A collection of meta-interpreters for your enjoyment

[hurufu]

I've started digging an old gold mine containing a lot of interesting historical Prolog code, and I found small collection of meta-interpreters. I find it useful in case you don't know what else can be done using basic meta-interpretation techniques.

[dnmfarrell]

This book chapter has some basic meta-interpreter examples including allowing the user to override a failed goal, and passing around state so it can be printed at terminal upon request:
https://www.cs.unm.edu/~luger/ai-final2/CH6_Three%20Meta-Intrepeters%20-%20Prolog%20in%20Prolog,%20EXSHELL,%20and%20a%20Planner.pdf

[dnmfarrell]

A layered meta-interpreter is an interesting approach as it allows negation and cut to be handled more easily:
https://www.ijcai.org/Proceedings/89-1/Papers/011.pdf

I typed out the mi code (and fixed a bug

[hurufu]

I've just learned about Futamura Projections – mind blown. It means that if you write an interpreter and have a program that is accepted by it – there is an automated way to compile the program to a faster representation. That means that you can introduce Prolog extensions via meta-interpretation and they can be "compiled" to a fast code without performance hit of the interpreter itself.

[UWN]

Two problems with Futamura Projections:

Most systems barely manage the first projection. Self application (2nd) is too far away. And the third illusory.

The speedups are not what people hoped for. Even more so in Prolog.

But there is another related approach: write the third projection by hand. This has produced a bit more of usable results.