[JonChesterfield]

m4 is like that. Lots of templating languages are too. Code generation is useful and roughly what you end up with if the language isn't expressive enough, e.g. C or go. Is yours doing anything different to those?

What the lisp approach gives you is the macro has the parsed representation of the program available to inspect and manipulate. It doesn't have to expand to some fixed text, it can expand into different things based on the context of the call site.

Various languages have decided that syntactically distinguishing macros from functions is a good idea. Maybe the function is foo and a macro would be $foo. For what it's worth I think that's wrong-headed, and the proper fix is to have 'foo' represent some transform on its arguments, where it doesn't matter to the caller whether that is by macro or by function, but it's certainly popular.

[hzhou321]

M4 only do token-level macros or inline macros. M4 macros are identifier that can't be distinguished from underlying language. M4 macros does not have scopes. M4 does not have context level macros. M4 does not have full programmable ability to extend syntax.

I can define any macro lisp can define, just not with LISP syntax. I do not have a full AST view of the code, due to the its generality that it does not marry to any specific underlying language. But I can have extensions that is tailored to specific language and do understand the syntax. For example, the C extension can check existing functions and inject C functions. MyDef always can query and obtain the entire view of the program, and it is up to the effort in writing extensions to which degree we want macro layer to be able to parse. Embedding a AST parser for a local code block is not that difficult.

It's like the innerHTML thing, for me, I always find the text layer (as string) is more intuitive for me to manipulate than an AST tree. If needed, make an ad-hoc parser in Perl is often simple and sufficient, for me at least.