There may be a bunch of other reasons, but the one that springs to mind is localisation of errors. You'd ideally like to know about any mistakes you've made in your macro before you run your macro. And when you do learn about a mistake you'd like the error to describe where the mistake is in macro-land rather than generated-code land.

Why just type errors? You might want to trace any error that occurs at runtime back to the original source code anyway.

I agree, you may wish to do that but I was answering the question "What is the utility of typed-templates", not "contrast type and runtime error tradeoffs".

Firstly, if your macro does not typecheck, it means that executing it might result in a segfault or something similar, since macros are regular OCaml functions. Secondly, if macro typechecks, then it is guaranteed to expand without errors, and I think that's a nice guarantee to have.

> Firstly, if your macro does not typecheck, it means that executing it might result in a segfault or something similar, since macros are regular OCaml functions.

I see. The only time I've extensively used (real) macros has been with racket, where they aren't regular functions at all. I suppose my ignorance of Ocaml macros is showing here.

Look at the C++ templates, I think that some of the clusterfuck that is C++ template compiler errors would go away if templates were type aware.