Rechercher dans le manuel MySQL

12.21.3 Window Function Frame Specification

Contents

The definition of a window used with a window function can include a frame clause. A frame is a subset of the current partition and the frame clause specifies how to define the subset.

Frames are determined with respect to the current row, which enables a frame to move within a partition depending on the location of the current row within its partition. Examples:

  • By defining a frame to be all rows from the partition start to the current row, you can compute running totals for each row.

  • By defining a frame as extending N rows on either side of the current row, you can compute rolling averages.

The following query demonstrates the use of moving frames to compute running totals within each group of time-ordered level values, as well as rolling averages computed from the current row and the rows that immediately precede and follow it:

  1. mysql> SELECT
  2.          time, subject, val,
  3.          SUM(val) OVER (PARTITION BY subject ORDER BY time
  4.                         ROWS UNBOUNDED PRECEDING)
  5.            AS running_total,
  6.          AVG(val) OVER (PARTITION BY subject ORDER BY time
  7.                         ROWS BETWEEN 1 PRECEDING AND 1 FOLLOWING)
  8.            AS running_average
  9.        FROM observations;
  10. +----------+---------+------+---------------+-----------------+
  11. | time     | subject | val  | running_total | running_average |
  12. +----------+---------+------+---------------+-----------------+
  13. | 07:00:00 | st113   |   10 |            10 |          9.5000 |
  14. | 07:15:00 | st113   |    9 |            19 |         14.6667 |
  15. | 07:30:00 | st113   |   25 |            44 |         18.0000 |
  16. | 07:45:00 | st113   |   20 |            64 |         22.5000 |
  17. | 07:00:00 | xh458   |    0 |             0 |          5.0000 |
  18. | 07:15:00 | xh458   |   10 |            10 |          5.0000 |
  19. | 07:30:00 | xh458   |    5 |            15 |         15.0000 |
  20. | 07:45:00 | xh458   |   30 |            45 |         20.0000 |
  21. | 08:00:00 | xh458   |   25 |            70 |         27.5000 |
  22. +----------+---------+------+---------------+-----------------+

For the running_average column, there is no frame row preceding the first one or following the last. In these cases, AVG() computes the average of the rows that are available.

Aggregate functions used as window functions operate on rows in the current row frame, as do these nonaggregate window functions:

  1. FIRST_VALUE()
  2. LAST_VALUE()
  3. NTH_VALUE()

Standard SQL specifies that window functions that operate on the entire partition should have no frame clause. MySQL permits a frame clause for such functions but ignores it. These functions use the entire partition even if a frame is specified:

  1. CUME_DIST()
  2. DENSE_RANK()
  3. LAG()
  4. LEAD()
  5. NTILE()
  6. PERCENT_RANK()
  7. RANK()
  8. ROW_NUMBER()

The frame clause, if given, has this syntax:

  1. frame_clause:
  2.     frame_units frame_extent
  3.  
  4. frame_units:
  5.     {ROWS | RANGE}

In the absence of a frame clause, the default frame depends on whether an ORDER BY clause is present, as described later in this section.

The frame_units value indicates the type of relationship between the current row and frame rows:

  • ROWS: The frame is defined by beginning and ending row positions. Offsets are differences in row numbers from the current row number.

  • RANGE: The frame is defined by rows within a value range. Offsets are differences in row values from the current row value.

The frame_extent value indicates the start and end points of the frame. You can specify just the start of the frame (in which case the current row is implicitly the end) or use BETWEEN to specify both frame endpoints:

  1. frame_extent:
  2.     {frame_start | frame_between}
  3.  
  4. frame_between:
  5.     BETWEEN frame_start AND frame_end
  6.  
  7. frame_start, frame_end: {
  8.     CURRENT ROW
  9.   | UNBOUNDED PRECEDING
  10.   | UNBOUNDED FOLLOWING
  11.   | expr PRECEDING
  12.   | expr FOLLOWING
  13. }

With BETWEEN syntax, frame_start must not occur later than frame_end.

The permitted frame_start and frame_end values have these meanings:

  • CURRENT ROW: For ROWS, the bound is the current row. For RANGE, the bound is the peers of the current row.

  • UNBOUNDED PRECEDING: The bound is the first partition row.

  • UNBOUNDED FOLLOWING: The bound is the last partition row.

  • expr PRECEDING: For ROWS, the bound is expr rows before the current row. For RANGE, the bound is the rows with values equal to the current row value minus expr; if the current row value is NULL, the bound is the peers of the row.

    For expr PRECEDING (and expr FOLLOWING), expr can be a ? parameter marker (for use in a prepared statement), a nonnegative numeric literal, or a temporal interval of the form INTERVAL val unit. For INTERVAL expressions, val specifies nonnegative interval value, and unit is a keyword indicating the units in which the value should be interpreted. (For details about the permitted units specifiers, see the description of the DATE_ADD() function in Section 12.7, “Date and Time Functions”.)

    RANGE on a numeric or temporal expr requires ORDER BY on a numeric or temporal expression, respectively.

    Examples of valid expr PRECEDING and expr FOLLOWING indicators:

    1. 10 PRECEDING
    2. INTERVAL 5 DAY PRECEDING
    3. 5 FOLLOWING
    4. INTERVAL '2:30' MINUTE_SECOND FOLLOWING

  • expr FOLLOWING: For ROWS, the bound is expr rows after the current row. For RANGE, the bound is the rows with values equal to the current row value plus expr; if the current row value is NULL, the bound is the peers of the row.

