Skip to main content

skijack

A surface language over the $SKI$ combinator calculus that compiles to closed {S,K,I} terms, with the interpreter of the companion paper as its conformance target.

The claim this package exists to support: the abstraction is free, and term equality proves it. The self-interpreter of Davis (2027)---618 atoms of pure {S,K,I}, built by hand---compiles from three lines of surface source to the same term, atom for atom. So do five more: the observability pair at 768 and 771, the blocking interpreters at 950 and 1,066, and structural equality on encoded terms at 240.

Everything is compile-time. Quotation happens in the compiler and nowhere else, because no closed term quotes every live term; the language's job is to make that line visible rather than to pretend it isn't there.

  • N E Davis (2027, in review) A Self-Interpreter for $SKI$: Authoring Semantics for New Symbols. Journal of Functional Programming. doi:10.5281/zenodo.22867957
  • N E Davis (2027a, in review) SKIjack: A Low-Level Native Language for the $SKI$ Calculus. The Art, Science, and Engineering of Programming.

Install

pip install skijack

Requires Python 3.10+ and aviary-kernel, which supplies bracket abstraction, the term representation and the reducer. This package reimplements none of them.

Avon, a C runtime, can run the reductions instead: SKIJACK_REDUCER=avon with AVON naming its binary, or SKIJACK_REDUCER=check to run both and compare (skijack/avon.py). Avon's copy mode gives aviary's statuses, counts and terms; SKIJACK_AVON_STRATEGY=share runs its share mode, each argument reduced once, with the same normal forms and its own counts.

Use

skijack FILE                     # a summary
skijack FILE --check             # Stage A, then B; exits 1 and lists every problem
skijack FILE --expand            # every compiled term, with atom counts
skijack FILE --dictionary        # the Tier 1 table: name, atoms, hash
skijack FILE --lift NAME         # NAME, with what the table knows named
skijack FILE --render unicode    # the same program in the other lexicon
skijack FILE --run NAME [--fuel N]   # run a level-1 declaration and decode it

The lexicon is taken from the filename (*.unicode.ski) unless --lexicon says otherwise. The twenty-two programs of the conformance corpus ship with the package; eighteen compile, and Stage A refuses the other four for stated reasons:

>>> import skijack
>>> from skijack import corpus
>>> skijack.compile(corpus.read("interp-whnff")).sizes["whnfF"]
618

From Python:

import skijack
exp   = skijack.compile(open("prog.ski").read())   # parse, check, generate, expand
d     = skijack.from_expansion(exp)                # the Tier 1 table
named = skijack.lift(exp.term("whnfF"), d)         # back to named source
out   = skijack.run_level1(exp.level1["answer"], 1_000_000)

Everything this package raises for a bad program derives from skijack.SkijackError, including a refusal by aviary, the host that abstracts and expands, which reaches the caller as an ExpandError.

Definitions

An author writes an equation f x y = body. The closed Y-tied term to which it compiles is the supercombinator, the unit of naming, sharing, jetting and lifting. The group an equation is declared in is a core. Names resolve in a compile-time table and are gone before anything runs; there is no environment at run time.

Tests

The suite is not in the wheel. Run it from a checkout of SKIjack:

pip install -e "python[test]"
python3 -m pytest -q python/tests

Roughly a third of the suite compares compiled terms against the hand-built artifact of the companion paper. Its oracle scripts are vendored under python/tests/artifact, so the suite runs with no setup; SKIJACK_ARTIFACT_DIR points it at a live checkout of the artifact instead.

More

The language is specified in SPEC.md at the repository root; NOTES.md beside this file is the decision log it was derived from.

NOTES.md records what the suite proves, every open question the specification left and how it was answered, and any discrepancies found in the original specification itself.

Metadata

Release files for skijack 0.2.6

For a detailed explanation of source distributions (sdists) and built distributions (wheels), please see the package formats documentation.

Source distribution (sdist)

Source distribution for skijack 0.2.6
File Size Uploaded Provenance Actions
skijack-0.2.6.tar.gz 300.4 kB Without provenance Details for skijack-0.2.6.tar.gz

Built distribution (wheel)

Table of built distributions (wheels) for skijack 0.2.6
File Interpreter ABI Platform Actions
skijack-0.2.6-py3-none-any.whl Python 3 none any Details for skijack-0.2.6-py3-none-any.whl

Total release size: 530.5 kB

Release files / skijack-0.2.6.tar.gz

Download URL skijack-0.2.6.tar.gz
Size 300.4 kB
Tags Source
SHA-256 checksum
How to use checksums
9b115cc662c7d3ee4749899f343f12ac89570835037868f7cc76c8092e6ab67b
BLAKE2b-256 checksum
How to use checksums
74fb9fe7627231bc7e2e4c997eab4d82dc32ed3af8216638507ee79f1fffd3c7
Upload date
Uploaded using Trusted Publishing?
What is trusted publishing?
No
Uploaded via twine/6.2.0 CPython/3.12.7

Release files / skijack-0.2.6-py3-none-any.whl

Download URL skijack-0.2.6-py3-none-any.whl
Size 230.1 kB
Tags Python 3
SHA-256 checksum
How to use checksums
00a0da6a434291d455ddaf4539cf499a54bed5c8178265e802f5cd37405aabb3
BLAKE2b-256 checksum
How to use checksums
319e2a8e7d807377f226f422aeb91fd21e082d6cf174379a04d9aa957fc331a8
Upload date
Uploaded using Trusted Publishing?
What is trusted publishing?
No
Uploaded via twine/6.2.0 CPython/3.12.7

Security for skijack 0.2.6

Provenance

Provenance describes where a file came from. On PyPI, provenance is shared via attestations, which provide a verifiable record of the build or publishing details. View details, limitations and caveats.

No provenance found

None of the files in this release were uploaded with provenance metadata.

A note on package security

PyPI is a repository for distributing packages, not a service for auditing them. While third-party security researchers and our staff do their best to detect and remove malicious packages, PyPI cannot guarantee the quality, security or suitability of any given package.

To learn more about protecting yourself from malicious actors, visit our package security documentation.

Anthropic, PBC Visionary sponsor Bloomberg Visionary sponsor Hudson River Trading Visionary sponsor Meta Visionary sponsor NVIDIA Visionary sponsor Microsoft Sustainability sponsor Depot Continuous Integration AWS Cloud computing and Security Sponsor Datadog Monitoring Fastly CDN Google Download Analytics Sentry Error logging StatusPage Status page