Documentation and tools for using Platypus with the Rust programming language
Rust:
#![crate_type = "cdylib"]
#[no_mangle]
pub extern "C" fn add(a: i32, b: i32) -> i32 {
a + b
}
Perl:
use FFI::Platypus 2.00;
use FFI::CheckLib qw( find_lib_or_die );
use File::Basename qw( dirname );
my $ffi = FFI::Platypus->new( api => 2, lang => 'Rust' );
$ffi->lib(
find_lib_or_die(
lib => 'add',
libpath => [dirname __FILE__],
systempath => [],
)
);
$ffi->attach( add => ['i32', 'i32'] => 'i32' );
print add(1,2), "\n"; # prints 3
This module provides native Rust types for FFI::Platypus in order to reduce cognitive load and concentrate on Rust and forget about C types. This document also documents issues and caveats that I have discovered in my attempts to work with Rust and FFI.
Note that in addition to using pre-compiled Rust libraries, you can bundle Rust code with your Perl distribution using FFI::Build and FFI::Build::File::Cargo.
The examples in this discussion are bundled with this distribution and
can be found in the examples
directory.
#![crate_type = "cdylib"]
#[no_mangle]
pub extern "C" fn add(a: i32, b: i32) -> i32 {
a + b
}
use FFI::Platypus 2.00;
use FFI::CheckLib qw( find_lib_or_die );
use File::Basename qw( dirname );
my $ffi = FFI::Platypus->new( api => 2, lang => 'Rust' );
$ffi->lib(
find_lib_or_die(
lib => 'add',
libpath => [dirname __FILE__],
systempath => [],
)
);
$ffi->attach( add => ['i32', 'i32'] => 'i32' );
print add(1,2), "\n"; # prints 3
$ rustc add.rs
$ perl add.pl
3
Basic types like integers and floating points are the easiest to pass
across the FFI boundary. The Platypus Rust language plugin (this module)
provides the basic types used by Rust (for example: bool
, i32
, u64
,
f64
, isize
and others) will all work as a Rust programmer would expect.
This is nice because you don't have to think about what the equivalent types
would be in C when you are writing your Perl extension in Rust.
Rust symbols are "mangled" by default, which means that you cannot use
the name of the function from the source code without knowing what the
mangled name is. Rust provides a function attribute #[no_mangle]
which will tell the compiler not to mangle the name, making lookup of
the symbol possible from other programming languages like Perl.
Rust functions do not use the same ABI as C by default, so if you want
to be able to call Rust functions from Perl they need to be declared
as extern "C"
as in this example.
We also set the "crate type" to cdylib
in the first line to tell the
Rust compiler to generate a dynamic library that will be consumed by
a non-Rust language like Perl.
#![crate_type = "cdylib"]
use std::ffi::CStr;
use std::os::raw::c_char;
#[no_mangle]
pub extern "C" fn how_many_characters(s: *const c_char) -> isize {
if s.is_null() {
return -1;
}
let s = unsafe { CStr::from_ptr(s) };
match s.to_str() {
Ok(s) => s.chars().count() as isize,
Err(_) => -2,
}
}
use FFI::Platypus 2.00;
use FFI::CheckLib qw( find_lib_or_die );
use File::Basename qw( dirname );
my $ffi = FFI::Platypus->new( api => 2, lang => 'Rust' );
$ffi->lib(
find_lib_or_die(
lib => 'argument',
libpath => [dirname __FILE__],
systempath => [],
)
);
$ffi->attach( how_many_characters => ['string'] => 'isize' );
print how_many_characters(undef), "\n"; # prints -1
print how_many_characters("frooble bits"), "\n"; # prints 12
$ rustc argument.rs
$ perl argument.pl
-1
12
Strings are considerably more complicated for a number of reasons, but for passing them into Rust code the main challenge is that the representation is different from what C uses. C Uses NULL terminated strings and Rust uses a pointer and size combination that allows NULLs inside strings. Perls internal representation of strings is actually closer to what Rust uses, but when Perl talks to other languages it typically uses C Strings.
Getting a Rust string slice &str
requires a few stems
-
We have to ensure the C pointer is not
NULL
We return
-1
to indicate an error here. As we can see from the calling Perl code passing anundef
from Perl is equivalent to passing inNULL
from C. -
Wrap using
Cstr
We then wrap the pointer using an
unsafe
block. Even though we know at this point that the pointer cannot beNULL
it could technically be pointing to uninitialized or unaddressable memory. Thisunsafe
block is unfortunately necessary, though it is relatively isolated so it is easy to reason about and review. -
Convert to UTF-8
If the string that we passed in is valid UTF-8 we can convert it to a
&str
usingto_str
and compute the length of the string. Otherwise, we return -2 error.