    For permitted values of expr, see the description of expr PRECEDING.

The following query demonstrates FIRST_VALUE(), LAST_VALUE(), and two instances of NTH_VALUE():

  1. mysql> SELECT
  2.          time, subject, val,
  3.          FIRST_VALUE(val)  OVER w AS 'first',
  4.          LAST_VALUE(val)   OVER w AS 'last',
  5.          NTH_VALUE(val, 2) OVER w AS 'second',
  6.          NTH_VALUE(val, 4) OVER w AS 'fourth'
  7.        FROM observations
  8.        WINDOW w AS (PARTITION BY subject ORDER BY time
  9.                     ROWS UNBOUNDED PRECEDING);
  10. +----------+---------+------+-------+------+--------+--------+
  11. | time     | subject | val  | first | last | second | fourth |
  12. +----------+---------+------+-------+------+--------+--------+
  13. | 07:00:00 | st113   |   10 |    10 |   10 |   NULL |   NULL |
  14. | 07:15:00 | st113   |    9 |    10 |    9 |      9 |   NULL |
  15. | 07:30:00 | st113   |   25 |    10 |   25 |      9 |   NULL |
  16. | 07:45:00 | st113   |   20 |    10 |   20 |      9 |     20 |
  17. | 07:00:00 | xh458   |    0 |     0 |    0 |   NULL |   NULL |
  18. | 07:15:00 | xh458   |   10 |     0 |   10 |     10 |   NULL |
  19. | 07:30:00 | xh458   |    5 |     0 |    5 |     10 |   NULL |
  20. | 07:45:00 | xh458   |   30 |     0 |   30 |     10 |     30 |
  21. | 08:00:00 | xh458   |   25 |     0 |   25 |     10 |     30 |
  22. +----------+---------+------+-------+------+--------+--------+

Each function uses the rows in the current frame, which, per the window definition shown, extends from the first partition row to the current row. For the NTH_VALUE() calls, the current frame does not always include the requested row; in such cases, the return value is NULL.

In the absence of a frame clause, the default frame depends on whether an ORDER BY clause is present:

  • With ORDER BY: The default frame includes rows from the partition start through the current row, including all peers of the current row (rows equal to the current row according to the ORDER BY clause). The default is equivalent to this frame specification:

    1. RANGE BETWEEN UNBOUNDED PRECEDING AND CURRENT ROW

  • Without ORDER BY: The default frame includes all partition rows (because, without ORDER BY, all partition rows are peers). The default is equivalent to this frame specification:

    1. RANGE BETWEEN UNBOUNDED PRECEDING AND UNBOUNDED FOLLOWING

Because the default frame differs depending on presence or absence of ORDER BY, adding ORDER BY to a query to get deterministic results may change the results. (For example, the values produced by SUM() might change.) To obtain the same results but ordered per ORDER BY, provide an explicit frame specification to be used regardless of whether ORDER BY is present.

The meaning of a frame specification can be nonobvious when the current row value is NULL. Assuming that to be the case, these examples illustrate how various frame specifications apply:

  • ORDER BY X ASC RANGE BETWEEN 10 FOLLOWING AND 15 FOLLOWING

    The frame starts at NULL and stops at NULL, thus includes only rows with value NULL.

  • ORDER BY X ASC RANGE BETWEEN 10 FOLLOWING AND UNBOUNDED FOLLOWING

    The frame starts at NULL and stops at the end of the partition. Because an ASC sort puts NULL values first, the frame is the entire partition.

  • ORDER BY X DESC RANGE BETWEEN 10 FOLLOWING AND UNBOUNDED FOLLOWING

    The frame starts at NULL and stops at the end of the partition. Because a DESC sort puts NULL values last, the frame is only the NULL values.

  • ORDER BY X ASC RANGE BETWEEN 10 PRECEDING AND UNBOUNDED FOLLOWING

    The frame starts at NULL and stops at the end of the partition. Because an ASC sort puts NULL values first, the frame is the entire partition.

  • ORDER BY X ASC RANGE BETWEEN 10 PRECEDING AND 10 FOLLOWING

    The frame starts at NULL and stops at NULL, thus includes only rows with value NULL.

  • ORDER BY X ASC RANGE BETWEEN 10 PRECEDING AND 1 PRECEDING

    The frame starts at NULL and stops at NULL, thus includes only rows with value NULL.

  • ORDER BY X ASC RANGE BETWEEN UNBOUNDED PRECEDING AND 10 FOLLOWING

    The frame starts at the beginning of the partition and stops at rows with value NULL. Because an ASC sort puts NULL values first, the frame is only the NULL values.


Find a PHP function

Document created the 26/06/2006, last modified the 26/10/2018
Source of the printed document:https://www.gaudry.be/en/mysql-rf-window-functions-frames.html

The infobrol is a personal site whose content is my sole responsibility. The text is available under CreativeCommons license (BY-NC-SA). More info on the terms of use and the author.

References

  1. View the html document Language of the document:en Manuel MySQL : https://dev.mysql.com/

These references and links indicate documents consulted during the writing of this page, or which may provide additional information, but the authors of these sources can not be held responsible for the content of this page.
The author This site is solely responsible for the way in which the various concepts, and the freedoms that are taken with the reference works, are presented here. Remember that you must cross multiple source information to reduce the risk of errors.

Contents Haut