select list as does the |
| 91 |
order by clause. This means when a ordinal references to a sub-select |
| 92 |
the sub-select is at least executed twice. |
| 93 |
|
| 94 |
|
| 95 |
Examples |
| 96 |
======== |
| 97 |
|
| 98 |
|
| 99 |
A) |
| 100 |
SELECT |
| 101 |
r.RDB$RELATION_NAME, |
| 102 |
MAX(r.RDB$FIELD_POSITION), |
| 103 |
(SELECT |
| 104 |
r2.RDB$FIELD_NAME |
| 105 |
FROM |
| 106 |
RDB$RELATION_FIELDS r2 |
| 107 |
WHERE |
| 108 |
r2.RDB$RELATION_NAME = r.RDB$RELATION_NAME and |
| 109 |
r2.RDB$FIELD_POSITION = MAX(r.RDB$FIELD_POSITION)) |
| 110 |
FROM |
| 111 |
RDB$RELATION_FIELDS r |
| 112 |
GROUP BY |
| 113 |
1 |
| 114 |
|
| 115 |
SELECT |
| 116 |
rf.RDB$RELATION_NAME AS "Relationname", |
| 117 |
(SELECT |
| 118 |
r.RDB$RELATION_ID |
| 119 |
FROM |
| 120 |
RDB$RELATIONS r |
| 121 |
WHERE |
| 122 |
r.RDB$RELATION_NAME = rf.RDB$RELATION_NAME) AS "ID", |
| 123 |
COUNT(*) AS "Fields" |
| 124 |
FROM |
| 125 |
RDB$RELATION_FIELDS rf |
| 126 |
GROUP BY |
| 127 |
rf.RDB$RELATION_NAME |
| 128 |
|
| 129 |
|
| 130 |
B) |
| 131 |
SELECT |
| 132 |
r.RDB$RELATION_NAME, |
| 133 |
MAX(i.RDB$STATISTICS) AS "Max1", |
| 134 |
(SELECT |
| 135 |
COUNT(*) || ' - ' || MAX(i.RDB$STATISTICS) |
| 136 |
FROM |
| 137 |
RDB$RELATION_FIELDS rf |
| 138 |
WHERE |
| 139 |
rf.RDB$RELATION_NAME = r.RDB$RELATION_NAME) AS "Max2" |
| 140 |
FROM |
| 141 |
RDB$RELATIONS r |
| 142 |
JOIN RDB$INDICES i on (i.RDB$RELATION_NAME = r.RDB$RELATION_NAME) |
| 143 |
GROUP BY |
| 144 |
r.RDB$RELATION_NAME |
| 145 |
HAVING |
| 146 |
MIN(i.RDB$STATISTICS) <> MAX(i.RDB$STATISTICS) |
| 147 |
|
| 148 |
Note! This query gives results in FB1.0, but they are WRONG! |
| 149 |
|
| 150 |
|
| 151 |
C) |
| 152 |
SELECT |
| 153 |
r.RDB$RELATION_NAME, |
| 154 |
SUM((SELECT |
| 155 |
COUNT(*) |
| 156 |
FROM |
| 157 |
RDB$RELATION_FIELDS rf |
| 158 |
WHERE |
| 159 |
rf.RDB$RELATION_NAME = r.RDB$RELATION_NAME)) |
| 160 |
FROM |
| 161 |
RDB$RELATIONS r |
| 162 |
JOIN RDB$INDICES i on (i.RDB$RELATION_NAME = r.RDB$RELATION_NAME) |
| 163 |
GROUP BY |
| 164 |
r.RDB$RELATION_NAME |
| 165 |
|
| 166 |
|