• CELKO (9/3/2012)


    You gave us no DDL, so this will be harder than it should. I assume that you want to get rid of this function. Good SQL never use them; they are proprietary, non-relational, screw up the optimizer and scream to the world that the author is still stuck in 1960's BASIC, right down to the “f_”, “fn_”, “udf_” prefixes that were required by FORTRAN II and BASIC. :w00t:

    Since we have VIEWs in SQL, a UDF that returns a table is the worst way to program SQL; but it lets BASIC avoid leaning how to do it right. Write a VIEW.

    We use the DUNS for company identifiers; it is the industry standard. Can you give me a 255 character example of your company_id? When you invite garbage data with absurdly sized columns, you will get it. I also see the magic default VARCHAR(50) declarations from ACCESS.

    Is the company identifier of a client or a supplier or what? In COBOL, which is what you are actually writing, the hierarchical record structure would would qualify the field; but SQL has tables with rows and columns, not records and fielsd.

    I also see that you layout your code as if you were still using punch cards; each parens, data element, etc on a separate card (line). That let us re-arrange the deck on the fly. Today, we use a “pretty printer” and make the code human readable instead.

    CREATE VIEW Count_Somethings_Days (work_date, client_duns)

    AS

    SELECT work_date, client_duns

    FROM dbo.Table1, dbo.Table2 -- not real names!

    WHERE Table1.vague_something = Table2.vague_something

    AND client_duns IN (..);

    Yes, you actually have to hard code the list, if it is what defines the set your want. Yes, I know it worked that way in 1960's BASIC, but SQL is compiled and not interpreted.

    If you want this to change infrequently, put their DUNS in a table and use “client_duns IN (SELECT client_duns FROM Bankrupt_Companies)”in the VIEW. There is also the long parameter list idiom, if you want to make this into a stored procedure.

    My guess is that you are about 2-3 years away from un-learning bad habits and old languages to be an SQL programmer. Keep at it and read everything you can find, starting with an intro to Set Theory and Logic. Eventually, you will get to ISO Standards which are boring as hell. 🙂

    Let's see, your view may take less space, but fails in some regards. The first is the use of the old ANSI-89 style join instead of the newer ANSI-92 style join. I though you were all for using standards? You should consider updating to the newer standard. In my opinion it makes code more readable be separating the JOIN criteria from the filter criteria in the FROM clause.

    Also, there really is nothing wrong with listing each column on its own line, in fact I prefer that as a coding style. Apparently so do many of the routines used to automatically format code. Personally, putting as many columns on one line leads to more difficult to read code.