asked on

Converting old SQL Server outer join syntax

I'm a Power Builder (PB) developer. I've migrated PB from 10 .5 to 12.5, and now I need to migrate the stored procedures from SQL Server 2005 to 2012 as well.

I get a invalid expression error when I run the procedures in Power Builder. We have previously been using a lower database compatibility model in 2005, which allowed the procedures to work there. I have learned that =* is the old way to write right outer joins.

For example:
select *
from A
right outer join
B
on A.bid = B.id

...is written in the old style like:
select *
from A,
B
where A.bid =* B.id

Here is a part of my procedure that uses *=:
SELECT DISTINCT
a_stmt.cyc_dte,
a_carr.carr_nm,
a_stmt.amt_du_ic
FROM a_stmt,
s_dataccs,
a_carr,
a_icc
WHERE ( a_stmt.ic_cntct_id *= a_icc.ic_cntct_id ) and -- HERE
( a_icc.usg_ind = 'S' ) and
( a_carr.acna = a_stmt.acna ) and
( s_dataccs.acna = a_stmt.acna ) and ( s_dataccs.user_id = @user_id ) and
( ( a_stmt.acna = @acna ) or ( @acna = '' ) ) AND
( ( a_stmt.juris_id = @juris_id ) or ( @juris_id = '' ) ) AND
( a_stmt.jrnl_mo_yr = @jrnl_mo_yr ) AND
( a_stmt.amt_du_ic < 0 )

I have converted that to:
SELECT DISTINCT
a_stmt.cyc_dte,
a_carr.carr_nm,
a_stmt.amt_du_ic
from
a_stmt left outer join a_icc on a_stmt.ic_cntct_id = a_icc.ic_cntct_id
join a_carr on a_stmt.acna = a_carr.acna join s_dataccs on a_stmt.acna = s_dataccs.acna and
( a_icc.usg_ind = 'S' ) and
( s_dataccs.user_id = @user_id ) and
( ( a_stmt.acna = @acna ) or ( @acna = '' ) ) AND
( ( a_stmt.juris_id = @juris_id ) or ( @juris_id = '' ) ) AND
( a_stmt.jrnl_mo_yr = @jrnl_mo_yr ) AND
( a_stmt.amt_du_ic < 0 )

Here is one more :
SELECT r_rj.juris_id, "GRAND TOTAL", "",
round( isnull (sum(a_ovrudr.bd_intl + a_ovrudr.bd_inter + a_ovrudr.bd_intra+
a_ovrudr.bd_fed_tx + a_ovrudr.bd_st_lcl_tx), 0), 2),
r_rj.ord
FROM a_ovrudr, r_rj
WHERE ( a_ovrudr.juris_id =* r_rj.juris_id) and
( a_ovrudr.acna = @acna ) AND
(( SUBSTRING (a_ovrudr.jrnl_mo_yr, 3,4) = @yr_1 ) OR
( SUBSTRING (a_ovrudr.jrnl_mo_yr, 3,4) = @yr_2 ) OR
( SUBSTRING (a_ovrudr.jrnl_mo_yr, 3,4) = @yr_3 ) OR
( SUBSTRING (a_ovrudr.jrnl_mo_yr, 3,4) = @yr_4 ) OR
( SUBSTRING (a_ovrudr.jrnl_mo_yr, 3,4) = @yr_5 ) )
GROUP BY a_ovrudr.acna, r_rj.juris_id, r_rj.ord
ORDER BY yr, mo, r_rj.ord

which I have converted as :
SELECT r_rj.juris_id, "GRAND TOTAL", "",
round( isnull (sum(a_ovrudr.bd_intl + a_ovrudr.bd_inter + a_ovrudr.bd_intra+
a_ovrudr.bd_fed_tx + a_ovrudr.bd_st_lcl_tx), 0), 2),
r_rj.ord
FROM a_ovrudr right outer join r_rj on a_ovrudr.juris_id = r_rj.juris_id and
( a_ovrudr.acna = @acna ) AND
(( SUBSTRING (a_ovrudr.jrnl_mo_yr, 3,4) = @yr_1 ) OR
( SUBSTRING (a_ovrudr.jrnl_mo_yr, 3,4) = @yr_2 ) OR
( SUBSTRING (a_ovrudr.jrnl_mo_yr, 3,4) = @yr_3 ) OR
( SUBSTRING (a_ovrudr.jrnl_mo_yr, 3,4) = @yr_4 ) OR
( SUBSTRING (a_ovrudr.jrnl_mo_yr, 3,4) = @yr_5 ) )
GROUP BY a_ovrudr.acna, r_rj.juris_id, r_rj.ord
ORDER BY yr, mo, r_rj.ord

Is my approach correct, or do I need to add a WHERE condition to it? I don't have access to the production database to check. The development database is in SQL Server 2012 too, sO I will not be able to run the old version there to check. It would be really kind, if someone could help me out with this!

Qlemo

Generally you should move anything not directly related to joins into a WHERE clause. For inner joins it does not matter, but outer joins can change their meaning.

Further you should write

(( SUBSTRING (a_ovrudr.jrnl_mo_yr, 3,4) = @yr_1 ) OR   
( SUBSTRING (a_ovrudr.jrnl_mo_yr, 3,4) = @yr_2 )  OR  
( SUBSTRING (a_ovrudr.jrnl_mo_yr, 3,4) = @yr_3 ) OR  
( SUBSTRING (a_ovrudr.jrnl_mo_yr, 3,4) = @yr_4 ) OR  
( SUBSTRING (a_ovrudr.jrnl_mo_yr, 3,4) = @yr_5 ) )

Open in new window

SUBSTRING (a_ovrudr.jrnl_mo_yr, 3,4) in (@yr_1, @yr_2, @yr_3 , @yr_4, @yr_5 )

Open in new window

for obvious reasons.