(This example is based on one provided in the Rust FFI Omnibus)
#![crate_type = "cdylib"]
use std::ffi::CString;
use std::iter;
use std::os::raw::c_char;
#[no_mangle]
pub extern "C" fn theme_song_generate(length: u8) -> *mut c_char {
let mut song = String::from("💣 ");
song.extend(iter::repeat("na ").take(length as usize));
song.push_str("Batman! 💣");
let c_str_song = CString::new(song).unwrap();
c_str_song.into_raw()
}
#[no_mangle]
pub extern "C" fn theme_song_free(s: *mut c_char) {
if s.is_null() {
return;
}
unsafe { CString::from_raw(s) };
}
use FFI::Platypus 2.00;
use FFI::CheckLib qw( find_lib_or_die );
use File::Basename qw( dirname );
my $ffi = FFI::Platypus->new( api => 2, lang => 'Rust' );
$ffi->lib(
find_lib_or_die(
lib => 'return',
libpath => [dirname __FILE__],
systempath => [],
)
);
$ffi->attach( theme_song_free => ['opaque'] => 'void' );
$ffi->attach( theme_song_generate => ['u8'] => 'opaque' => sub {
my($xsub, $length) = @_;
my $ptr = $xsub->($length);
my $str = $ffi->cast( 'opaque' => 'string', $ptr );
theme_song_free($ptr);
$str;
});
print theme_song_generate(42), "\n";
$ rustc return.rs
$ perl return.pl
💣 na na na na na na na na na na na na na na na na na na na na na na na na na na na na na na na na na na na na na na na na na na Batman! 💣
The big challenge of returning strings from Rust into Perl is handling the ownership. In this example we have a C API implemented in Rust that returns a C NULL terminated string, but we have to pass it back into Rust in order to deallocate it when we are done.
Unfortunately Platypus' string
type assumes that the callee
retains ownership of the returned string, so we have to get the
pointer instead as an opaque
so that we can later free it.
Before freeing it though we cast it into a Perl string.
In order to hide the complexities from caller of our
theme_song_generate
function, we use a function wrapper to
do all of that for us.
(This example is based on one provided in the Rust FFI Omnibus)
#![crate_type = "cdylib"]
use std::cell::RefCell;
use std::ffi::CString;
use std::iter;
use std::os::raw::c_char;
#[no_mangle]
pub extern "C" fn theme_song_generate(length: u8) -> *const c_char {
thread_local! {
static KEEP: RefCell<Option<CString>> = RefCell::new(None);
}
let mut song = String::from("💣 ");
song.extend(iter::repeat("na ").take(length as usize));
song.push_str("Batman! 💣");
let c_str_song = CString::new(song).unwrap();
let ptr = c_str_song.as_ptr();
KEEP.with(|k| {
*k.borrow_mut() = Some(c_str_song);
});
ptr
}
use FFI::Platypus 2.00;
use FFI::CheckLib qw( find_lib_or_die );
use File::Basename qw( dirname );
my $ffi = FFI::Platypus->new( api => 2, lang => 'Rust' );
$ffi->lib(
find_lib_or_die(
lib => 'keep',
libpath => [dirname __FILE__],
systempath => [],
)
);
$ffi->attach( theme_song_generate => ['u8'] => 'string' );
print theme_song_generate($_), "\n" for 1..10;
$ rustc keep.rs
$ perl keep.pl
💣 na Batman! 💣
💣 na na Batman! 💣
💣 na na na Batman! 💣
💣 na na na na Batman! 💣
💣 na na na na na Batman! 💣
💣 na na na na na na Batman! 💣
💣 na na na na na na na Batman! 💣
💣 na na na na na na na na Batman! 💣
💣 na na na na na na na na na Batman! 💣
💣 na na na na na na na na na na Batman! 💣
For frequently called functions with smaller strings it may make more
sense to keep ownership of the string and just return a pointer. Perl
makes its own copy on return anyway when you use the string
type.
In this example we use thread local storage to keep the CString
until the next call when it will be freed. Since we are using thread
local storage, it should even be safe to use this interface from a
threaded Perl program (although you should probably not be using
threaded Perl).
