Skip to content

jacquard-core-0.2.0

Latest

Choose a tag to compare

@github-actions github-actions released this 15 Aug 18:57
· 12 commits to main since this release
aceca65

Jacquard Core 0.2.0

Jacquard 0.2.0 is the first integrated release of the public .jac authoring
surface and the successor research systems built over the 27-form kernel.

Highlights:

  • readable .jac source with deterministic formatting, kernel lowering,
    direct native build, and bootstrap export;
  • explicit effect rows and runtime grants, deep multi-shot handlers, and
    affine once resumptions;
  • finite discrete inference, Warp properties, replay, fault exploration, and
    relational schedule/Secret/grant variation;
  • interpreter structured concurrency with scoped Tasks, cooperative
    cancellation, deterministic and exhaustive scheduling, strict replay, and
    scoped typed Channels;
  • canonical Audit chains, opaque Secret handling, proposal-bound Approval,
    and a typed Workspace v0 governance reference implementation;
  • native AOT differential, memory, leak, and seeded fuzz evidence under both
    Clang and GCC;
  • checksum-verified archives for Linux x86-64, macOS Intel, and macOS Apple
    Silicon, including the prelude, demos, and native runtime.

Install the final release:

curl -fsSL https://raw.githubusercontent.com/jbwinters/jacquard-lang/jacquard-core-0.2.0/scripts/install.sh | sh
export PATH="$HOME/.local/bin:$PATH"
jac --version
jac run "$HOME/.local/share/jacquard/demos/basics/m1-fact.jac"

For RC verification before final promotion:

curl -fsSL https://raw.githubusercontent.com/jbwinters/jacquard-lang/jacquard-core-0.2.0-rc1/scripts/install.sh \
  | JACQUARD_INSTALL_VERSION=jacquard-core-0.2.0-rc1 sh

Jacquard remains a research prototype. Language grants are not an OS sandbox;
structured concurrency and Channels are interpreter-only; probability is
finite and discrete; Workspace governance is not a production authorization
system; the public .jac projection remains evolving v0 syntax; and no human
readability result is claimed. Read the bounded claims, full
limits
, and reproduction procedure before relying on a
feature claim.