Question

Sorry, I am pretty much an SQL noob. This has to work in MSFT SQL, Oracle as well as Sybase. In the following snippet I need to change an inner join between IJ and KL on IJ.PO_id = KL.PO_id into a left join also on IJ.PO_id = KL.PO_id. So, I believe I have to re-factor this. Well, implicit joins are not the most readable, at least in my co-worker’s eyes. I guess I will agree until I develop my own taste. Sorry, I mangled the table and field names just in case.

/* @IJ_id is an input stored proc patrameter. */
from AB, 
     CD,
     EF,
     GH,
     IJ,
     KL
where
    EF.EF_id = IJ.EF_id and
    IJ.EF_id = AB.EF_id and
    EF.ZY_id = IJ.ZY_id and
    IJ.ZY_id = AB.ZY_id and
    IJ.IJ_id = AB.IJ_id and
    IJ.IJ_id = @IJ_id and
    EF.XW_id = GH.GH_id and
    AB.VU_code = CD.VU_code and
    IJ.TS > 0 and
    IJ.RQ = 0 and
    EF.RQ = 0 and
    AB.RQ = 0 and
    IJ.PO_id = KL.PO_id;

Now, my difficulty is that there is a lot going on in the where clause. Things that do not look like a.b = c.d will remain in the where clause, but not all stuff that does look like a.b = c.d look easy to convert into an explicit join. The difficult part is that ideally the conditions would be between neighbors – AB+CD, CD+EF, EF+GH, GH+IJ, IJ+KL but they are not that organized right now. I could re-order some, but ultimately I do not want to forget my goal: I want the new query to be no slower, and I want the new query to be no less readable. It seems that I might be better off hacking just the part that I need to change, and leave it mostly the same. I am not sure if I can do that.

If you understood my intent, please suggest a better query. if you did not, then please tell me how I can improve the question. Thanks.

Answer 1

I think it should be something like this:

FROM AB
JOIN CD ON AB.VU_code = CD.VU_code
JOIN IJ ON IJ.EF_id = AB.EF_id AND IJ.ZY_id = AB.ZY_id AND IJ.IJ_id = AB.IJ_id
JOIN EF ON EF.EF_id = IJ.EF_id AND EF.ZY_id = IJ.ZY_id
JOIN GH ON EF.XW_id = GH.GH_id
JOIN KL ON IJ.PO_id = KL.PO_id
WHERE
    IJ.IJ_id = @IJ_id AND
    IJ.TS > 0 AND
    IJ.RQ = 0 AND
    EF.RQ = 0 AND
    AB.RQ = 0

I have tried to arrange the tables such that the following rules hold:

• Every join condition mentions the new table that it joining on one side.
• No table is mentioned in a join condition if that table has not been joined yet.
• Conditions where one of the operands is a constant are left as a WHERE condition.

The last rule is a difficult one – it is not possible to tell from your mangled names whether a condition ought to be part of a join or part of the where clause. Both will give the same result for an INNER JOIN. Whether the condition should be part of the join or part of the where clause depends on the semantics of the relationship between the tables.

You need to consider each condition on a case-by-case basis:

• Does it define the relationship between the two tables? Put it in the JOIN.
• Is it a filter on the results? Put it in the WHERE clause.

Some guidelines:

• A condition that includes a parameter from the user is unlikely to be something that should be moved to a join.
• Inequalities are not usually found in join conditions.