(This example is based on one provided in the Rust FFI Omnibus)
#![crate_type = "cdylib"]
#[no_mangle]
pub extern "C" fn hello_rust() -> *const u8 {
"Hello, world!\0".as_ptr()
}
use FFI::CheckLib qw( find_lib_or_die );
use File::Basename qw( dirname );
use FFI::Platypus 1.00;
my $ffi = FFI::Platypus->new( api => 1, lang => 'Rust');
$ffi->lang('Rust');
$ffi->lib(
find_lib_or_die(
lib => 'static',
libpath => [dirname __FILE__],
systempath => [],
)
);
$ffi->lib('./libstring.so');
$ffi->attach(hello_rust => [] => 'string');
print hello_rust(), "\n";
$ rustc static.rs
$ perl static.pl
Hello, world!
Sometimes you just want to return a static NULL terminated string from Rust to Perl. This can sometimes be useful for returning error messages.
#![crate_type = "cdylib"]
use std::ffi::CString;
use std::os::raw::c_char;
type PerlLog = extern "C" fn(line: *const c_char);
#[no_mangle]
pub extern "C" fn rust_log(logf: PerlLog) {
let lines: [&str; 3] = ["Hello from rust!", "Something else.", "The last log line"];
for line in lines.iter() {
// convert string slice to a C style NULL terminated string
let line = CString::new(*line).unwrap();
logf(line.as_ptr());
}
}
use FFI::Platypus 2.00;
use FFI::CheckLib qw( find_lib_or_die );
use File::Basename qw( dirname );
my $ffi = FFI::Platypus->new( api => 2, lang => 'Rust' );
$ffi->lib(
find_lib_or_die(
lib => 'callback',
libpath => [dirname __FILE__],
systempath => [],
)
);
$ffi->type( '(string)->void' => 'PerlLog' );
$ffi->attach( rust_log => ['PerlLog'] );
my $perl_log = $ffi->closure(sub {
my $message = shift;
print "log> $message\n";
});
rust_log($perl_log);
$ rustc callback.rs
$ perl callback.pl
log> Hello from rust!
log> Something else.
log> The last log line
Calling back into Perl from Rust is easy, so long as you have the correct types defined. The above Rust function takes a C function pointer. We can crate a Platypus closure object from Perl from a plain Perl sub and pass the closure into Rust.
#![crate_type = "cdylib"]
use std::slice;
#[no_mangle]
pub extern "C" fn sum_of_even(numbers: *const u32, len: usize) -> i64 {
if numbers.is_null() {
return -1;
}
let numbers = unsafe { slice::from_raw_parts(numbers, len) };
let sum: u32 = numbers.iter().filter(|&v| v % 2 == 0).sum();
sum as i64
}
use FFI::Platypus 2.00;
use FFI::CheckLib qw( find_lib_or_die );
use File::Basename qw( dirname );
my $ffi = FFI::Platypus->new( api => 2, lang => 'Rust' );
$ffi->lib(
find_lib_or_die(
lib => 'slice',
libpath => [dirname __FILE__],
systempath => [],
)
);
$ffi->attach( sum_of_even => ['u32*', 'usize'] => 'i64' );
print sum_of_even(undef, 0), "\n"; # print -1
print sum_of_even([1,2,3,4,5,6], 6), "\n"; # print 12
$ rustc slice.rs
$ perl slice.pl
-1
12
A Rust slice is a pointer to a chunk of homogeneous data, and the number of elements in the slice. We can pass these two pieces in from Perl and combine them into a slice in Rust.
This example sums the even numbers from a slice and returns the result.
(This example is based on one provided in the Rust FFI Omnibus)
#![crate_type = "cdylib"]
use std::convert::From;
// A Rust function that accepts a tuple
fn flip_things_around_rust(tup: (u32, u32)) -> (u32, u32) {
let (a, b) = tup;
(b + 1, a - 1)
}
// A struct that can be passed between C and Rust
#[repr(C)]
pub struct Tuple {
x: u32,
y: u32,
}
// Conversion functions
impl From<(u32, u32)> for Tuple {
fn from(tup: (u32, u32)) -> Tuple {
Tuple { x: tup.0, y: tup.1 }
}
}
impl From<Tuple> for (u32, u32) {
fn from(tup: Tuple) -> (u32, u32) {
(tup.x, tup.y)
}
}
// The exported C method
#[no_mangle]
pub extern "C" fn flip_things_around(tup: Tuple) -> Tuple {
flip_things_around_rust(tup.into()).into()
}
use FFI::Platypus 2.00;
use FFI::CheckLib qw( find_lib_or_die );
use File::Basename qw( dirname );
my $ffi = FFI::Platypus->new( api => 2, lang => 'Rust' );
$ffi->lib(
find_lib_or_die(
lib => 'tuple',
libpath => [dirname __FILE__],
systempath => [],
)
);
package Tuple;
use FFI::Platypus::Record;
use overload
'""' => sub { shift->as_string },
bool => sub { 1 }, fallback => 1;
record_layout_1($ffi, qw(
u32 x
u32 y
));
sub as_string {
my $self = shift;
sprintf "[%d,%d]", $self->x, $self->y;
}
package main;
$ffi->type('record(Tuple)' => 'tuple_t');
$ffi->attach( flip_things_around => ['tuple_t'] => 'tuple_t' );
print flip_things_around(Tuple->new(x => 10, y => 20)), "\n";
$ rustc tuple.rs
$ perl tuple.pl
[21,9]
Rust's tuples do not have a standard representation that can be used
directly from Perl, but if your tuple contains only simple types you
can use the Platypus Record class and translate
in Rust between the tuple and the struct
.
