Getting started
A Lyric program is a file with the extension .lyr. Execution begins at main.
import std.io.console { println };
fn main(): int {
println("Hello, Lyric!");
return 0;
}
Run it:
lyric run hello.lyr
The return value of main becomes the process exit code, masked with & 0xFF.
The tools
| Command | Purpose |
|---|---|
lyric run <file> |
compile and execute |
lyric check <file> |
compile without writing a file |
lyric build <file> -o <out> |
compile to .lyrbc |
lyric disasm <file> |
print the bytecode |
lyric repl |
interactive prompt |
In a project — a directory with a lyric.json, which lyric new writes — the same verbs take
no file at all: lyric run, lyric build, lyric check, lyric test and lyric pack answer
for the project. Chapter 16 is about that.
A compile is the debug profile unless you say --release: every frame kept, every name
kept, the shape a backtrace and a debugger read honestly. Chapter 16
has the table.
lyric check goes as far as the intermediate representation, so a construct the code generator
cannot lower is reported there rather than at the build. It stops before the bytes: a module can
pass everything above and still be one the loader refuses. lyric check <file> --emit runs that
last step too — the bytes are produced, loaded and dropped, and no file is written. It is the
answer for a project that compiles each of its files on its own and wants to know they will all
load.
Arguments
To receive command-line arguments, declare main with one parameter of type string[]:
import std.io.console { println };
fn main(args: string[]): int {
println(f"got {args.length} argument(s)");
return 0;
}
Everything after -- on the command line belongs to the program:
lyric run wc.lyr -- README.md