Your first transformation is not correct. You are using the result of the "nullable" table in a restriction ( a_icc.usg_ind = 'S' ) outside of the outer join condition, and that makes it an inner join. You should use the following code instead; and it is also a good idea to apply the outer join last to keep away from more issues:

SELECT  DISTINCT  
   a_stmt.cyc_dte,   
   a_carr.carr_nm,   
   a_stmt.amt_du_ic  
from a_stmt
join  a_carr on a_stmt.acna = a_carr.acna
join  s_dataccs on a_stmt.acna = s_dataccs.acna
left join a_icc on a_stmt.ic_cntct_id = a_icc.ic_cntct_id  and  a_icc.usg_ind = 'S'
where
   ( s_dataccs.user_id = @user_id ) and   
   ( ( a_stmt.acna = @acna ) or ( @acna = '' ) ) AND    
   ( ( a_stmt.juris_id = @juris_id ) or ( @juris_id = '' ) ) AND    
   ( a_stmt.jrnl_mo_yr = @jrnl_mo_yr ) AND  
   ( a_stmt.amt_du_ic < 0 )

Open in new window

(you can leave out INNER and OUTER as you can see, and I prefer to do so for both).

On another note, right outer joins are difficult to read. It is much easier to think about a query using left joins, just more natural. But the second query is correct, it needs to have the restrictions in the ON clause because all restrictions are against the "nullable" table, But: Is it really correct you do not apply any restriction against the outer table? This leads to getting the complete r_rj table content, and just joined additional data to it if found and appliable.

kanagu ramasamy

ASKER

Hi Qlemo! First of all thanks for your detailed explanation! I was down with cold and fever for the past two days so I was not able to respond :(

I have completed most of my procedures having two tables..thanks for your explanation .. But in your explanation you have not mentioned the procedure for multiple interconnected tables.

 SELECT a_trueup.jrnl_mo_yr,     
      SUM(a_mthly.tot_acct_rcv_inter) as tot_acct_rcv_inter,     
      SUM(a_mthly.tot_acct_rcv_intra) as tot_acct_rcv_intra,     
  FROM a_mthly,     
   a_trueup,     
   a_bac,     
   a_bdfctr,  
   #tmp_purch_mo_yr    
  WHERE ( a_trueup.acna *= a_bac.acna) and    
      ( a_trueup.ba_cntct_id *= a_bac.ba_cntct_id) and    
      ( a_trueup.acna = a_bdfctr.acna ) and    
      ( a_trueup.juris_id = a_bdfctr.juris_id ) and    
      ( a_mthly.acna = a_trueup.acna ) and    
      ( a_mthly.juris_id = a_trueup.juris_id ) and    
      ( a_mthly.jrnl_mo_yr = #tmp_purch_mo_yr.purch_mo_yr ) and    
      ( a_mthly.intrm_ind = 'EOM' ) AND    
      ( a_mthly.ip_typ <= '2' )and    
      ( a_trueup.rlzd_mo_yr = a_bdfctr.jrnl_mo_yr ) and    
      ( ( a_trueup.acna = @acna ) AND    
        ( a_trueup.juris_id = @juris_id ) AND    
        ( a_trueup.jrnl_mo_yr = @jrnl_mo_yr )    
      )

Open in new window

which I have converted as

 
SELECT a_trueup.jrnl_mo_yr,     
      SUM(a_mthly.tot_acct_rcv_inter) as tot_acct_rcv_inter,     
      SUM(a_mthly.tot_acct_rcv_intra) as tot_acct_rcv_intra,     
  FROM a_mthly join a_trueup on a_mthly.juris_id = a_trueup.juris_id,     
  join a_bdfctr on a_trueup.juris_id = a_bdfctr.juris_id and a_trueup.acna = a_bdfctr.acna  and  a_trueup.rlzd_mo_yr = a_bdfctr.jrnl_mo_yr    
   a_bac,     
   a_bdfctr,  
   #tmp_purch_mo_yr    
  WHERE ( a_trueup.acna *= a_bac.acna) and    
      ( a_trueup.ba_cntct_id *= a_bac.ba_cntct_id) and    
      ( a_mthly.acna = a_trueup.acna ) and    
      ( a_mthly.juris_id = a_trueup.juris_id ) and    
      ( a_mthly.jrnl_mo_yr = #tmp_purch_mo_yr.purch_mo_yr ) and    
      ( a_mthly.intrm_ind = 'EOM' ) AND    
      ( a_mthly.ip_typ <= '2' )and    
          
      ( ( a_trueup.acna = @acna ) AND    
        ( a_trueup.juris_id = @juris_id ) AND    
        ( a_trueup.jrnl_mo_yr = @jrnl_mo_yr )    
      )

Open in new window

Here Im not sure how to connect the tables a_trueup and a_bac as I have already connected them through joins.. that is the sequencing of joins is very difficult.. It would be really so kind of you if you could convert this so that I can do the same for the remaining ..

Thanks in advance!

ASKER CERTIFIED SOLUTION

Qlemo

membership

This solution is only available to members.

To access this solution, you must be a member of Experts Exchange.

Start Free Trial

kanagu ramasamy

ASKER

Thanks a lot Qlemo ! that was really helpful!