在Oracle sql中，plsql创建了一个构建变量的sql

Question

Wonder if someone could guide me in the right direction.

I have a data table structure as below:

table2_type char (1)

table2_key char (40)

table2_reftype char (20)

table2_seq char (10)

table2_ref char (50)

and data rows like(comma separated for readability only):

I,123456789,typea,0000000001,abc

I,123456789,typeb,0000000002,999

I,123456789,typec,0000000003,9z9

I,123456789,typed,0000000004,zyx

I,123456789,typee,0000000005,qwe

I,987654321,typea,0000000006,bcd

I,987654321,typeb,0000000007,444

What I have to do at the moment to get all the table_ref values for the same table_key values is below (spaced for ease of reading):

select table1_bkey,

nvl((case when (select table2_ref from table2 where table2_key = table1_bkey and table2_reftype = rpad('typea',20)) is not null then (select table2_ref from table2 where table2_key = table1_bkey and table2_reftype = rpad('typea',20)) else (select table2_ref from table2 where table2_key = table1_alt and table2_reftype = rpad('typea',20)) end),' ') as REF_A,

nvl((case when (select table2_ref from table2 where table2_key = table1_bkey and table2_reftype = rpad('typeb',20)) is not null then (select table2_ref from table2 where table2_key = table1_bkey and table2_reftype = rpad('typeb',20)) else (select table2_ref from table2 where table2_key = table1_alt and table2_reftype = rpad('typeb',20)) end),' ') as REF_B,

nvl((case when (select table2_ref from table2 where table2_key = table1_bkey and table2_reftype = rpad('typec',20)) is not null then (select table2_ref from table2 where table2_key = table1_bkey and table2_reftype = rpad('typec',20)) else (select table2_ref from table2 where table2_key = table1_alt and table2_reftype = rpad('typec',20)) end),' ') as REF_C,

nvl((case when (select table2_ref from table2 where table2_key = table1_bkey and table2_reftype = rpad('typed',20)) is not null then (select table2_ref from table2 where table2_key = table1_bkey and table2_reftype = rpad('typed',20)) else (select table2_ref from table2 where table2_key = table1_alt and table2_reftype = rpad('typed',20)) end),' ') as REF_D,

nvl((case when (select table2_ref from table2 where table2_key = table1_bkey and table2_reftype = rpad('typee',20)) is not null then (select table2_ref from table2 where table2_key = table1_bkey and table2_reftype = rpad('typee',20)) else (select table2_ref from table2 where table2_key = table1_alt and table2_reftype = rpad('typee',20)) end),' ') as REF_E

from table1 where table1_bkey = '123456789';

I currently could have up to 25 different table2_reftype per table2_key so I have to do 25 case statements. And if/when new table2_reftype are added I have to do new case statements.

My question, is there a way to get all the table2_reftype values per table2_key without having to do multiple case statements eg:

illustration purpose only, not to be taking literal as code in any way!!

while table2_key = table1_bkey

get table2_reftype 

then for each table2_reftype 

nvl((case when (select table2_ref from table2 where table2_key = table1_bkey and table2_reftype = rpad(<table2_reftype variable>,20)) is not null then (select table2_ref from table2 where table2_key = table1_bkey and table2_reftype = rpad(<table2_reftype variable>,20)) else (select table2_ref from table2 where table2_key = table1_alt and table2_reftype = rpad(<table2_reftype variable>,20)) end),' ') as REF_<table2_reftype  variable>

where table1_bkey = '123456789';

First post so hope I have put enough information.

Answer 1

I think you could use something like this:

select key, ref_a, ref_b, ref_c
  from (
    select table1_bkey key, table2_ref ref, trim(table2_reftype) type, 
           row_number() over (partition by table1_bkey, table2_reftype 
                              order by case table2_key when table1_bkey then 1 else 2 end) rn
      from table1
      left join table2 
        on trim(table2_reftype) in ('typea', 'typeb', 'typec')
        and table2_key in (table1_bkey, table1_alt))
  pivot (max(ref) for type in ('typea' ref_a, 'typeb' ref_b, 'typec' ref_c))
  where rn = 1

^{dbfiddle demo}

What is going on here? I left join -ed both tables using your conditions. Then I assigned rn which depends on if match is done on base key or alternate key. Base has higher priority, so in case both were found second is removed ( where rn = 1 ). Rest was easy, just pivot data for types which you hardcoded. Also please note that I trimmed ref_type .

This is example for 3 types, when you add new, just put it in both lists. As you can see in my dbfiddle query works, but I had to imagine some data which you did not describe (for table1 ).

In case something is wrong and you would not be able to correct / adapt this solution please edit your question and provide sample data for both tables in the form of working create and insert statements and show us desired result for this data.

You can also solve it like you described in your pseudo-code, but this requires either dynamic PLSQL, either constructing query in some other code (I did similiar things in .NET). Why? Because Oracle has to know all columns when it compiles query.

I hope this helps :)