Order by, cursor, rownum

Question

i want to select 10 database entires
from an 8i database with the selector variable and sort the entries by a date
column.
but the problem is that my sql statement
(see below) selects first the entries
and then the entries get ordered.
what i need is first the entries ordered
and then the select. otherwise it is possible
that i get on the next page with the next
10 entries that this entries may be with a date that should be on the prev. page.
has anyone an idea?
thx
SELECT *
FROM (SELECT col1,
col2
FROM table
WHERE ROWNUM < (selector + 10)
MINUS
SELECT col1,
col2
FROM table
WHERE ROWNUM < selector)
ORDER BY TO_DATE (col2, 'DD.MM.YYYY HH24:MI');

Answer 1

In Oracle 8i, the general format for this sort of "top-n" analysis is:
SELECT column1, column2, ROWNUM
FROM (SELECT column1, column2
FROM table_name
ORDER BY top_n_column DESC)
WHERE ROWNUM <= number_of_rows;
Applying that to this situation, the query would be something like:
SELECT col1, col2, ROWNUM
FROM (SELECT col1, col2
FROM table
ORDER BY TO_DATE(col2, 'DD.MM.YYYY HH24:MI') DESC)
WHERE ROWNUM <= 10;
When you put the ORDER BY clause in a sub-query as above, the ORDER BY is applied before the numbering of the rows, so the pseudo-column ROWNUM has some meaning and usefulness.
However, for those of us who are still using versions prior to Oracle 8i, we can't have an ORDER BY clause in a sub-query; It will generate a somewhat confusing error about a missing parenthesis. Therefore, the pseudo-column ROWNUM has no meaning or usefulness. So, prior to Oracle 8i, for "top-n" analysis, the general format is something like this (It will also still work in 8i):
SELECT outer.column1, outer.column2
FROM table_name outer
WHERE number_of_rows >=
(SELECT COUNT (*) + 1
FROM table_name inner
WHERE inner.top_n_column > outer.top_n_column)
ORDER BY outer.top_n_column DESC;
Applying that to this situation, the query would be something like:
SELECT outer.col1, outer.col2
FROM table outer
WHERE 10 >=
(SELECT COUNT (*) + 1
FROM table inner
WHERE TO_DATE(inner.col2,'DD.MM.YYYY HH24:MI') > TO_DATE(outer..col2,'DD.MM.YYYY HH24:MI'))
ORDER BY TO_DATE(outer.col2,'DD.MM.YYYY HH24:MI') DESC;
A lot of people keep asking for queries to select the first few or last few rows entered in a table, without ordering by a specific column (for what purpose I am not sure). It is a common mistake to attempt to use ROWNUM for this purpose. However, ROWID is assigned an incremental value upon input, not upon ordering or retrieval, like ROWNUM. Therefore, the pseudo-column ROWID can be used as a comparison column for the same sort of "top-n" analysis like this:
For the first n rows:
SELECT outer.column1, outer.column2, ROWID
FROM table_name outer
WHERE n >=
(SELECT COUNT (*) + 1
FROM table_name inner
WHERE inner.ROWID < outer.ROWID)
ORDER BY ROWID;
For the last n rows:
SELECT outer.column1, outer.column2, ROWID
FROM table_name outer
WHERE n >=
(SELECT COUNT (*) + 1
FROM table_name inner
WHERE inner.ROWID > outer.ROWID)
ORDER BY ROWID DESC;
If you are only attempting to retrieve one row, with or without a comparison column, then you can use things like MIN, MAX, NOT IN, NOT EXISTS, and set operators to find that one row.
I have seen a lot of confusion on this topic and I hope this explanation was thorough enough to help clarify it for some.
null