Because we are passing in and out the entire struct
, not pointers
to a struct
we don't have to worry about freeing them from Perl.
They just get allocated and freed on the stack.
(This example is based on one provided in the Rust FFI Omnibus)
use std::cell::RefCell;
use std::ffi::c_void;
use std::ffi::CStr;
use std::ffi::CString;
use std::os::raw::c_char;
struct Person {
name: String,
lucky_number: i32,
}
impl Person {
fn new(name: &str, lucky_number: i32) -> Person {
Person {
name: String::from(name),
lucky_number: lucky_number,
}
}
fn get_name(&self) -> String {
String::from(&self.name)
}
fn set_name(&mut self, new: &str) {
self.name = new.to_string();
}
fn get_lucky_number(&self) -> i32 {
self.lucky_number
}
}
type CPerson = c_void;
#[no_mangle]
pub extern "C" fn person_new(
_class: *const c_char,
name: *const c_char,
lucky_number: i32,
) -> *mut CPerson {
let name = unsafe { CStr::from_ptr(name) };
let name = name.to_string_lossy().into_owned();
Box::into_raw(Box::new(Person::new(&name, lucky_number))) as *mut CPerson
}
#[no_mangle]
pub extern "C" fn person_name(p: *mut CPerson) -> *const c_char {
thread_local!(
static KEEP: RefCell<Option<CString>> = RefCell::new(None);
);
let p = unsafe { &*(p as *mut Person) };
let name = CString::new(p.get_name()).unwrap();
let ptr = name.as_ptr();
KEEP.with(|k| {
*k.borrow_mut() = Some(name);
});
ptr
}
#[no_mangle]
pub extern "C" fn person_rename(p: *mut CPerson, new: *const c_char) {
let new = unsafe { CStr::from_ptr(new) };
let p = unsafe { &mut *(p as *mut Person) };
if let Ok(new) = new.to_str() {
p.set_name(new);
}
}
#[no_mangle]
pub extern "C" fn person_lucky_number(p: *mut CPerson) -> i32 {
let p = unsafe { &*(p as *mut Person) };
p.get_lucky_number()
}
#[allow(non_snake_case)]
#[no_mangle]
pub extern "C" fn person_DESTROY(p: *mut CPerson) {
unsafe { Box::from_raw(p as *mut Person) };
}
#[cfg(test)]
mod test;
Main class:
package Person;
use strict;
use warnings;
use FFI::Platypus 2.00;
our $VERSION = '2.00';
my $ffi = FFI::Platypus->new( api => 2, lang => 'Rust' );
# use the bundled code as a library
$ffi->bundle;
# use the person_ prefix
$ffi->mangler(sub {
my $symbol = shift;
return "person_$symbol";
});
# Create a custom type mapping for the person_t (C) and Person (perl)
# classes.
$ffi->type( 'object(Person)' => 'person_t' );
$ffi->attach( new => [ 'string', 'string', 'i32' ] => 'person_t' );
$ffi->attach( name => [ 'person_t' ] => 'string' );
$ffi->attach( rename => [ 'person_t', 'string' ] );
$ffi->attach( lucky_number => [ 'person_t' ] => 'i32' );
$ffi->attach( DESTROY => [ 'person_t' ] );
1;
Test:
use Test2::V0;
use Person;
my $plicease = Person->new("Graham Ollis", 42);
is $plicease->name, "Graham Ollis";
is $plicease->lucky_number, 42;
$plicease->rename("Graham THE Ollis");
is $plicease->name, "Graham THE Ollis";
done_testing;
$ prove -lvm t/basic.t
t/basic.t ..
# Seeded srand with seed '20221023' from local date.
ok 1
ok 2
ok 3
1..3
ok
All tests successful.
