← All modules

SHA-256 through a C implementation rather than a Raku one — two subs, one returning the raw digest and one the hex string, for when hashing is on a path that runs often enough for the speed to show.

Version

1.0.2 zef:bduggan

Depends

LibraryMake

License

Artistic-2.0

Its own test suite

9 files, green

Checked

2026-09-14 against Raku++ 3.28.0 and Rakudo 2026.08

Where it lives

raku.land · source

Install it #

$ rakupp install Digest::SHA256::Native

zef install Digest::SHA256::Native writes the same store; either installer leaves the module usable by both engines.

What it is for #

SHA-256 is the hash you reach for when you want a content address: a cache key that changes when the content does, an ETag, a fingerprint proving two files are the same, the leaves of a Merkle tree. Pure-Raku implementations exist and are correct, but a hash is usually in a loop — every file in a tree, every chunk of an upload — and that is where the cost of doing thirty-two-bit arithmetic in an interpreter shows up.

This distribution binds the C implementation through NativeCall. The distribution builds the shared library at install time, which is why LibraryMake is a dependency and why installing it needs a working C compiler; after that it is two subs.

Both subs #

use Digest::SHA256::Native;

say sha256-hex('hello');
say sha256-hex('');
say sha256-hex('hello').chars;

2cf24dba5fb0a30e26e83b2ac5b9e29e1b161e5c1fa7425e73043362938b9824
e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855
64

Those are the canonical answers — the second is the SHA-256 of the empty string, the value worth memorising because it is what you get when a file you meant to read came back empty. sha256-hex gives the 64-character lowercase string; sha256 gives the 32 raw bytes, which is what you want when the digest is going into a binary format or straight into another hash rather than onto a screen.

The one thing to know #

There is no incremental interface. Both subs take the whole message as one value and return the digest of it — there is no object to .add to and no .finalize, so hashing a four-gigabyte file means having four gigabytes in memory. For anything larger than comfortable, either shell out to the system's own shasum, or use a pure-Raku Digest that offers a streaming API and accept the slowdown.

The other thing to check before depending on it is the build. Because the shared library is compiled during installation, this module can install cleanly on your machine and fail on a container image with no toolchain — a failure that arrives at deploy time rather than at test time. If that matters, pin a pure-Raku fallback behind the same two-sub interface; the digests are identical, so nothing above the call site needs to know which one ran.