query? "Is there a SQL query to check for gaps in Postgresql entries for each day?"

Kurasao
May 17, 2023
256 views
0 votes
2 Answers

I have a log table "table1" with values appended every day and need to find "column3" values and dates from "column1", for which the data wasn’t recorded. For example, my table looks like this:

column1	column2	column3
2022-07-14	274,5	markus
2022-07-14	251,2	tess
2022-07-14	162,6	mike
2022-07-15	286,9	markus
2022-07-15	254,8	tess
2022-07-16	289,1	markus
2022-07-17	295,2	markus
2022-07-17	260,0	tess
2022-07-17	182,3	mike

Everything is ok with column3 = 'markus', but I need to get something like this as output:

column1	column3
2022-07-15	mike
2022-07-16	tess
2022-07-16	mike

Answers

- lemon
- May 17, 2023 at 1:58 pm
- 0 votes
0
One way of addressing this problem is by:
- rebuilding all combinations of names and dates
- left-joining this table with your original tables
- filtering on records whose table values is null
```
WITH cte_dates AS (
    SELECT DISTINCT column1 AS "date" FROM tab
), cte_names AS (
    SELECT DISTINCT column3 AS "name" FROM tab
)
SELECT cte_dates.date,
       cte_names.name
FROM       cte_dates
CROSS JOIN cte_names
LEFT JOIN tab
       ON cte_dates.date = tab.column1
      AND cte_names.name = tab.column3
WHERE tab.column2 IS NULL
```
Output:

date name

2022-07-16T00:00:00.000Z mike

2022-07-16T00:00:00.000Z tess

2022-07-15T00:00:00.000Z mike

Check the demo here.

If gaps can be found among your dates, you need to use generate_series with boundary dates to generate the corresponding calendar in the first cte:
```
WITH cte_dates AS (
    SELECT "date"
    FROM (SELECT MIN(column1) AS startdt, 
                 MAX(column1) AS enddt 
          FROM tab) boundaries
    CROSS JOIN generate_series( startdt :: timestamp 
                              , enddt   :: timestamp
                              , '1 day' :: interval ) "date"
),
...
```
Check the demo here.
Login or Signup to reply.

- MHA
- May 17, 2023 at 2:19 pm
- 0 votes
0
At first, you must build a time base for the whole period, then cross-join it with distinct of column3 to create all possibilities, then subtract it from your base data as follows:
```
with cross_sql as (select * from 
(SELECT timebase 
FROM   generate_series(timestamp '2022-07-14'
                     , timestamp '2022-07-17'
                     , interval  '1 day') as timebase) A
cross join                   
(select distinct column3 from table1) B )

select column3, timebase from cross_sql 
except
select column3, column1 from table1;
```
Login or Signup to reply.