Files=1, Tests=3, 0 wallclock secs ( 0.02 usr 0.00 sys + 0.19 cusr 0.05 csys = 0.26 CPU)
Result: PASS
This example includes excerpts from a full Person
dist which you can
find in the examples/Person
directory of this distribution. You can
install it like a normal Perl distribution using ExtUtils::MakeMaker,
or you can simply run the test file by using App::Prove. That is
because we are using FFI::Build and FFI::Build::File::Cargo to
build the Rust parts for us, which know how to work in either mode.
There are some stuff that we don't show you here for brevity: the
Makefile.PL
for example, and also the rust tests in ffi/src/test.rs
which test the Rust crate by calling both its Rust and C interface.
What we have done here is created a Rust struct
and then written
C wrappers to create, query and modify the object. We've also created
a destructor to free the object when we are done with it.
In terms of naming conventions, we use person_
prefix to denote that
these are methods for the Person class that we are creating. This is
a common convention in C, where the only namespaces are adding prefixes
like this. We also break the convention of using snake case for the
destructor person_DESTROY
because that will make it easier to bind
to from Perl.
When we creat the object we use Box::new
and Box::into_raw
to
create the object on the heap, and to return the opaque pointer back
to Perl.
For methods we can convert the raw pointers back into a Person struct
using &*(p as *mut Person)
inside an unsafe
block. In the case
of person_rename
we need a mutable version so we use &mut *(p as *mut Person)
instead.
Finally when we are done with the object we can free it by simply
calling Box::from_raw
. When it falls out of scope it will be freed.
On the Perl side, we use the mangler
method to prepend all symbols
with the person_
prefix, so that we can attach with just the method
name.
We also create a Platypus type for object(Person)
and give it the
alias person_t
. Now we can use it as an argument and return type.
This is really a pointer to an opaque (to perl) struct
.
If you look at just the test, then you can't even tell that the implementation for our Person class is in Rust, which is good because your users shouldn't have to care!
#![crate_type = "cdylib"]
use std::panic::catch_unwind;
fn might_panic(i: u32) -> u32 {
if i % 2 == 1 {
panic!("oops!");
}
i / 2
}
#[no_mangle]
pub extern "C" fn oopsie(i: u32) -> i64 {
let result = catch_unwind(|| might_panic(i));
match result {
Ok(i) => i as i64,
Err(_) => -1,
}
}
use FFI::Platypus 2.00;
use FFI::CheckLib qw( find_lib_or_die );
use File::Basename qw( dirname );
my $ffi = FFI::Platypus->new( api => 2, lang => 'Rust' );
$ffi->lib(
find_lib_or_die(
lib => 'panic',
libpath => [dirname __FILE__],
systempath => [],
)
);
$ffi->attach( oopsie => ['u32'] => 'i64' );
print oopsie(5), "\n"; # -1
print oopsie(10), "\n"; # 5
$ perl panic.pl
thread '<unnamed>' panicked at 'oops!', panic.rs:7:9
note: run with `RUST_BACKTRACE=1` environment variable to display a backtrace
-1
5
Be cautious about code that might panic!
. A panic!
across the FFI
boundary is undefined behavior and usually results in a crash. You will
want to catch the panic with a catch_unwind
and map to an appropriate
error result. In this example, we have a function that returns the
integer passed in divided by 2. It does not like odd numbers though and
will panic. So we catch the panic and return -1 to indicate an error.
As you can see from the run we also get a rather ugly diagnostic, but
at least our program didn't crash!
Generally you will not use this class directly, instead interacting with the FFI::Platypus instance. However, the public methods used by Platypus are documented here.
my $hashref = FFI::Platypus::Lang::Rust->native_type_map;
This returns a hash reference containing the native aliases for the Rust programming languages. That is the keys are native Rust types and the values are libffi native types.
-
The
bool
typeAs of this writing, the
bool
type is in practice always a signed 8 bit integer, but this has not been guaranteed by the Rust specification. This module assumes that it is asint8
type, but if that ever changes this module will need to be updated.
-
The Core Platypus documentation.
-
Bundle Rust code with your FFI / Perl extension.
-
Includes a number of examples of calling Rust from other languages.
-
The Rustonomicon - Foreign Function Interface
Detailed Rust documentation on crossing the FFI barrier.
-
The Rust Programming Language - Unsafe Rust
Unsafe Rust in the Rust Programming Language book.
Author: Graham Ollis [email protected]
Contributors:
Andrew Grangaard (SPAZM)
This software is copyright (c) 2015-2022 by Graham Ollis.
This is free software; you can redistribute it and/or modify it under the same terms as the Perl 5 programming language system itself.