3 ms·
He's really wrong about LISP macros. Would suggest pg takes a look at camlp4. Edit: maybe instead of downvotes, you could reply explaining what's wrong with t
by rwmj 15y ago
He's really wrong about LISP macros. Would suggest pg takes a look at camlp4.
Edit: maybe instead of downvotes, you could reply explaining what's wrong with this position. Or just look at camlp4 and see how it provides macros that are better (with a better underlying language) than LISP. And yes, I've written a LISP compiler.
- zedshaw 15y agoYou questioned PG and LISP in one comment then advocated another language with a lesser Nerd Lore Composite Score. About the only thing you could have done worse is to admit you wrote a child porn site in both and found LISP's parentheses to be annoying. :-)
- Goladus 15y agoPG's assertions are often hard to question, probably because he studied philosophy and is very good at logic. He doesn't say it's impossible and his definition of lisp is very inclusive. I read some of the docs for camlp4. It looks like a library for extending ocaml using syntax grammars, similar to Langscape(formerly EasyExtend) for Python. It's not clear whether that really addresses PG's point (a). It would be really nice if rwmj would make his point explicit instead of whining about downvotes and telling people to go slog through camlp4's incomplete, broken-english documentation.
- biot 15y agoThe downvotes are likely because your position consists of little more than "google why he's wrong". If you wish to refute pg's statement you should explain what's right about your position, such as why camlp4's macros aren't just a prettier syntax on top of a LISP-like abstract syntax tree as the Wikipedia article states.
- colanderman 15y agoEasy reason: camlp4's macros are typechecked. Meaning, it's impossible to build an invalid syntax tree using them.
- Robin_Message 15y agoMy personal opinion is that a type system on top of macros needs types that are too painful for anyone to actually use. The winning element of Lisp macros is that they are dirt simple and untyped, which makes them easier to write (but makes run-time errors more likely). At a meta level, you might as well argue that Python programmers should be using ML instead, since the features are similar but ML is statically typed so more errors can be found at compile-time. The logic of the argument is correct, and yet Python (and Ruby and Javascript and so on) win. Again, I think its to do with easy of expression and less thought up front, but I'm not really sure.