Skip to content

Version 0.9.5

Choose a tag to compare

View all tags
@zipproth zipproth released this 16 Mar 19:46
· 79 commits to main since this release
v0.9.5
ae7d7be

This is a beta release with major new capabilities, following version 0.9.4.

🚀 Context: The 0.9.x series

Version 0.9.0 was a major milestone supported by a Wikimedia Rapid Fund.
For details on the architecture and the full grant report, please see:
👉 Full 0.9.0 Grant Report

📦 Changes in v0.9.5

  • Janet Scripting Integration:

    • Embedded Janet as a full scripting layer — every zelph statement is parsed into Janet before execution.
    • Rich API: zelph/fact, zelph/query, zelph/resolve, zelph/set, zelph/list, zelph/list-chars, zelph/rule, zelph/negate, zelph/exists, zelph/name, zelph/sources, zelph/targets, zelph/car, zelph/cdr.
    • Inline (%expr), block (%...%), and unquote (,var) syntax for seamless interplay between zelph and Janet.
    • JSON import/export and CSV support via Janet's standard library and spork.

  • Lists, Sets & Deep Unification:

    • Lisp-style cons-list topology (< a b c > and compact <abc> syntax).
    • Unordered sets ({ a b c }) with proper rule matching.
    • Deep unification over arbitrarily nested structures, enabling pattern matching on statements-about-statements.

  • Reasoning Engine:

    • Negation as failure with ¬(pattern) syntax sugar and zelph/negate API.
    • Inequality constraints (!=) as built-in guard — filters variable bindings after unification, essential for Wikidata ontology validation.
    • Fresh variables in rule consequences — the engine generates new anonymous nodes for constructive reasoning.
    • Conjunction syntax sugar — comma-separated conditions: (cond1, cond2) => consequence.
    • Reasoning speedup by factor 22× through optimized evaluation order.

  • Proof of Concept: Rule-based Arithmetic:

    • Multi-digit addition implemented purely via graph inference rules — no hard-coded arithmetic.
    • Numbers represented as cons-lists of digit nodes; a 200-entry lookup table and eight inference rules handle arbitrary-precision addition with carry propagation.

  • Performance & Memory:

    • Integrated mimalloc; significantly reduced peak RAM for Wikidata import.
    • Optimized adjacency_set with custom Flexible Array Member block allocator.
    • Removed Boost dependency entirely; migrated from std::wstring to UTF-8 std::string.

  • Testing:

    • Adopted doctest as the test framework with CTest integration.
    • Comprehensive test suite covering parsing, nested unification, negation, inequality semantics, Janet integration, meta-rules, contradiction detection, and multi-digit arithmetic — all run in both parallel and single-core mode.

  • Documentation & Packaging:

  • Build & Refactoring:

    • Refactored large source files into thematic compilation units; restructured namespaces.
    • Cross-compilation support for Linux ARM64 via QEMU in CI.
    • Improved Mermaid diagram rendering (dynamic depth, subgraph support).
    • Added reasoning profiler for performance diagnostics.

Binaries for Windows, Linux (x64 & ARM64), and macOS are automatically attached below, both as .zip archives and as .deb packages (Linux).

Assets 8
Loading