I have a requirement to fetch previous row or lag records where there are some missing previous ids.
Database : Oracle 12c
Example data:
BRANCH | PERIOD | QTY |
---|---|---|
105 | 319 | 17 |
105 | 320 | 20 |
105 | 321 | 32 |
105 | 322 | 61 |
107 | 319 | 17 |
107 | 321 | 18 |
107 | 322 | 16 |
108 | 319 | 21 |
108 | 322 | 27 |
I want the results in below format: If you see for branch 107 : the period 319 is missing and for branch 108 : 320,321 are missing. So if there are any missing previous records then the prev_period_<>_Qty columns should be 0.
Can you please help in achieving this.
BRANCH | PERIOD | QTY | PREV_PERIOD_1_QTY | PREV_PERIOD_2_QTY | PREV_PERIOD_3_QTY |
---|---|---|---|---|---|
105 | 319 | 17 | 0 | 0 | 0 |
105 | 320 | 20 | 17 | 0 | 0 |
105 | 321 | 32 | 20 | 17 | 0 |
105 | 322 | 61 | 32 | 20 | 17 |
107 | 319 | 17 | 0 | 0 | 0 |
107 | 321 | 18 | 0 | 17 | 0 |
107 | 322 | 16 | 18 | 0 | 17 |
108 | 319 | 21 | 0 | 0 | 0 |
108 | 322 | 27 | 0 | 0 | 21 |
From Oracle 12, you can use MATCH_RECOGNIZE
to do row-by-row processing:
SELECT branch,
period,
qty,
COALESCE(prev_period_1_qty, 0) AS prev_period_1_qty,
COALESCE(prev_period_2_qty, 0) AS prev_period_2_qty,
COALESCE(prev_period_3_qty, 0) AS prev_period_3_qty
FROM table_name
MATCH_RECOGNIZE (
PARTITION BY branch
ORDER BY period DESC
MEASURES
curr.period AS period,
curr.qty AS qty,
prev1.qty AS prev_period_1_qty,
prev2.qty AS prev_period_2_qty,
prev3.qty AS prev_period_3_qty
ONE ROW PER MATCH
AFTER MATCH SKIP TO NEXT ROW
PATTERN (curr prev1? prev2? prev3?)
DEFINE
prev1 AS curr.period - 1 = period,
prev2 AS curr.period - 2 = period,
prev3 AS curr.period - 3 = period
)
ORDER BY branch, period
Or, using LAG
:
SELECT branch,
period,
qty,
CASE
WHEN p1 = period - 1
THEN q1 ELSE 0
END AS prev_period_1_qty,
CASE
WHEN p1 = period - 2 THEN q1
WHEN p2 = period - 2 THEN q2
ELSE 0
END AS prev_period_2_qty,
CASE
WHEN p1 = period - 3 THEN q1
WHEN p2 = period - 3 THEN q2
WHEN p3 = period - 3 THEN q3
ELSE 0
END AS prev_period_3_qty
FROM (
SELECT t.*,
LAG(period, 1) OVER (PARTITION BY branch ORDER BY period) AS p1,
LAG(period, 2) OVER (PARTITION BY branch ORDER BY period) AS p2,
LAG(period, 3) OVER (PARTITION BY branch ORDER BY period) AS p3,
LAG(qty, 1, 0) OVER (PARTITION BY branch ORDER BY period) AS q1,
LAG(qty, 2, 0) OVER (PARTITION BY branch ORDER BY period) AS q2,
LAG(qty, 3, 0) OVER (PARTITION BY branch ORDER BY period) AS q3
FROM table_name t
)
Which, for the sample data:
CREATE TABLE table_name (BRANCH, PERIOD, QTY) AS
SELECT 105, 319, 17 FROM DUAL UNION ALL
SELECT 105, 320, 20 FROM DUAL UNION ALL
SELECT 105, 321, 32 FROM DUAL UNION ALL
SELECT 105, 322, 61 FROM DUAL UNION ALL
SELECT 107, 319, 17 FROM DUAL UNION ALL
SELECT 107, 321, 18 FROM DUAL UNION ALL
SELECT 107, 322, 16 FROM DUAL UNION ALL
SELECT 108, 319, 21 FROM DUAL UNION ALL
SELECT 108, 322, 27 FROM DUAL;
Both output:
BRANCH PERIOD QTY PREV_PERIOD_1_QTY PREV_PERIOD_2_QTY PREV_PERIOD_3_QTY 105 319 17 0 0 0 105 320 20 17 0 0 105 321 32 20 17 0 105 322 61 32 20 17 107 319 17 0 0 0 107 321 18 0 17 0 107 322 16 18 0 17 108 319 21 0 0 0 108 322 27 0 0 21
db<>fiddle here