Oracle's SQL select should have a syntax like this

Question

Oracles SQL is very powerful but lacks a basic thing the LIMIT and OFFSET syntax which
is present in postgres sql, I recommend oracle to support this syntax in future release.
more details of this syntax can be found at :
http://www.postgresql.org/idocs/index.php?sql-select.html
SELECT [ ALL | DISTINCT [ ON ( expression [, ...] ) ] ]
* | expression [ AS output_name ] [, ...]
[ FROM from_item [, ...] ]
[ WHERE condition ]
[ GROUP BY expression [, ...] ]
[ HAVING condition [, ...] ]
[ { UNION | INTERSECT | EXCEPT } [ ALL ] select ]
[ ORDER BY expression [ ASC | DESC | USING operator ] [, ...] ]
[ FOR UPDATE [ OF tablename [, ...] ] ]
[ LIMIT { count | ALL } ]
[ OFFSET start ]

Answer 1

I've executed the above queries against my table and the results follows :
The structure of the table
Name Null? Type
FATWAID NOT NULL NUMBER(11)
FATWASTATUS NUMBER(1)
SUBJ_NO NUMBER(4)
LANG CHAR(1)
SCHOLAR NUMBER(2)
REVIEWER NUMBER(2)
AUDITOR NUMBER(2)
PUBLISHER NUMBER(2)
SUBSCRIBER NUMBER(2)
ENTRY NUMBER(2)
FATWATITLE VARCHAR2(100)
FATWATEXT CLOB
FATWADATE DATE
TRANSLATE CHAR(1)
SELECTED NUMBER(1)
SUBJ_MAIN NUMBER(4)
I have set timing on and executed each query above, following are the statistics.
First the number of rows in the table.
SQL>set timing on
SQL>select count (*) from fatwa;
COUNT(*)
1179651
Elapsed: 00:00:13.05
The query of Andrew Clarke
SQL>SELECT fatwaid FROM (SELECT fatwaid, rownum as ranking FROM fatwa) r
WHERE r.ranking BETWEEN &OFFSET AND &LIMIT
SQL>/
Enter value for offset: 100000
Enter value for limit: 100009
old 2: WHERE r.ranking BETWEEN &OFFSET AND &LIMIT
new 2: WHERE r.ranking BETWEEN 100000 AND 100009
FATWAID
96592
96593
96594
96595
96596
96597
96598
96599
96600
96601
10 rows selected.
Elapsed: 00:00:12.02
SQL> /
Enter value for offset: 1000000
Enter value for limit: 1000009
old 2: WHERE r.ranking BETWEEN &OFFSET AND &LIMIT
new 2: WHERE r.ranking BETWEEN 1000000 AND 1000009
FATWAID
994621
994622
994623
994624
994625
994626
995769
995770
995771
995772
10 rows selected.
Elapsed: 00:00:12.00
The response time is decreasing because of use of bind variables,
but 12 seconds is a sign of poor performance.
Now a slight modification to Clarke's query
I will add order by clause
SQL> ed
Wrote file afiedt.buf
1 SELECT fatwaid FROM (SELECT fatwaid, rownum as ranking FROM fatwa ORDER BY fatwaid) r
2* WHERE r.ranking BETWEEN &OFFSET AND &LIMIT
SQL> /
Enter value for offset: 100001
Enter value for limit: 100010
old 2: WHERE r.ranking BETWEEN &OFFSET AND &LIMIT
new 2: WHERE r.ranking BETWEEN 100001 AND 100010
FATWAID
100032
100033
100034
100035
100036
100037
100038
100039
100040
100041
10 rows selected.
Elapsed: 00:00:04.00 -- time reduced from 12 to 4 seconds
A time of 4 seconds is acceptable but not good,
response time should be in milli seconds.
SQL> /
Enter value for offset: 1000001
Enter value for limit: 1000010
old 2: WHERE r.ranking BETWEEN &OFFSET AND &LIMIT
new 2: WHERE r.ranking BETWEEN 1000001 AND 1000010
FATWAID
1000032
1000033
1000034
1000035
1000036
1000037
1000038
1000039
1000040
1000041
10 rows selected.
Elapsed: 00:00:03.09 -- this reduction is because of bind variables
The query of Chris Gates
SQL>select fatwaid
2 from ( select a.*, rownum r
3 from ( select *
