Oracle sql 动态比较表/数据集中的行

Question

following issue:

I want to compare the rows within a table with each other.

Let's assume I have a testcase table with below data constellation:

ID  Result  Date

123 success 29.04.2021
123 error   28.04.2021
123 success 27.04.2021
123 success 26.04.2021
123 error   25.04.2021

234 success 29.04.2021
234 success 28.04.2021
234 success 27.04.2021
234 success 26.04.2021
234 error   25.04.2021

I want my query to return only those ID's where the result was identical for the last 3 days. But I want to handle the days parameter dynamically without using PL/SQL. That mean's - if I need to compare the last 5 days I just want to change the paramter in my sql statement. Can that be realized?

Back to my example above:

I want my query to give back:

ID
234

Best regards

Answer 1

From Oracle 12c, you can use MATCH_RECOGNIZE :

SELECT id
FROM   table_name
MATCH_RECOGNIZE (
  PARTITION BY id
  ORDER BY "DATE"
  ONE ROW PER MATCH
  PATTERN ( same{3} $ )
  DEFINE same AS FIRST( result ) = result
)

Or, in earlier versions, you can use:

SELECT id
FROM   (
  SELECT id,
         ROW_NUMBER() OVER ( PARTITION BY id ORDER BY "DATE" DESC )
           AS id_rn,
         ROW_NUMBER() OVER ( PARTITION BY id, result ORDER BY "DATE" DESC )
           AS result_rn
  FROM   table_name
)
WHERE id_rn = 3
AND   result_rn = 3

Which, for the sample data:

CREATE TABLE table_name ( ID, Result, "DATE" ) AS
SELECT 123, 'success', DATE '2021-04-29' FROM DUAL UNION ALL
SELECT 123, 'error',   DATE '2021-04-28' FROM DUAL UNION ALL
SELECT 123, 'success', DATE '2021-04-27' FROM DUAL UNION ALL
SELECT 123, 'success', DATE '2021-04-26' FROM DUAL UNION ALL
SELECT 123, 'error',   DATE '2021-04-25' FROM DUAL UNION ALL
SELECT 234, 'success', DATE '2021-04-29' FROM DUAL UNION ALL
SELECT 234, 'success', DATE '2021-04-28' FROM DUAL UNION ALL
SELECT 234, 'success', DATE '2021-04-27' FROM DUAL UNION ALL
SELECT 234, 'success', DATE '2021-04-26' FROM DUAL UNION ALL
SELECT 234, 'error',   DATE '2021-04-25' FROM DUAL;

Outputs:

ID
234

db<>fiddle here

Answer 2

You can use just simple group by + having clause:

select id
from t
where dt>=sysdate-3
group by id
having count(distinct result)=1 -- just one distinct `result`
and count(*)=3 -- only 3 rows >= sysdate-3
and count(distinct dt)=3 -- all 3 days are present;

Full test cast with test data:

with t(ID, Result, Dt) as (
select 123, 'success' , to_date('29.04.2021','dd.mm.yyyy') from dual union all
select 123, 'error  ' , to_date('28.04.2021','dd.mm.yyyy') from dual union all
select 123, 'success' , to_date('27.04.2021','dd.mm.yyyy') from dual union all
select 123, 'success' , to_date('26.04.2021','dd.mm.yyyy') from dual union all
select 123, 'error  ' , to_date('25.04.2021','dd.mm.yyyy') from dual union all
select 234, 'success' , to_date('29.04.2021','dd.mm.yyyy') from dual union all
select 234, 'success' , to_date('28.04.2021','dd.mm.yyyy') from dual union all
select 234, 'success' , to_date('27.04.2021','dd.mm.yyyy') from dual union all
select 234, 'success' , to_date('26.04.2021','dd.mm.yyyy') from dual union all
select 234, 'error  ' , to_date('25.04.2021','dd.mm.yyyy') from dual
)
select id
from t
where dt>=sysdate-3
group by id
having count(distinct result)=1 -- just one distinct `result`
and count(*)=3 -- only 3 rows >= sysdate-3
and count(distinct dt)=3 -- all 3 days are present
;

Results:

        ID
----------
       234

Modified query as per a clarification in the comment above:

I want to check last 3 days starting from today. If I only have two entries then I would like to compare only two entries...

select id
from t
where dt>=sysdate-3
group by id
having count(distinct result)=1 -- just one distinct `result`
;

Full test case:

with t(ID, Result, Dt) as (
select 123, 'success' , to_date('29.04.2021','dd.mm.yyyy') from dual union all
select 123, 'error  ' , to_date('28.04.2021','dd.mm.yyyy') from dual union all
select 123, 'success' , to_date('27.04.2021','dd.mm.yyyy') from dual union all
select 123, 'success' , to_date('26.04.2021','dd.mm.yyyy') from dual union all
select 123, 'error  ' , to_date('25.04.2021','dd.mm.yyyy') from dual union all
select 234, 'success' , to_date('29.04.2021','dd.mm.yyyy') from dual union all
select 234, 'success' , to_date('28.04.2021','dd.mm.yyyy') from dual union all
select 234, 'success' , to_date('27.04.2021','dd.mm.yyyy') from dual union all
select 234, 'success' , to_date('26.04.2021','dd.mm.yyyy') from dual union all
select 234, 'error  ' , to_date('25.04.2021','dd.mm.yyyy') from dual
)
select id
from t
where dt>=sysdate-3
group by id
having count(distinct result)=1 -- just one distinct `result`
;
----------------------
--Results:

        ID
----------
       234

PS. It's better to give test data in this format:

with t(ID, Result, Dt) as (
select 123, 'success' , to_date('29.04.2021','dd.mm.yyyy') from dual union all
select 123, 'error  ' , to_date('28.04.2021','dd.mm.yyyy') from dual union all
select 123, 'success' , to_date('27.04.2021','dd.mm.yyyy') from dual union all
select 123, 'success' , to_date('26.04.2021','dd.mm.yyyy') from dual union all
select 123, 'error  ' , to_date('25.04.2021','dd.mm.yyyy') from dual union all
select 234, 'success' , to_date('29.04.2021','dd.mm.yyyy') from dual union all
select 234, 'success' , to_date('28.04.2021','dd.mm.yyyy') from dual union all
select 234, 'success' , to_date('27.04.2021','dd.mm.yyyy') from dual union all
select 234, 'success' , to_date('26.04.2021','dd.mm.yyyy') from dual union all
select 234, 'error  ' , to_date('25.04.2021','dd.mm.yyyy') from dual
)
select *
from t;

Answer 3

Assuming you have one row per id per day, then you can use lag() :

select t.*
from (select t.*,
             lag(result, 2) over (partition by id order by date) as date_2,
             lag(result, 2) over (partition by id, result order by date) as date_result_2
      from t
     ) t
where date_2 = date_result_2;

This returns all rows where the previous two rows (days assuming that there is one row per id per day) have the same result.

You can add filters on the time frame you care about. This should probably be in the outer query.

What this does is look two rows in the past, but in two different ways:

• Once per id .
• Once per id and result

When these are the same, then the row two rows back has the same result and the row in-between does as well.