SQL is a very rich and expressive language to query relational databases. But it is awkward in many aspects, and a lot of people perceive it to be evolving only slowly - even if that is not true, considering the pace of SQL standards. But the standard is one thing, the implementations another - especially in the enterprise. When we blog about SQL, we’re constantly surprised ourselves, how awesome the PostgreSQL dialect is. But often, PostgreSQL actually just implements the standard. So there is hope that we’re getting somewhere.
Nonetheless, in Leland’s article, there are a couple of ideas worth picking up. From our point of view, these are mainly:
Flexibility in ordering the
SELECT clause and the table expression
SELECTis always the first keyword. It must be expressed before the table expression. We’ve shown in a previous article that this is quite confusing for many SQL users. While the existing syntax should continue to exist, it would be good to be able to inverse the
SELECTclause and the table expression.
Remember, the table expression contains
GROUP BYclauses, as well as vendor-specific
CONNECT BYclauses and others:
This language feature is already available in LINQ, by the way.
Implicit KEY JOINs
This feature is also available in jOOQ, using the
ON KEYclause. Note that Sybase also supports
This is one of the features we really wish that the SQL language had. However, we wouldn’t count on specifying projections in a dedicated syntax. We had rather use an extension to the table expression syntax, allowing for a table to produce “side-tables” as such:
In the above example,
projectionis really nothing else than another table expression that is derived from the
userstable. From a SQL syntax semantics, this would be extremely powerful, because such projections would inherit all syntactic features of a regular table. We’ve blogged about this before, when we called that feature “common column expressions”.
Leland has lots of other ideas. He’s just at the beginning of a project that will still need a lot of refinement. The feedback he got on reddit, however, is rather good. Clearly, there is a lot of potential in creating “BQL” for SQL what
- less is for CSS
- Groovy is for Java
- Xtend is for Java