Window functions offer yet another way to implement pagination in SQL. This is a flexible, and above all, standards-compliant method. However, only SQL Server, the Oracle database and PostgreSQL 15+ can use them for a pipelined top-N query. MySQL, MariaDB0and Db2 (LUW) do not abort the index scan after fetching enough rows and therefore execute these queries very inefficiently.
The following example uses the window function ROW_NUMBER for a pagination query:
SELECT *
FROM ( SELECT sales.*
, ROW_NUMBER() OVER (ORDER BY sale_date DESC
, sale_id DESC) rn
FROM sales
) tmp
WHERE rn between 11 and 20
ORDER BY sale_date DESC, sale_id DESCThe ROW_NUMBER function enumerates the rows according to the sort order defined in the over clause. The outer where clause uses this enumeration to limit the result to the second page (rows 11 through 20).
If you like this page, you might also like …
… to subscribe my mailing lists, get free stickers, buy my book or join a training.
The Oracle database recognizes the abort condition and uses the index on SALE_DATE and SALE_ID to produce a pipelined top-N behavior:
- Db2 (LUW)
Explain Plan ------------------------------------------------------------- ID | Operation | Rows | Cost 1 | RETURN | | 65658 2 | FILTER | 100933 of 1009326 | 65658 3 | FETCH SALES | 1009326 of 1009326 | 65295 4 | IXSCAN (REVERSE) SL_DTID | 1009326 of 1009326 | 5679 Predicate Information 2 - RESID (11 <= Q3.$C8) RESID (Q3.$C8 <= 20)Please note that Db2 (LUW) 10.5 does not execute this query as top-n query. Although it prevents the sort operation it still reads through the entire index—it does not abort execution after fetching 20 rows.
To get a proper top-n abort after reading 20 rows, you must double-wrap the query to first apply the top-n abort condition and then filtering away the first 10 rows:
SELECT * FROM (SELECT * FROM (SELECT sales.* , ROW_NUMBER() OVER (ORDER BY sale_date DESC , sale_id DESC) rn FROM sales ) tmp WHERE rn <= 20 ) tmp2 WHERE rn > 10 ORDER BY sale_date DESC, sale_id DESC;Explain Plan ----------------------------------------------------------------- ID | Operation | Rows | Cost 1 | RETURN | | 21 2 | FILTER | 7 of 20 | 21 3 | FETCH SALES | 20 of 1009326 | 65352 4 | IXSCAN (REVERSE) SALES_DT_ID | 1009326 of 1009326 | 5736 Predicate Information 2 - RESID (10 < Q3.$C8)Note that the filter
rn <= 20does not appear in the Predicate Information section, yet the row count estimate reflects it.- Oracle
--------------------------------------------------------------- |Id | Operation | Name | Rows | Cost | --------------------------------------------------------------- | 0 | SELECT STATEMENT | | 1004K| 36877 | |*1 | VIEW | | 1004K| 36877 | |*2 | WINDOW NOSORT STOPKEY | | 1004K| 36877 | | 3 | TABLE ACCESS BY INDEX ROWID | SALES | 1004K| 36877 | | 4 | INDEX FULL SCAN DESCENDING | SL_DTID | 1004K| 2955 | --------------------------------------------------------------- Predicate Information (identified by operation id): --------------------------------------------------- 1 - filter("RN">=11 AND "RN"<=20) 2 - filter(ROW_NUMBER() OVER ( ORDER BY "SALE_DATE" DESC, "SALE_ID" DESC )<=20)- PostgreSQL
Since version 15 the execution plan shows the “Run Condition”, which may abort the downstream execution.
QUERY PLAN ------------------------------------------------------- Subquery Scan on tmp (cost=0.42..141724.98 rows=334751 width=249) (actual time=0.040..0.052 rows=10 loops=1) Filter: (tmp.rn >= 11) Rows Removed by Filter: 10 Buffers: shared hit=5 -> WindowAgg (cost=0.42..129171.80 rows=1004254 width=249) (actual time=0.028..0.049 rows=20 loops=1) Run Condition: (row_number() OVER (?) <= 20) Buffers: shared hit=5 -> Index Scan Backward using sl_dtid on sales (cost=0.42..111597.36 rows=1004254 width=241) (actual time=0.018..0.025 rows=22 loops=1) Buffers: shared hit=5
The WINDOW NOSORT STOPKEY operation indicates that there is no sort operation (NOSORT) and that the database aborts the execution when reaching the upper threshold (STOPKEY). Considering that the aborted operations are executed in a pipelined manner, it means that this query is as efficient as the offset method explained in the previous section.
Support of this optimization is by no means common among SQL products.
- Only with
row_number() - Not with
partition byclause (see below)
While this optimization could conceptually work with any monotonic function, there is a clear focus on the ROW_NUMBER function among the analyzed implementations.
- Only with
whereclause:OVER(ORDER BY…)+WHERE x=?
Caution needs to be taken when using partition by: Even if the where clause limits the rows to a single partition, the pure presence of partition by disables this optimization in some products. This might happen if the partitioned window function is part of a view, but the outer query limits the view to a single partition.
- Not all functions that work in presence of a
whereclause
The strength of window functions is not pagination, however, but analytical calculations. If you have never used window functions before, you should definitely spend a few hours studying the respective documentation.
Links
Oracle: Analytic Functions in 19
PostgreSQL: Window Functions
SQL Server: OVER Clause in SQL Server
