Every now and then, SQLJ pops up somewhere, mostly in a very dusty/enterprisey or in an academic context.
If you give SQLJ some thought, though, it isn’t such a bad idea. It is:
- An ANSI and ISO standard
- Part of the SQL standard
- Quite easy to understand
- Quite a powerful extension to JDBC
So why did it die (or rather, why did it never really take off)? This question was asked on Stack Overflow, and I gave an answer.
Let’s assume that you have already decided to embed your SQL (as opposed to externalising it through a templating mechanism, hiding it with an ORM, or with stored procedures). Here are a couple of reasons why SQLJ is not an optimal solution for embedding SQL:
While Pro*C worked well for C and C++ in the 90s, Java really took off in the early 2000’s. With Java, there were also an increasing number of powerful IDEs such as Eclipse, NetBeans, JBuilder, and others. Java preprocessors and IDEs have never become friends, though, as parsing one language is hard enough. Parsing (and providing tooling) for two languages is much harder.
In fact, SQLJ made the surrounding Java code type-unsafe as IDEs and compilers couldn’t process .sqlj files before they had been pre-processed.
There was a time when people started thinking that SQL itself was going to be dead. First, they did so with the advent of ORMs, then they did so with the advent of NoSQL. People thought that the DBA is dead. Again.
Well, this has been proven to be wrong a couple of times, but certainly not because of SQLJ.
In the late 2000’s, there had now been typesafe alternatives to SQLJ, such as jOOQ in Java, or LINQ-to.SQL in .NET, which leverage IDE features such as syntax autocompletion. By being internal domain-specific languages / query DSLs, these APIs not only bring typesafety to embedded SQL, but they also allow for dynamic SQL building, which SQLJ doesn’t support.
While embedding SQL into other languages is a useful thing, SQLJ never solved this problem adequately. Hence, R.I.P., SQLJ