Port Math::Random::MT as a bundled Java XS implementation

未關閉
#2,005 0 則留言 0 個 reaction 已指派 0 人 在 GitHub 檢視

維護者通常 1 天內回覆

還沒有人認領這個 Issue。

評估

難度
5/5
預估耗時
一週以上
新手友好度
45/100
Issue 類型
功能
描述清晰度
描述清楚
活躍度
活躍
技術堆疊
java, perl
領域
backend, devtools

研究方向

Start by inspecting the bundled Java XS bridge pattern and registry, then read the upstream Math-Random-MT 1.17 tests and implementation to identify the required seeded vectors and API behavior. Add the bridge, Perl shim, and focused project tests; validate those tests with system Perl first. Done means the upstream suite and both named Games-Solitaire-BlackHole-Solver tests pass on JVM and interpreter backends, with no runtime need for MT.xs or _mt.c.

由索引模型根據 Issue 內容生成。

描述

area:cpan-port enhancement

Summary

Add a bundled Java implementation of Math::Random::MT so PerlOnJava can load the CPAN module without its XS binary. This should also unblock CPAN consumers that require Math::Random::MT at runtime, including Games-Solitaire-BlackHole-Solver.

Evidence

CPAN random run 20261109-091152-39587 tested Games-Solitaire-BlackHole-Solver 0.18.0. Under PerlOnJava, t/00-compile.t could not load Math::Random::MT because no Java XS implementation was available; t/01-run.t then failed after the solver could not run. The system-Perl oracle passed the complete target suite: 27/27 tests across both test programs. The same system-Perl run built and passed Math-Random-MT 1.17's nine test files (125 tests).

The failed solver requires Math::Random::MT at runtime. Targeted MetaCPAN reverse-dependency data reports 13 unique released CPAN distributions with runtime dependencies, 4 recent; examples include Games-Solitaire-BlackHole-Solver, Task-BeLike-SHLOMIF, Log-GELF-Util, and Bencher-Scenario-RandomNumbers.

Proposed implementation

Provide the Java XS bridge and Perl shim for the Math::Random::MT API, following the project's existing Java XS bridge pattern. Bouncy Castle is already a project dependency, but its documented RandomGenerator is an interface and does not itself provide an MT19937 engine. Avoid assuming that a Bouncy Castle DRBG has MT19937-compatible sequences. A small pure-Java MT19937 implementation avoids adding another runtime dependency; Apache Commons RNG is an alternative only if its additional dependency and seed/output compatibility are justified.

Preserve the CPAN module's behavior and seeded sequence, including:

  • Object interface: new, set_seed, get_seed, rand, and irand.
  • Functional interface and optional exports: srand, rand, and irand.
  • Single unsigned 32-bit seed and array seeding (up to 624 words), including the published high-value seed cases.
  • Automatic seeding when no seed is supplied, without changing Perl's CORE::rand state beyond the upstream behavior.
  • Unsigned 32-bit results from irand and [0, n) floating-point results from rand(n).

Acceptance criteria

  • Add the module to the bundled Java XS registry and provide the required Perl-facing shim, without modifying upstream CPAN tests.
  • Add project-owned focused tests for seeded MT19937 output, array seeding, unsigned output, OO calls, function imports, and automatic seeding. Validate new Perl unit tests with system Perl first as required by the repository.
  • Run the upstream Math-Random-MT 1.17 suite with the Java implementation and confirm its deterministic vectors match system Perl.
  • Re-run Games-Solitaire-BlackHole-Solver 0.18.0 with PerlOnJava and confirm both t/00-compile.t and t/01-run.t pass; test JVM and interpreter backends.
  • Confirm native MT.xs/_mt.c is not needed at runtime when the Java implementation is selected.

References

主要語言
Perl
星號
67
分支
8
平均合併
5 小時 12 分鐘
30 天內合併 PR
131

環境準備

  • 提供 Dockerfile 或 Docker Compose 檔案
  • 沒有 Pull Request 範本
  • 閱讀貢獻指南

從這裡開始

  1. 先讀完整個 Issue,再讀專案的貢獻指南。
  2. 在 Issue 下留言說明你要接手 —— 這能避免兩個人做同樣的事。
  3. Fork 儲存庫,在一個分支上完成修改。
  4. 送出 Pull Request,並在描述裡引用這個 Issue 編號。

fglock/PerlOnJava 的其他 Issue

查看 fglock/PerlOnJava 的全部 Issue

相似的 Issue

更多 Perl Issue

把新 issue 寄到你的電子郵件信箱

精選適合新手參與的 GitHub issue 摘要。