[such_a_casual]

Lisp macros do not rely on S-expressions. They rely on the fact that arguments are not evaluated, the code is passed in raw to the function, and then the code is replaced by the result of the function. If macros existed in Python for example, you could do this (sorry if my python isn't 100% correct. haven't used it in a while):

  defmacro exists (code list):
    for x in ,list:
      if ,code:
        return True
    return False

  exists(x > 10, [1 5 333]) => t

Macros are great in Lisp, because lisp makes it easy to code your code. It does this by treating your file of code like you might think of an array or a list or another data structure in another language. This data structure is really simple and is called a list. Lisp provides a bunch of help to make it easy to work with lists, just like Python provides a bunch of help for working with strings, lists, and tuples by giving you functions and variables for manipulating, printing, and debugging these things. Because the code you write is a list and lisp provides a bunch of help for dealing with lists, it's as easy to manipulate a list of code as it is to manipulate a list of songs or emails or zombies like you're probably already accustomed to doing.

In other words, macros are awesome and you could easily put them in another language. Lisp isn't great because of macros, it's great because of how well (mind blowingly well) macros play with with all of the other parts of lisp. Macros are empowered by the rest of the langauge, and the rest of the language is empowered by macros. And this is really the theme of lisp, it's not one feature that makes the language impressive, it's how the features seem to amplify one another.

[bhnmmhmd]

The Python example looks really inspiring, and could possibly be implemented in future versions of Python.

But as for your statement that "macros don't rely on homoiconicity", I find it hard to agree, because if that were the case, then why haven't Java or Python implemented the macro feature by now?

[jwdunne]

Macros don't rely on homoiconicity. The raw power of Lisp macros rely on it. This is because, as you know, lisp code is lisp data. All of the functions for manipulating lists, which form the basis of a lisp language, can be used to build and emit code.

In reality, lisp macros are functions that accept lists and emit lists. The difference is they operate at compile time, so they transform code.

In fact, a common macro implementation pattern is to delegate the actual work to a function that transforms lists, with the macro passing the code as a list and returning the resultant list as expanded code. I imagine this makes macros an order of magnitude easier to test.

In other languages, it's possible and can be done as in Rust etc. It's just far far more natural to do so in a homoiconic language due to the above.

[such_a_casual]

Perl 6 has macros:

Copied from: https://perl6advent.wordpress.com/2012/12/23/day-23-macros/

    macro checkpoint {
      my $i = ++(state $n);
      quasi { say "CHECKPOINT $i"; }
    }

    checkpoint;
    for ^5 { checkpoint; }
    checkpoint;

No "homoiconicity". Just two operators: "macro" which will make a macro instead of a normal function, and "quasi" which says the stuff in here is code that shouldn't be evaluated. In lisp these two operators could be "defmacro" and "quote".

Lisp has a lot of features that other languages haven't implemented. We could spend all day asking why other languages haven't implemented them yet. And it's unfortunate, because just adding the feature for macros would allow you to add a lot of those other features. Indeed, macros are such a big deal that they are probably the sole defining reason that lispers can add features from other languages so easily.

However, we need to stop pretending that code substitution is some magical, genius feature that can only exist in the utopian environment of a common lisp system. It's a simple idea that's been used since before I was born. Macros in lisp are nice because you have so much access to the language, as well as (perhaps more so than any popular language) really smart tools for manipulating the data structure your code is written in. If you have to code code, if you have to write code that manipulates other code, it's nice if that other code is in an easy to understand data structure. And even easier if you already have a bunch of functions, classes, variables, etc given to you by the language for changing that data structure. Which lisp does, and that's what makes lisp lisp.

[bhnmmhmd]

Thank you. I never thought Perl 6 had macros. Plus, now I can see that "homoiconicity" is not required for macro systems.

> However, we need to stop pretending that code substitution is some magical, genius feature that can only exist in the Utopian environment of a common lisp system.

I couldn't agree more.

[dragonwriter]

> Lisp macros do not rely on S-expressions.

They depend on homoiconicity, which S-expressions are the mechanism for in lisp.

[such_a_casual]

No they don't. A macro just takes code and returns code.