first_value(), nth_value(), last_value()

These three window functions fall into the second group, Window functions that return column(s) of another row within the window in the section List of all window functions. Each of the functions in the second group makes obvious sense when the scope within which the specified row is found is the entire window. Only this use will be described here. When used this way, each of these functions, as their names suggest, return the same result for each row in the current window. See Examples that illustrate all three functions below.

If you have a use case that requires a specifically tailored window frame, then see the section The frame_clause.

first_value()

Signature:

input value:       anyelement
return value:      anyelement

Purpose: Return the specified value from the first row, in the specified sort order, in the current window frame. If you specify the frame_clause to start at a fixed offset before the current row, then first_value() would produce the same result as would the correspondingly parameterized lag(). If this is your aim, then you should use lag() for clarity.

nth_value()

Signature:

input value:       anyelement, int
return value:      anyelement

Purpose: Return the specified value from the "Nth" row, in the specified sort order, in the current window frame. The second, mandatory, parameter specifies "N" in "Nth".

last_value()

Signature:

input value:       anyelement
return value:      anyelement

Purpose: Return the specified value from the last row, in the specified sort order, in the current window frame.

Examples that illustrate all three functions

This example uses table "t1". Notice that it has been contrived so that the last "v" (ordered by "k") for each value of "class" is NULL.

Use the technique shown in the section Using nth_value() and last_value() to return the whole row so that each of the three window functions produces all of the fields in each row:

drop type if exists rt cascade;
create type rt as (class int, k int, v int);

select
  class,
  k,
  first_value((class, k, v)::rt::text)    over w as fv,
  nth_value  ((class, k, v)::rt::text, 3) over w as nv,
  last_value ((class, k, v)::rt::text)    over w as lv
from t1
window w as (
  partition by class
  order by k
  range between unbounded preceding and unbounded following);

Here is the result. To make it easier to see the pattern, a break has been manually inserted here between each successive set of rows with the same value for "class".

 class | k  |    fv     |    nv     |   lv    
-------+----+-----------+-----------+---------
     1 |  1 | (1,1,1)   | (1,3,3)   | (1,5,)
     1 |  2 | (1,1,1)   | (1,3,3)   | (1,5,)
     1 |  3 | (1,1,1)   | (1,3,3)   | (1,5,)
     1 |  4 | (1,1,1)   | (1,3,3)   | (1,5,)
     1 |  5 | (1,1,1)   | (1,3,3)   | (1,5,)

     2 |  6 | (2,6,6)   | (2,8,8)   | (2,10,)
     2 |  7 | (2,6,6)   | (2,8,8)   | (2,10,)
     2 |  8 | (2,6,6)   | (2,8,8)   | (2,10,)
     2 |  9 | (2,6,6)   | (2,8,8)   | (2,10,)
     2 | 10 | (2,6,6)   | (2,8,8)   | (2,10,)

     3 | 11 | (3,11,11) | (3,13,13) | (3,15,)
     3 | 12 | (3,11,11) | (3,13,13) | (3,15,)
     3 | 13 | (3,11,11) | (3,13,13) | (3,15,)
     3 | 14 | (3,11,11) | (3,13,13) | (3,15,)
     3 | 15 | (3,11,11) | (3,13,13) | (3,15,)

     4 | 16 | (4,16,16) | (4,18,18) | (4,20,)
     4 | 17 | (4,16,16) | (4,18,18) | (4,20,)
     4 | 18 | (4,16,16) | (4,18,18) | (4,20,)
     4 | 19 | (4,16,16) | (4,18,18) | (4,20,)
     4 | 20 | (4,16,16) | (4,18,18) | (4,20,)

     5 | 21 | (5,21,21) | (5,23,23) | (5,25,)
     5 | 22 | (5,21,21) | (5,23,23) | (5,25,)
     5 | 23 | (5,21,21) | (5,23,23) | (5,25,)
     5 | 24 | (5,21,21) | (5,23,23) | (5,25,)
     5 | 25 | (5,21,21) | (5,23,23) | (5,25,)

Notice that the ::text typecast of a "row" type value renders NULL simply as an absence. This explains why you see, for example, "(1,5,)" for each value produced by last_value() in the window where "k=1". This basic example certainly demonstrates the meaning of "first", "Nth" (for "N=3"), and "last". But it isn't very useful because, just as these names suggest, the output is the same for each row in a particular window. The following query adds a conventional GROUP BY clause. It also extracts the interesting fields from the "row" type value that each window function produces as individual values.

drop type if exists rt cascade;
create type rt as (class int, k int, v int);
\pset null '??'

with a as (
  select
    class,
    first_value((class, k, v)::rt)    over w as fv,
    nth_value  ((class, k, v)::rt, 3) over w as nv,
    last_value ((class, k, v)::rt)    over w as lv
  from t1
  window w as (
    partition by class
    order by k
    range between unbounded preceding and unbounded following))
select
  (fv).class as fv_class,
  (fv).k     as fv_k,
  (fv).v     as fv_v,
  (nv).k     as nv_k,
  (nv).v     as nv_v,
  (lv).k     as lv_k,
  (lv).v     as lv_v
from a
group by
  (fv).class,
  (fv).k,
  (fv).v,
  (nv).k,
  (nv).v,
  (lv).k,
  (lv).v
order by 1;

This is the result:

 fv_class | fv_k | fv_v | nv_k | nv_v | lv_k | lv_v 
----------+------+------+------+------+------+------
        1 |    1 |    1 |    3 |    3 |    5 |   ??
        2 |    6 |    6 |    8 |    8 |   10 |   ??
        3 |   11 |   11 |   13 |   13 |   15 |   ??
        4 |   16 |   16 |   18 |   18 |   20 |   ??
        5 |   21 |   21 |   23 |   23 |   25 |   ??