Oracle SQL - 获取每列的最新值

Question

Suppose I have a table that looks like this:

There are 2 ISIN's in the table (ISIN 1 and 2). If you look at the first row, you'll see that all fields are filled with values. In the second row there is only one field filled, which represents an update of the field (all empty fields didn't changed).

Basically I want the latest value of each field (grouped by ISIN).

I was thinking of creating a Materialized view, if there is a build-in function or an easy way to achive my result. If there is not, I'll create a pipe function and iterate over the records.

Answer 1

You can use the LAST_VALUE analytic function:

SELECT ts,
       id,
       isin,
       value1,
       value2,
       value3,
       value4,
       value5,
       value6
FROM   (
  SELECT ts,
         id,
         isin,
         LAST_VALUE(value1) IGNORE NULLS OVER ( PARTITION BY isin ORDER BY ts, id ) AS value1,
         LAST_VALUE(value2) IGNORE NULLS OVER ( PARTITION BY isin ORDER BY ts, id ) AS value2,
         LAST_VALUE(value3) IGNORE NULLS OVER ( PARTITION BY isin ORDER BY ts, id ) AS value3,
         LAST_VALUE(value4) IGNORE NULLS OVER ( PARTITION BY isin ORDER BY ts, id ) AS value4,
         LAST_VALUE(value5) IGNORE NULLS OVER ( PARTITION BY isin ORDER BY ts, id ) AS value5,
         LAST_VALUE(value6) IGNORE NULLS OVER ( PARTITION BY isin ORDER BY ts, id ) AS value6,
         ROW_NUMBER() OVER ( PARTITION BY isin ORDER BY ts DESC, id DESC ) AS rn
  FROM   table_name
)
WHERE rn = 1

or, you can use MAX()... KEEP ( DENSE_RANK... ) :

SELECT MAX(ts) AS ts,
       MAX(id) KEEP ( DENSE_RANK LAST ORDER BY ts, id ) AS id,
       isin,
       MAX(value1) KEEP (
         DENSE_RANK FIRST
         ORDER BY
           CASE WHEN value1 IS NOT NULL THEN ts END DESC NULLS LAST,
           id DESC
       ) AS value1,
       MAX(value2) KEEP (
         DENSE_RANK FIRST
         ORDER BY
           CASE WHEN value2 IS NOT NULL THEN ts END DESC NULLS LAST,
           id DESC
       ) AS value2,
       MAX(value3) KEEP (
         DENSE_RANK FIRST
         ORDER BY
           CASE WHEN value3 IS NOT NULL THEN ts END DESC NULLS LAST,
           id DESC
       ) AS value3,
       MAX(value4) KEEP (
         DENSE_RANK FIRST
         ORDER BY
           CASE WHEN value4 IS NOT NULL THEN ts END DESC NULLS LAST,
           id DESC
       ) AS value4,
       MAX(value5) KEEP (
         DENSE_RANK FIRST
         ORDER BY
           CASE WHEN value5 IS NOT NULL THEN ts END DESC NULLS LAST,
           id DESC
       ) AS value5,
       MAX(value6) KEEP (
         DENSE_RANK FIRST
         ORDER BY
           CASE WHEN value6 IS NOT NULL THEN ts END DESC NULLS LAST,
           id DESC
       ) AS value6
FROM   table_name
GROUP BY isin

Which, for the sample data:

CREATE TABLE table_name ( ts, id, isin, value1, value2, value3, value4, value5, value6 ) AS
SELECT SYSDATE - 5, 1, 1, 'A',  'B',  'C',  'D',  'E',  'F'  FROM DUAL UNION ALL
SELECT SYSDATE - 4, 2, 1, NULL, NULL, 'G',  NULL, NULL, NULL FROM DUAL UNION ALL
SELECT SYSDATE - 3, 3, 1, NULL, 'H',  NULL, NULL, 'I',  NULL FROM DUAL UNION ALL
SELECT SYSDATE - 2, 4, 1, NULL, NULL, NULL, 'J',  NULL, NULL FROM DUAL UNION ALL
SELECT SYSDATE - 2, 5, 2, 'K',  'L',  'M',  'N',  'O',  'P'  FROM DUAL UNION ALL
SELECT SYSDATE - 1, 6, 2, NULL, 'Q',  NULL, NULL, NULL, NULL FROM DUAL UNION ALL
SELECT SYSDATE - 0, 7, 2, NULL, NULL, NULL, 'R',  NULL, NULL FROM DUAL;

Both output:

 TS |  ID | ISIN | VALUE1 |  VALUE2 |  VALUE3 |  VALUE4 |  VALUE5 |  VALUE6:-------- | -: |  ---: |:----- |:----- |:----- |:----- |:----- |:----- 09-NOV-20 |  4 |  1 |  A | H |  G | J | I | F 11-NOV-20 | 7 |  2 |  K |  Q | M | R |  O | P 

db<>fiddle here

Answer 2

This is it:

SELECT SUBSTR(LISTAGG("col1" ,', ') WITHIN GROUP (ORDER BY "col1"),-1,1)as value1,
 SUBSTR(LISTAGG("col2" ,', ') WITHIN GROUP (ORDER BY "col2"),-1,1)as value2,
 SUBSTR(LISTAGG("col3" ,', ') WITHIN GROUP (ORDER BY "col3"),-1,1)as value3,
 SUBSTR(LISTAGG("col4" ,', ') WITHIN GROUP (ORDER BY "col4"),-1,1)as value4,
 SUBSTR(LISTAGG("col5" ,', ') WITHIN GROUP (ORDER BY "col5"),-1,1)as value5,
 SUBSTR(LISTAGG("col6" ,', ') WITHIN GROUP (ORDER BY "col6"),-1,1)as value6

FROM Table1
group by "isin";

check: http://sqlfiddle.com/#!4/f02bd/5

Answer 3

You can use the KEEP clause as follows:

SELECT ISIN, 
       MAX(VALUE1) KEEP (DENSE_RANK FIRST ORDER BY CASE WHEN VALUE1 IS NULL THEN NULL ELSE "TIMESTAMP" END DESC NULLS LAST) AS VALUE1,
       MAX(VALUE2) KEEP (DENSE_RANK FIRST ORDER BY CASE WHEN VALUE2 IS NULL THEN NULL ELSE "TIMESTAMP" END DESC NULLS LAST) AS VALUE2,
       ...
  FROM YOUR_TABLE
GROUP BY ISIN;