For a collection of useful macros, see the associated Magma project
Macros are written directly in the source, and the cmc program is used to
process a file with macros to a macroexpanded file.
cmc code_with_macros.c -o macroexpanded_code.cIf you're running Arch or a similarly bleeding-edge distro, just install sbcl
from Pacman and skip to step 5. Otherwise, you need to manually download the
latest SBCL[1].
- Download SBCL
- Unpack it, for example, through
bzip2 -cd sbcl-1.1.17-x86-linux-binary.tar.bz2 | tar xvf - - Install git, curl and flex through your favorite package manager.
- Build SBCL:
cd <sbcl dir>; sudo sh install.sh - Build cmacro:
make,sudo make install
[1]: Buildapp doesn't work on older versions of SBCL, and it is required to build the executable.
A macro is a function that operates on your code's abstract syntax tree rather than values. Macros in cmacro have nothing to do with the C preprocessor except they happen at compile time, and have no knowledge of run-time values.
In cmacro, a macro maps patterns in the code to templates. A macro may have multiple cases, each matching multiple patterns, but each producing code through the same template.
Macros are not primarily about safety and performance: They are about the programmer. Macros give you automation, plain and simple. They allow you to abstract away and remove repetition in places where a functional or object-oriented approach can't. For example, Common Lisp's WITH-OPEN-FILE macro helps with the common pattern of 'acquire a resource, apply something to it, and close it'. While this can be done in languages that support (And have simple syntax for) anonymous functions, macros help reduce this syntactic overhead.
cmacro has a very lenient notion of C syntax, which means you can write macros to implement DSLs with any syntax you like. You could implement Lisp-like prefix notation, or a DSL for routing URLs, or the decorator pattern, for example.
For a very simple example, this macro matches anything of the form unless <cond>, where <cond> is any arbitrary expression, and performs a simple
transformation:
macro unless {
case {
match {
$(cond)
}
template {
if(!$(cond))
}
}
}With this definition, code like unless(buffer.empty) becomes if(!(buffer.empty)).
A more complicated macro can match multiple patterns, like the route macro
which implements a DSL for defining routes in a hypothetical C web framework.
macro route {
/* Route all requests to 'url' to 'route'. Optionally discriminate by HTTP
method (GET by default). */
case {
match {
$(url) => $(route)
}
template {
register_route($(url), $(route), HTTP_GET);
}
}
case {
match {
$(url) [$(method)] => $(route)
}
template {
register_route($(url), $(route), $(method));
}
}
}
// Usage with the lambda macro (See below)
route "/profile/<user>" =>
lambda(Req* request) -> Resp { return Authenticate(request.user); }Because a language without macros is a tool: You write applications with it. A language with macros is building material: You shape it and grow it into your application.
There is a sweet spot between low-level performance and control and high-level metaprogramming that is not yet occupied by any language: Metaprogramming, being an inherently compile-time thing, can be done in the absence of automatic memory management or dynamic typing. Rust seems to want to fill this spot, and I also approached this problem with Corvus, but I feel this approach of adding metaprogramming to C - A simple language, with a long history, that runs truly everywhere - can become useful.
macro lambda {
case {
match {
$(args) -> $(ret) $(body)
}
template {
$(@getsym lambda 0)
}
toplevel {
$(ret) $(@gensym lambda) $(args) $(body)
}
}
}Usage:
/* Input */
fn = lambda (int x, int y) -> int { return x + y; };
/* After macroexpansion */
int cmacro_lambda_0(int x, int y) { return x + y; }
fn = cmacro_lambda_0;A more complicated example, using the qsort function:
int main() {
int array[] = {423, 61, 957, 133, 969,
829, 821, 390, 704, 596};
qsort(array, 10, sizeof(int),
lambda (const void* a, const void* b) -> int
{ return *(int*)a - *(int*)b; });
for(size_t i = 0; i < 10; i++){
printf("%i ", array[i]);
}
return 0;
}This stores the result of the condition in the variable it. See
Anaphora for a collection of similar
anaphoric macros.
macro aif {
case {
match {
$(cond)
}
template {
typeof($(cond)) it = $(cond);
if(it)
}
}
}Usage:
/* Input*/
aif(get_buffer(a,b,c)) {
write_string(it, text);
}
/* After macroexpansion */
typeof(get_buffer(a,b,c)) it = get_buffer(a,b,c);
if(it) {
write_string(it, text);
}macro forEach {
case {
match {
($(item), $(collection)) $(body)
}
template {
{
size_t index;
typeof($(collection)[0]) $(item);
for(index = 0, item = nth($(collection), 0);
index < length($(collection));
index++)
$(body)
}
}
}
}The syntax for variables is just a name followed by an optional,
space-separated list of qualifiers, enclosed in the $() operator, eg:
$(var), $(body ident), $(arg const).
- None: The variable matches any expression.
rest: Match multiple expressions (Like C's...).ident: Matches Identifiers.int: Integers.float: Floats.num: Integers and floats.string: String literals.const: The equivalent of(or int float string).op: Operators.list,array,block: Matches expressions of the form(...),[...],{...}.
These use regular variable syntax but the text starts with a '@'.
gensym <label>: Generates a unique identifier associated withlabel.getsym <label> [n]: Gets the latest identifier associated withlabel, or optionally then-th last identifier.to-string <var>: Since string literals in C can contain variable notation, you have to explicity use this to stringify a variable. Note, also, that C concatenates string that appear next to each other in the source.splice <var>: Ifvaris a block (ie(...),[...],{...}) this expression removes the block separators, leaving just the block body.
The lex and yacc grammars were originally posted by Jeff Lee in 1985, and rescued and updated to the recent standards by Jutta Degener.
The syntax for macro definition was inspired by Mozilla's great sweet.js library. Originally I considered multiple different ways of defining them, including external YAML files or just piping the AST to an XSLT or similar program, but this seemed like the best way.
The Makefile is largely based on that of Dimitri Fontaine's pgloader utility.
Peter Norvig's Paradigms of Artificial Intelligence Programming chapter on Eliza was used as a reference for the pattern-matching engine.
Copyright (c) 2014-2015 Fernando Borretti ([email protected])
Permission is hereby granted, free of charge, to any person obtaining a copy of this software and associated documentation files (the "Software"), to deal in the Software without restriction, including without limitation the rights to use, copy, modify, merge, publish, distribute, sublicense, and/or sell copies of the Software, and to permit persons to whom the Software is furnished to do so, subject to the following conditions:
The above copyright notice and this permission notice shall be included in all copies or substantial portions of the Software.
THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
