Question

I have 3 columns doc_date, chem_date, st_date

all 3 columns belong to different tables:

doc
doc_date      count(x)
01-02-2012      2
02-02-2012      3

chem
chem_date      count(x)
04-02-2012      1
06-02-2012      0

stock
st_date      count(x)
01-02-2012      1
03-02-2012      5

I want to write select clause like this

case doc_date 
    when '01' then count(x),
    case chem_date
    when '01' then count(x),
    case st_date 
    when '01' then count(x)
    end as '01',
case doc_date 
    when '02' then count(x),
    case chem_date
    when '02' then count(x),
    case st_date 
    when '02' then count(x)
    end as '02',

....up to 31 days

If some case statements have an entry on same date e.g if doc_date and st_date both exist on ’01-02-2012′ then their respective count should be added to the final count(x) means count(x) + count(x)

So, the answer should be:

01  02  03  04  05  06  07  08  09 10  11  12.. up to 31 days
3   3   5   1       0                          count(x) value for particular date

Explanation of the output: Here, the starting value for date 01 is 2 for doc table and for stock table the value is 1. So, the final value will become addition of them which is 3

The same rule will be applicable for others.

Any suggestions how can I achieve this output in SQL? Or any other way?

Answer 1

Here’s a pseudo code showing how it can be done:

select

case when date_format(doc_date,"%d") = '01' then sum_x else null end as '01',
case when date_format(doc_date,"%d") = '02' then sum_x else null end as '02',
--and etc

from (

    select doc_date, sum(doc_x) as sum_x              --sum the x value
    from
    (

        select doc_date,doc_x                         --combine all 3 table stuff
        from doc a

        union all

        select chem_date,chem_x
        from chem a

        union all

        select st_date,st_x
        from st a
    ) h
    group by doc_date               --use first col name as the union's col name

) g
limit 1,1       

If the doc, chem and st are combined using joins and have different schema, you may need to do pivot, either using multiple join to itself or use case... when...:

select

case when date_format(doc_date,"%d") = '01' then sum_x else null end as '01',
case when date_format(doc_date,"%d") = '02' then sum_x else null end as '02',
--and etc

from (

        select doc_date,                               
               a_x + b_x + c_x as sum_x            --sum them according to date
        from (select ... from doc where ...) a
        left join (select ... from chem where ...) b on a.doc_date=b.doc_date and --some join condition
        left join (select ... from st where ...) c on a.doc_date=c.doc_date and --some join condition

) g
limit 1,1       

I am working with an ambiguous question here, so clarify as needed,