Skip to content

I like to build … Reusable libraries & packages#

Redistributable code with no entry point -- a Python wheel for PyPI, a JavaScript/TypeScript package for npm, or a C-ABI shared library any language can link. The public surface is whatever you mark :pub. These map to the py-package, js-package, and native-lib project kinds.

Your 5-minute quick win#

A reusable library packaged as a standard pip wheel. Any def:pub is part of the public API:

# greetlib.jac
def:pub greet(name: str) -> str { return f"Hello, {name}!"; }

# jac.toml
[project]
name = "greetlib"
version = "0.1.0"

jac build --as wheel
# → dist/greetlib-0.1.0-py3-none-any.whl

Upload it with twine, then pip install greetlib anywhere. The wheel ships your compiled modules and runs under the jac binary -- it does not list jaclang as a runtime dependency.

npm package#

The client-side counterpart: a cl component (or function) library published to npm so any JavaScript/TypeScript project can npm install it. jac build --as npm compiles your client modules to ES-module JavaScript, generates package.json, and emits .d.ts declarations. (Modules that cross a server boundary can't ship as standalone npm packages -- keep server-coupled code in your app.)

Shared library (C ABI)#

The native counterpart: an na module compiled to a C-ABI shared library (.so / .dylib / .dll) that any language with a C FFI -- C, C++, Rust, Go (cgo), Python (ctypes) -- can link or dlopen:

jac nacompile mathlib.na.jac --shared                    # → ./libmathlib.so
jac nacompile mathlib.na.jac --shared --target macos     # → ./libmathlib.dylib
jac nacompile mathlib.na.jac --shared --target windows   # → ./libmathlib.dll

Scalars pass by value; Jac objects and strings cross as opaque handles with jac_retain/jac_release for lifetime management. Jac's own linker emits the ELF/Mach-O/PE file -- no gcc/ld, and the --target cross-builds need no extra toolchain.

Your learning path#

Going further#