- ROW_NUMBER(): This one generates a new row number for every row, regardless of duplicates within a partition.
- RANK(): This one generates a new row number for every distinct row, leaving gaps between groups of duplicates within a partition.
- DENSE_RANK(): This one generates a new row number for every distinct row, leaving no gaps between groups of duplicates within a partition.

```
CREATE TABLE t AS
SELECT 'a' v UNION ALL
SELECT 'a' UNION ALL
SELECT 'a' UNION ALL
SELECT 'b' UNION ALL
SELECT 'c' UNION ALL
SELECT 'c' UNION ALL
SELECT 'd' UNION ALL
SELECT 'e'
```

```
SELECT
v,
ROW_NUMBER() OVER (window) row_number,
RANK() OVER (window) rank,
DENSE_RANK() OVER (window) dense_rank
FROM t
WINDOW window AS (ORDER BY v)
ORDER BY v
```

+---+------------+------+------------+ | V | ROW_NUMBER | RANK | DENSE_RANK | +---+------------+------+------------+ | a | 1 | 1 | 1 | | a | 2 | 1 | 1 | | a | 3 | 1 | 1 | | b | 4 | 4 | 2 | | c | 5 | 5 | 3 | | c | 6 | 5 | 3 | | d | 7 | 7 | 4 | | e | 8 | 8 | 5 | +---+------------+------+------------+

^{(See also this SQLFiddle)}

### How DENSE_RANK() can help when writing SELECT DISTINCT

No doubt,`ROW_NUMBER()`

is the most useful ranking function among the above, specifically when you need to emulate LIMIT .. OFFSET clauses as in DB2, Oracle (11g or less), Sybase SQL Anywhere (prior to version 12), SQL Server (2008 and less). Read on here about how jOOQ emulates this SQL clause in various SQL dialects.
But using `ROW_NUMBER()`

has a subtle problem when used along with `DISTINCT`

or `UNION`

. It prevents the database from being able to remove duplicates, because `ROW_NUMBER`

will always produce distinct values within a partition. In the above example, duplicate values for `T.V`

were added on purpose. How can we first remove the duplicates and only then enumerate row numbers? Clearly, we can no longer use `ROW_NUMBER()`

. The following query:
```
SELECT DISTINCT
v,
ROW_NUMBER() OVER (window) row_number
FROM t
WINDOW window AS (ORDER BY v)
ORDER BY v, row_number
```

+---+------------+ | V | ROW_NUMBER | +---+------------+ | a | 1 | | a | 2 | | a | 3 | | b | 4 | | c | 5 | | c | 6 | | d | 7 | | e | 8 | +---+------------+

^{(See also this SQLFiddle)}But we can use

`DENSE_RANK()`

, instead! With `DENSE_RANK()`

the ranking is applied in a way that duplicate records will receive the same rank. And there are no gaps between ranks. Hence:
```
SELECT DISTINCT
v,
DENSE_RANK() OVER (window) row_number
FROM t
WINDOW window AS (ORDER BY v)
ORDER BY v, row_number
```

+---+------------+ | V | ROW_NUMBER | +---+------------+ | a | 1 | | b | 2 | | c | 3 | | d | 4 | | e | 5 | +---+------------+

^{(See also this SQLFiddle)}

### Thus, remember…

Thus, remember:`ROW_NUMBER()`

is to`SELECT`

what`DENSE_RANK()`

is to`SELECT DISTINCT`

### Caveats

In order for the above to be true, however, you must ensure that all expressions from the`SELECT DISTINCT`

clause are used in the `DENSE_RANK()`

‘s `OVER(ORDER BY ...)`

clause. For example:
```
SELECT DISTINCT
v1,
v2,
v3,
DENSE_RANK() OVER (window) row_number
FROM t
WINDOW window AS (ORDER BY v1, v2, v3)
```

`v1, v2, v3`

are other ranking functions or aggregate functions, or non-deterministic expressions, etc., the above trick won’t work. But it’s still a nice trick to keep up one’s sleeves for the odd corner-case query, where distinct rows need row numbers.
And if you don’t want to think about this trick all the time, do note that we’ve baked this into jOOQ, because we think that you should worry about business logic, not about SQL standardiation:
I have a input like

i need to pick the most top 3 played movie , i.e,

please help

That’s not really a query where you need window functions. It can be solved easily with SQL-92 aggregate functions, which run on any database:

Note that a great platform to ask these questions is Stack Overflow

select top 3 count(movie_name) total, movie_name from movies

group by movie_name

order by count(movie_name) desc;

with cteMovie as

(

select movie_name, played_time, row_number() over (partition by movie_name order by played_time desc ) as number from my_table

)

select * from cteMovie where number=1

Thank you for your contribution. Yes, that is how you can filter by such a row number. In what context did you think this fits the article at hand?

How to use the “where” function for row_number

for example:

row-number > 3

You’ll have to calculate the window function value in a derived table:

can any one please post the query to find the 10th highest record in HIVE.

Sounds like an excellent question for Stack Overflow! http://stackoverflow.com

We are sybase database, dont have DENSE_RANK, PARTITION BY functionality. How to achieve the DENSE_RANK() OVER (PARTITION BY with out using In-built functions.

You migrate off Sybase ;) Cheers

On a more serious note: DENSE_RANK() can be emulated with an ordinary subquery.

is the same as:

Chances are, that this query is quite slow, though.

i want to generate rank order by 2 columns.Is there is any solution for that

Just

`RANK() OVER (ORDER BY col1, col2)`