How to select timestamp values in PostgreSQL under conditions? - sql

Question

I have a database table 'table1' as follows:
f_key | begin | counts|
1 | 2018-10-04 | 15 |
1 | 2018-10-06 | 20 |
1 | 2018-10-08 | 34 |
1 | 2018-10-09 | 56 |
I have another database table 'table2' as follows:
f_key | p_time | percent|
1 | 2018-10-05 | 80 |
1 | 2018-10-07 | 90 |
1 | 2018-10-08 | 70 |
1 | 2018-10-10 | 60 |
The tables can be joined by the f_key field.
I want to get a combined table as shown below:
If the begin time is earlier than any of the p_time then the p_time value in the combined table would be the same as begin time and the percent value would be 50. (As shown in row 1 in the following table)
If the begin time is later than any of the p_time then the p_time value in the combined table would be the very next available p_time and the percent value would be the corresponding value of the selected p_time.
(As shown in row 2, 3 and 4 in the following table)
row | f_key | begin | counts| p_time | percent|
1 | 1 | 2018-10-04 | 15 | 2018-10-04 | 50 |
2 | 1 | 2018-10-06 | 20 | 2018-10-05 | 80 |
3 | 1 | 2018-10-08 | 34 | 2018-10-07 | 90 |
4 | 1 | 2018-10-09 | 56 | 2018-10-08 | 70 |

Answer 1

You can try to use row_number window function to make row number which is the closest row from table1 by begin.
then use coalesce function to let begin time is earlier than any of the p_time then the p_time value in the combined table would be the same as begin time and the percent value would be 50
PostgreSQL 9.6 Schema Setup:
CREATE TABLE table1(
f_key INT,
begin DATE,
counts INT
);
INSERT INTO table1 VALUES (1,'2018-10-04',15);
INSERT INTO table1 VALUES (1,'2018-10-06',20);
INSERT INTO table1 VALUES (1,'2018-10-08',34);
INSERT INTO table1 VALUES (1,'2018-10-09',56);
CREATE TABLE table2(
f_key INT,
p_time DATE,
percent INT
);
INSERT INTO table2 VALUES (1, '2018-10-05',80);
INSERT INTO table2 VALUES (1, '2018-10-07',90);
INSERT INTO table2 VALUES (1, '2018-10-08',70);
INSERT INTO table2 VALUES (1, '2018-10-10',60);
Query 1:
SELECT ROW_NUMBER() OVER(ORDER BY begin) "row",
t1.f_key,
t1.counts,
coalesce(t1.p_time,t1.begin) p_time,
coalesce(t1.percent,50) percent
FROM (
SELECT ROW_NUMBER() OVER(PARTITION BY t1.begin,t1.f_key order by t2.p_time desc) rn,
t2.p_time,
t2.percent,
t1.counts,
t1.f_key,
t1.begin
FROM table1 t1
LEFT JOIN table2 t2 ON t1.f_key = t2.f_key and t1.begin > t2.p_time
)t1
WHERE rn = 1
Results:
| row | f_key | counts | p_time | percent |
|-----|-------|--------|------------|---------|
| 1 | 1 | 15 | 2018-10-04 | 50 |
| 2 | 1 | 20 | 2018-10-05 | 80 |
| 3 | 1 | 34 | 2018-10-07 | 90 |
| 4 | 1 | 56 | 2018-10-08 | 70 |