[Java][arrow-jdbc] Ability to customize JdbcConsumer construction
- 主要言語
- Java
- スター
- 94
- フォーク
- 152
- 平均マージ
- 3日 16時間
- マージ済み PR(30日)
- 11
説明
### Describe the enhancement requested
I'm working on a project that heavily uses the arrow-jdbc library to convert results from a number of different JDBC drivers to Arrow. When the integration was initially written a bit over 2 years ago it forked `ArrowVectorIterator` to add a number of customization points in the JDBC->Arrow conversion which were missing at the time. I'm updating the project and happily in the intervening time it looks like a lot of this customization is now possible natively via customization points in `JdbcToArrowConfig` - in particular the ability to override `JdbcFieldInfo` per column and most importantly the ability to override a `Function`.
I'd love to fully remove our forked `ArrowVectorIterator` but there's one last customization point we're depending on that I don't see a way to handle currently: The ability to customize the `JdbcConsumer` instances constructed for each column. We depend on overriding these to paper over idiosyncrasies with various JDBC drivers. As an example on the top of my mind this morning: The Snowflake JDBC driver [does not support](https://docs.snowflake.com/en/user-guide/jdbc-api.html#id14) `getBinaryStream` and so is not compatible with the default `BinaryConsumer` and must be overriden with a consumer that uses `getBytes` instead. We also add a layer of wrapping to all of our `JdbcConsumers` to wrap error messages to aid debugging (by adding info on the specific column an error originated from).
Would it be possible to add an extension point to the `JdbcToArrowConfig` to customize this, similar to `jdbcToArrowTypeConverter`? I wish we didn't need it but in practice we've found enough variability across JDBC drivers to necessitate it.
### Component(s)
Java
コントリビューションガイド
調査の方向性
まず ArrowVectorIterator と JdbcToArrowConfig を読み、次に各列について JdbcConsumer のインスタンスがどのように構築されるかを追跡します。ドライバー固有の consumer と wrapping を可能にする、jdbcToArrowTypeConverter に類似した拡張ポイントを定義し、Snowflake の getBytes workaround をそこから提供できることを検証します。
索引モデルが issue の本文から書いたものです。
評価
- 技術スタック
- java
- 領域
- databases
- issue の種類
- 機能追加
- 難易度
- 5/5
- 見積もり時間
- 1週間以上
- 活発さ
- 停滞
- 明瞭さ
- おおむね明確
- 初心者へのやさしさ
- 38/100