apache / apache/arrow-java

Truly unchecked getters

未关闭
#1,252 0 条评论 0 个 reaction 已指派 0 人 在 GitHub 查看
Type: enhancement
主要语言
Java
星标
94
派生
152
平均合并
3 天 16 小时
30 天内合并 PR
11

描述

### Describe the enhancement requested

Right now the library integrating with arrow-java needs to resort to hacks of setting environment variable or system property at runtime if it knows its usages of arrow-java can omit null and bounds checking. However, that is not true for all of the usages in the given process leading to invalid data reads or segfaults.

To avoid this pitfall I suggest we add getUnchecked apis that NEVER do any null and bound checking but put the responsibility on the caller to guard against their possibility. For example in apache iceberg https://github.com/apache/iceberg/blob/23b5ce8eeca2d894c7973e19caa5c5ec02b4e4b8/spark/v4.1/spark/src/main/java/org/apache/iceberg/spark/data/vectorized/VectorizedSparkParquetReaders.java#L44-L51

These apis would make it easy to mix libraries that use arrow as an implementation detail along with user code that operates directly on arrow arrays

贡献指南

打开贡献指南

调研方向

首先检查现有的 Arrow 数组 getter API,以及链接行中 Iceberg 的 VectorizedSparkParquetReaders.java 对未经检查访问的使用情况。Issue 未指出任何 arrow-java 文件或测试,因此请先确定受影响的 getter 系列和测试位置。完成标准是所请求的未经检查 API 均可一致地使用,并且测试覆盖由调用方负责的行为。

由索引模型根据 Issue 内容生成。

评估

技术栈
java
领域
data-engineering
Issue 类型
功能
难度
5/5
预计耗时
一周以上
活跃度
冷清
描述清晰度
需要澄清
新手友好度
35/100

把新 issue 发到你的邮箱

精选适合新手参与的 GitHub issue 摘要。