Exceptions Should Not Contain Business Logic

Steven Feuerstein’s PL/SQL book has a remarkably good chapter about collections. In addition to showing the different kinds of collections, their methods and how to use them, he also shows some cool non-trivial uses for collections, a taste of the great things you can do.
Few of the examples involve the use of collections for caching. Associative arrays are perfect for caching, and caching is a useful programing technique. One of the examples is as follows (The code is available on Oreilly’s site, so I don’t think I’m breaking copyright here):

   FUNCTION description (code_in IN hairstyles.code%TYPE)
      RETURN hairstyles.description%TYPE
      return_value   hairstyles.description%TYPE;

      FUNCTION desc_from_database RETURN hairstyles.description%TYPE
         CURSOR desc_cur IS
            SELECT description FROM hairstyles WHERE code = code_in;
         desc_rec   desc_cur%ROWTYPE;
         OPEN desc_cur;
         FETCH desc_cur INTO desc_rec;
         RETURN desc_rec.description;
      RETURN descriptions (code_in);
         descriptions (code_in) := desc_from_database;
         RETURN descriptions (code_in);

This code is short, simple to understand, can be adapted for many different uses, and is an excellent demonstration of the power of associative arrays.

I have just one problem with this code: Most of its logic is implemented inside the exception clause.

Technically speaking, exceptions are very similar to “goto” and historically goto has been used for exactly the same purpose – get out of the current block when things have gone wrong. Exceptions (and goto) make the code easier to read by putting all the clean up work in one place and jumping to this one place whenever needed, but other uses of exceptions can make the code more confusing.

The development standards I’m working with specify that exceptions should be used for handling exceptional conditions, not situations that will occur during the normal flow of the program. The advantage of this standard is that when someone reads the code, she can understand the main logic without jumping back and forth between different blocks. Then when she gets to the exception block she can see which problems you anticipated and how you dealt with those.
How do we know if we did this right? We look at the code and ask ourselves what would happen if we deleted the exception block and then ran it in ideal circumstances (enough memory, disk space, smart users, etc) – would it behave the same from business logic perspective? It should.

So, I would rewrite the function above as follows:

    If  (not descriptions.exists(code_in)) then
        descriptions (code_in) := desc_from_database;
    End if;
    RETURN descriptions (code_in);

I’m not saying that Steven’s original example is confusing. It is not. I enjoyed the smart use of exceptions. But I know from bitter experience that examples have ways of propagating, and beginners that see this example in the book may copy this bit without thinking about the finer points of code readability.

If you are not sure what belongs in the exception clause and what doesn’t, keep your business logic out of it. Even if an example in your favorite book is doing the opposite. 99% of the time, your code will be more readable.


Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s