[window function] add percent_rank window function

[jimexist]

Which issue does this PR close?

[window function] add percent_rank function

Closes #667
Related #1076

Rationale for this change

What changes are included in this PR?

Are there any user-facing changes?

[alamb]

Thanks @jimexist ! I updated the title of this PR so that it reflects this PR implementing dense_rank as well as percent_rank

I think the code looks (very) good -- though it might be beneficial to have a unit test for dense_rank.

I also spot checked the answers using postgres:

Postgres:

alamb=# select * from arr;
 x | y
---+----
 1 | 10
 1 | 11
 1 |  3
 1 |  6
 2 | 20
 1 |  3
(6 rows)

alamb=# select x, percent_rank() over (partition by x order by y) from arr;
 x | percent_rank
---+--------------
 1 |            0
 1 |            0
 1 |          0.5
 1 |         0.75
 1 |            1
 2 |            0
(6 rows)

alamb=# select x, dense_rank() over (partition by x order by y) from arr;
 x | dense_rank 
---+------------
 1 |          1
 1 |          1
 1 |          2
 1 |          3
 1 |          4
 2 |          1
(6 rows)

DataFusion

echo "1,10" > /tmp/arr.csv
echo "1,11" >> /tmp/arr.csv
echo "1,3" >> /tmp/arr.csv
echo "1,6" >> /tmp/arr.csv
echo "2,20" >> /tmp/arr.csv
echo "1,3" >> /tmp/arr.csv

cargo run -p datafusion-cli

> CREATE EXTERNAL TABLE arr(x int, y int)
STORED AS CSV
LOCATION '/tmp/arr.csv';
0 rows in set. Query took 0.000 seconds.
>
>
> select * from arr;
+---+----+
| x | y  |
+---+----+
| 1 | 10 |
| 1 | 11 |
| 1 | 3  |
| 1 | 6  |
| 2 | 20 |
| 1 | 3  |
+---+----+
6 rows in set. Query took 0.013 seconds.
> select x, percent_rank() over (partition by x order by y) from arr;
+---+----------------+
| x | PERCENT_RANK() |
+---+----------------+
| 1 | 0              |
| 1 | 0              |
| 1 | 0.5            |
| 1 | 0.75           |
| 1 | 1              |
| 2 | 0              |
+---+----------------+

> select x, dense_rank() over (partition by x order by y) from arr;
+---+--------------+
| x | DENSE_RANK() |
+---+--------------+
| 1 | 1            |
| 1 | 1            |
| 1 | 2            |
| 1 | 3            |
| 1 | 4            |
| 2 | 1            |
+---+--------------+

👍

[alamb]

I think we should also add a test for dense_rank() here as well

[jimexist]

@alamb in fact dense_rank function was already implemented, this PR only added percent rank and thus i needed to change the dense: bool to a three element enum. same goes to unit test, as the code originally covered both rank and dense_rank.

[Dandandan]

Thanks @jimexist !

[alamb]

@alamb in fact dense_rank function was already implemented, this PR only added percent rank and thus i needed to change the dense: bool to a three element enum. same goes to unit test, as the code originally covered both rank and dense_rank.

I have (unbroken) the title of this PR now -- thank you for the clarification. Sorry for the mistake