Andy Powell's Blog

I was reading an old post by a buddy of mine about bad SQL being the cause of a lot of application problems and decided to weigh in my two cents.

A lot of developers get forced into writing SQL as part of their jobs. Should they be doing it? I don't think so. It's not necessarily the best of ideas, and in MOST cases should probably be avoided at all costs.

Besides, developers cannot be experts in every language or technology right? Something has got to give somewhere. It's usually their SQL skills that suffer. Developers are, sometimes, forced into situations where they have no choice but to write their own SQL. There is either no DBA, a DBA who isn't interested in helping developers with their queries, or a DBA who isn't even in the development loop (never a good sign). In these cases, developers may have to write their own SQL.

Sometimes developers have to know their limitations when it comes to writing queries, especially complex queries. I don't think a lot of developers do truly know their skill limitations. Yet, these intrepid souls will trudge on thinking they can write SQL just fine. When, in reality, they really and truly do not know the little tricks and tweaks that can make the SQL perform better.

I've seen it a thousand times. It's not an indictment of the developer, just a limitation of their skills that needs to be recognized.

Don't lose hope though, there are ways to combat this and make your SQL as good as it should be, or as good as it can be. What are these answers, then? Personally, I opt for something like Hibernate (Java) or Transfer (ColdFusion) to abstract my SQL for me. What does this mean? It means that my queries will be optimized and I can spend my time focused on developing the business logic rather than spending that time mired in persistence.

It also gives me the agility of being (most of the time) database agnostic. This lets us easily develop in one database locally, and port the same code to the server on a different database with minimal configuration changes. If ORMs are not an option for you and you have access to a good DBA, engage them. Talk to them and mine them for all the information they're worth. It's their job to know the database you're using and how to optimize queries. Make them earn their money by at least helping you write your queries, if not transferring the database functionality to them altogether (good luck).

None of these two options are viable? Well maybe you'll have to engage an outside consultancy to help you with your persistence layer optimizations, or pursue another path for optimizing your queries that has not been mentioned here. The bottom line is this: If you optimize your persistence today, then you'll have to spend less time dealing with it later when it causes you problems tomorrow. You may be able to write SQL, you may be able to do complex joins and create some wicked stored procedures, but are you a SQL expert? If so, then never mind this post. If not, then don't be afraid to swallow your pride and ask for help.