4 from fatwa
5 --where x = :host_variable
6 order by fatwaid ) a
7 where rownum < &HigerBound )
8 where r > &LowerBound
SQL> /
Enter value for higerbound: 100011
old 7: where rownum < &HigerBound )
new 7: where rownum < 100011 )
Enter value for lowerbound: 100000
old 8: where r > &LowerBound
new 8: where r > 100000
FATWAID
100032
100033
100034
100035
100036
100037
100038
100039
100040
100041
10 rows selected.
Elapsed: 00:00:02.04
This seems to be fast
SQL> /
Enter value for higerbound: 1000011
old 7: where rownum < &HigerBound )
new 7: where rownum < 1000011 )
Enter value for lowerbound: 1000000
old 8: where r > &LowerBound
new 8: where r > 1000000
FATWAID
1000032
1000033
1000034
1000035
1000036
1000037
1000038
1000039
1000040
1000041
10 rows selected.
Elapsed: 00:01:14.02
but this is worst when upper bound is 1 million.
Finally Myers query
SQL> select fatwaid from
2 (select /*+ INDEX(fawtaid pk_fatwa) */ fatwaid, rownum x from fatwa
3 where rownum < &UpperBound )
4 where x > &LowerBound;
Enter value for upperbound: 100011
old 3: where rownum < &UpperBound )
new 3: where rownum < 100011 )
Enter value for lowerbound: 100000
old 4: where x > &LowerBound
new 4: where x > 100000
FATWAID
122418
122419
122420
122421
122422
122423
122424
122425
122426
122427
10 rows selected.
Elapsed: 00:00:00.03 -- too fast
SQL> /
Enter value for upperbound: 1000011
old 3: where rownum < &UpperBound )
new 3: where rownum < 1000011 )
Enter value for lowerbound: 1000000
old 4: where x > &LowerBound
new 4: where x > 1000000
FATWAID
984211
984212
984213
984214
984215
984216
984217
984218
984219
984220
10 rows selected.
Elapsed: 00:00:02.02 -- with 1 million rows also satisfactory but it is not is milliseconds
The same query after using order by clause
SQL> select fatwaid from
2 (select /*+ INDEX(fawtaid pk_fatwa) */ fatwaid, rownum x from fatwa
3 where rownum < &UpperBound ORDER BY fatwaid)
4 where x > &LowerBound;
Enter value for upperbound: 100011
old 3: where rownum < &UpperBound ORDER BY fatwaid)
new 3: where rownum < 100011 ORDER BY fatwaid)
Enter value for lowerbound: 100000
old 4: where x > &LowerBound
new 4: where x > 100000
FATWAID
100032
100033
100034
100035
100036
100037
100038
100039
100040
100041
10 rows selected.
Elapsed: 00:00:00.06
SQL> /
Enter value for upperbound: 1000011
old 3: where rownum < &UpperBound ORDER BY fatwaid)
new 3: where rownum < 1000011 ORDER BY fatwaid)
Enter value for lowerbound: 1000000
old 4: where x > &LowerBound
new 4: where x > 1000000
FATWAID
1000032
1000033
1000034
1000035
1000036
1000037
1000038
1000039
1000040
1000041
10 rows selected.
Elapsed: 00:00:07.03 -- slow
SQL> /
Enter value for upperbound: 1000011
old 3: where rownum < &UpperBound ORDER BY fatwaid)
new 3: where rownum < 1000011 ORDER BY fatwaid)
Enter value for lowerbound: 1000000
old 4: where x > &LowerBound
new 4: where x > 1000000
FATWAID
1000032
1000033
1000034
1000035
1000036
1000037
1000038
1000039
1000040
1000041
10 rows selected.
Elapsed: 00:00:00.06
when I execute the same query again it is bringing records from
the SGA so it is very fast
Now which one to choose from ?
Andrew and Myers queries are good and currently I am using
Myers query.
There should be some technique to do this in the most efficient way.
any input is appreciated.