──────────────────────────────────
Oryx — Programming Made Better
──────────────────────────────────
Oryx is named after the oryx animal. This means that when referring to
Oryx the programming language in languages other than English you should
use the given language’s translation of the animal’s name (e.g. ‘Órix’ in
Portuguese or ‘Όρυξ’ in Greek) as opposed to using the English name.
──────────────────────
Build Instructions
──────────────────────
Building the Oryx compiler is rather trivial. The steps are as follows:
1. Install the LLVM libraries and -headers. They should be available
through your systems package manager. Do note that as of 17/07/2024
the version of LLVM being utilized is 18.1.x. The compiler may work
with other versions, but it isn’t guaranteed.
2. Install Gperf. It should be available in your systems package
manager, and if not then you can easily find instructions online to
build from source. Gperf 3.1 is the oldest version that is actively
tested on.
3. Clone the compiler repository.
$ git clone https://github.com/Mango0x45/oryx.git
4. Bootstrap and run the build script.
$ cc -o make make.c
$ ./make # See below for more details
If you followed the above steps, you should find the compiler located in
the root directory of the git repository under the name ‘oryx’.
The build script takes a few optional parameters that might be of
interest. They are as follows:
-F Force rebuild the compiler and its dependencies in vendor/.
-f Force rebuild the compiler but not its dependencies in vendor/.
-r Build a release build with optimizations enabled.
-S Do not build with the GCC sanitizer. This option is not required
if -r was specified.
The build script also accepts some subcommands. They are as follows:
clean Delete all build artifacts and compiled binaries.
distclean Delete all build artifacts and compiled binaries, as well
as those creates by any dependencies in vendor/.
test Run the tests in test/. This subcommand should only be
run after a regular invocation of the build script so
that the tests get compiled.
──────────────────────────────
Existing Language Features
──────────────────────────────
1. The following datatypes are supported. The unsized integer types
default to the systems word size (typically 64 bits). The rune type
is an alias for the i32 type and serves a purely semantic purpose.
In the future it will be a distinct type.
/* Integer types */
i8, i16, i32, i64, i128, int
u8, u16, u32, u64, u128, uint
rune
/* Floating-point types */
f16, f32, f64, f128
2. C-style block comments. Line comments are intentionally not
included.
3. Declaration of mutable variables with optional type-inference. The
syntax is simple and consistent regardless of if type-inference is
used or not. Variables are also zero-initialized unless ‘…’
(U+2026 HORIZONTAL ELLIPSIS) or ‘...’ is given as a value.
x: int; /* Declare a zero-initialized integer */
x: int = 69; /* Declare an integer and set it to 69 */
x: = 69; /* Same as above but infer the type */
x := 69; /* Recommended style when inferring types */
x: int = …; /* Declare an uninitialized integer */
x: int = ...; /* Same as above when Unicode is not possible */
When declaring an uninitialized variable, the recommended style is to
use U+2026 HORIZONTAL ELLIPSIS. If you cannot bind that codepoint to
your keyboard, you should investigate the key-remapping faculties of
your text editor. For example, (Neo)Vim users may try the following:
inoremap ... …
" or if you don’t like the above…
inoremap <C-.> …
4. Declaration of constant variables with optional type-inference
including constants of arbitrary precision. The syntax is
intentionally designed to be consistent with mutable variable
declaration.
Constants are unordered, meaning that a constant may refer to another
constant that is declared later in the source file.
FOO: u8 : BAR
BAR: u8 : 69;
REALLY_BIG :: 123'456'789'876'543'210;
pub my_func :: () int {
return BAR;
}
5. Constants of arbitrary precision (overflow is not possible), with ‘'’
(U+0027 APOSTROPHE) as an optional digit seperator.
REALLY_BIG :: 123'456'789'876'543'210;
6. No implicit type conversions between types. This includes between
different integer types which may have the same size (i.e. int and
int64)
pub my_func :: () {
x: int = 69;
y: i64 = x; /* Compile-time error */
}
7. Nested functions are supported, but not closures. Closures will
never be supported in the language.
/* Recall that constants (including functions!) can be declared
in any order. This lets us define inner *after* it gets
called by the assignment to ‘x’. */
outer :: () {
x := inner(5);
inner :: (x: int) int {
return x;
}
}
8. No increment/decrement operators. The following functions both
return 42 as the return values are parsed as (+ (+ 42)) and
(- (- 42)) respectively.
x := 42;
returns_42 :: () int {
return ++x;
}
returns_42′ :: () int {
return --x;
}
9. Assignment statements (not expressions). Unlike in C, you cannot put
an assignment inside of an expression.
return_42 :: () int {
x := 4;
y := 2;
x = x*10 + y;
return x;
}
Due to quirks of the language grammar identifiers may be wrapped in
(arbitrary levels of) parenthesis, however assignments are only
permitted if the left-hand side with parenthesis removed is a lone
identifier. The rationale behind this is to allow in the future
assignments to expressions that return pointers.
((x)) = x*10 + y; /* legal */
(true ? x : y) = x*10 + y; /* illegal */
10. Static local variables allow for block-scoped global variables. This
is useful for having function state persist across multiple calls.
iota :: () int {
static x := -1;
x = x + 1;
return x;
}
pub main :: () {
zero := iota();
one := iota();
two := iota();
three := iota();
}