Recursive descent parsing was probably the first “real” algorithm that I learned (in the 1980s).
Like most of these, I’ve never actually seen one “in the wild,” but it was a great example of the power of recursion.
Like most of these, I’ve never actually seen one “in the wild,” but it was a great example of the power of recursion.
But, like I said, the idea behind it was quite useful, to me. I use recursion all the time, and RDPs are a pretty "pure" form of recursion.
Is there a reason for that? None that I can think of. SQL systems tend to have worse error messages but I don't believe parser generators require error messages to be as bad as SQL error messages (I'm talking about how they often don't give line number or column number info).
So my guess (and it sounds kind of absurd to say) is that most SQL systems just don't put much thought into parsing. And not that parser generators fit the domain better.
Also, recursion can be very bad for parallel stuff, if not done carefully.
We used to have to do a lot of optimization, in my old job, involving use of GPUs, ISPs, multi-thread, etc.
In that kind of environment, things can get quite crazy, and seemingly innocuous stuff, can have huge knock-on effects.
For example, cache hits. If you want your code to stay in a shallow cache (which could result in 100X improvement in speed), you may do things like write something over and over, inline, instead of calling a method or subroutine.
The original C++ compiler was (after horrible experiences with YACC, or similar) writen mostly as a recursive descent compiler.