SQL 查询：根据其他列给列加值

Question

I have a dataset which looks like this:

Part          Runs          Duration          Date
-------------------------------------------------------
random_1       NULL          20              2020-01-01           
random_2       NULL          1               2020-01-01            
random_3       NULL          4               2020-01-01           
tot_rand       40            NULL            2020-01-01           
random_1       NULL          60              2020-01-02           
random_2       NULL          12              2020-01-02            
random_3       NULL          3               2020-01-02           
tot_rand       100           NULL            2020-01-02           
random_1       NULL          9               2020-01-10           
random_2       NULL          4               2020-01-10            
tot_rand       30            NULL            2020-01-10           

Now instead of NULL values in column Runs I want to add the tot_rand value for the same Date .

So that the result will look like this:

Part          Runs          Duration          Date     
-------------------------------------------------------
random_1       40           20              2020-01-01           
random_2       40           1               2020-01-01            
random_3       40           4               2020-01-01           
tot_rand       40           NULL            2020-01-01           
random_1       100          60              2020-01-02           
random_2       100          12              2020-01-02            
random_3       100          3               2020-01-02           
tot_rand       100          NULL            2020-01-02           
random_1       30           9               2020-01-10           
random_2       30           4               2020-01-10            
tot_rand       30           NULL            2020-01-10   

The reason for this is that in the end I want to create a new column called All which will count:

(Runs - Duration) / Runs

Final result

Part          Runs          Duration          Date           All
---------------------------------------------------------------------
random_1       40           20              2020-01-01       0.5
random_2       40           1               2020-01-01       0.975     
random_3       40           4               2020-01-01       0.9    
tot_rand       40           NULL            2020-01-01       NULL    
random_1       100          60              2020-01-02       0.4  
random_2       100          12              2020-01-02       0.88    
random_3       100          3               2020-01-02       0.97    
tot_rand       100          NULL            2020-01-02       NULL    
random_1       30           9               2020-01-10       0.7    
random_2       30           4               2020-01-10       0.86     
tot_rand       30           NULL            2020-01-10       NULL

I'm working in MariaDB / MySQL environment.

Maybe there is some other way to do this? I'm all open for suggestions.

Answer 1

SELECT t1.Part, t2.Runs, t1.Duration, `Date`, 1 - t1.Duration / t2.Runs `All`
FROM test t1
JOIN test t2 USING (`Date`)
WHERE t2.Part = 'tot_rand'

Only one row for each Date value must contain Part = 'tot_rand' .

The value of Runs will be replaced even if it is not NULL. Or you may use original t1.Runs but t2.Runs or COALESCE(t1.Runs, t2.Runs) in All calculation.

https://dbfiddle.uk/?rdbms=mysql_8.0&fiddle=440969fc417980aa3688bb4dac67bab2

Answer 2

You can use window functions:

select t.*,
       coalesce(runs, max(runs) over (partition by date)) as imputed_runs
from t;

If you specifically wanted the 'tot_rand' value, you can use conditional aggregation:

select t.*,
       coalesce(runs,
                 max(case when part = 'tot_rand' then runs end) over (partition by date)
               ) as imputed_runs
from t;

If you wanted an update , then one approach is:

update t join
       (select date, max(runs) as max_runs
        from t
        where part = 'tot_rand'
        group by date
       ) tt
       on t.date = tt.date
    set runs = tt.max_runs
    where tt.runs is null;

The subquery may not require aggregation, but it is unclear if there can be more than on 'tot_rand' value on a given date.