Hacktoberfest 2026：維護者為十月標記出來的 issue，仍然開放、適合新手。 瀏覽 Hacktoberfest issue

Gate pyo3/extension-module so the crate can be linked as a Rust dependency and tested

未關閉
  • 哪裡不對？
#1,703 1 則留言 0 個 reaction 已指派 1 人 在 GitHub 檢視

維護者通常 1 天內回覆

@emecii 已經在處理了。

開始於 2026年9月10日。

評估

這個 Issue 還沒有評估資料。

描述

development-process enhancement rust

Is your feature request related to a problem or challenge? Please describe what you are trying to do.

crates/core/Cargo.toml enables pyo3/extension-module unconditionally. That feature tells pyo3 not to link against libpython, which is right for the extension module the wheel ships, but it makes the crate unusable in any build that needs to link Py_* symbols itself. Two things are blocked by this today.

The first is Rust tests. No workflow invokes cargo test; the only Rust checks in CI are cargo fmt --check and cargo clippy --no-deps --all-targets. --all-targets compiles #[cfg(test)] code, so a Rust test cannot rot into a non-compiling state, but it is never executed and a behavioral regression will not fail the build. Adding the job is not a one-line change, because the test binary fails to link on Linux for exactly the reason above. This is recorded in AGENTS.md as the reason Rust tests are currently dead weight in this repository.

The second is consuming datafusion-python as an ordinary Rust dependency. crate-type already includes rlib, so this looks supported, but a downstream crate that builds a binary hits the same link failure. This came up in https://github.com/apache/datafusion-python/pull/1678#pullrequestreview-5100366976, where the request was to make some of the Python UDF serialization internals public so they could be plugged into an existing physical codec in a distributed setup. Marking those items pub would advertise an API that a downstream crate cannot actually link against, so the visibility change is not the useful part on its own.

Describe the solution you'd like

Put pyo3/extension-module behind a Cargo feature that is on by default (so the wheel build and maturin develop are unchanged) and can be turned off by a consumer or a test build. Then add a cargo test job to CI in the same change, so the tests that exist actually run and the gate does not silently regress.

Describe alternatives you've considered

Leaving it as is and keeping all Rust behavior covered from Python. That is the current practice and it works well for the user-facing surface, which is the primary focus anyway. It does not help the downstream-consumer case, and it means genuinely Rust-only invariants have no executable coverage.

Splitting the crate, with the reusable pieces in a library crate that does not depend on extension-module and the pyo3 bindings in a thin crate on top. Cleaner in the long run and a much larger change; worth considering if the feature gate turns out to be awkward.

Additional context

Follow-up from #1678. This one is a prerequisite for exposing any Rust-facing API from this crate, including the codec work requested in that review.

主要語言
Python
星號
606
分支
176
平均合併
1 天 7 小時
30 天內合併 PR
14

環境準備

  • 沒有 Dockerfile 或 Docker Compose 檔案
  • 有 Pull Request 範本
  • 沒有貢獻指南

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。

apache/datafusion-python 的其他 Issue

查看 apache/datafusion-python 的全部 Issue

相似的 Issue

更多 Python 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。