[Feature] Support File Index V2 and streaming index writes
維護者通常 3 天內回覆
評估
研究方向
Start by reading the existing File Index container and write path, then compare the V2 layout with the Java reference PR #10101 and the related work in #173. The issue names no source files or tests, so locate the container, writer, plugin interfaces, and spillable output stream before estimating the changes. Done means V1 compatibility remains, V2 reads and writes Java-compatible bytes, and the listed payload, empty-index, and spill-threshold cases are covered; an open linked PR #446 is already working on it.
由索引模型根據 Issue 內容生成。
描述
Search before asking
- I searched in the issues and found no existing issue specifically tracking File Index V2.
Motivation
The V1 File Index container uses 32-bit payload offsets and lengths. The existing write path also materializes serialized index payloads and the complete container in memory before deciding whether to embed the index or write an external file. These limitations make it difficult to support large index artifacts, especially vector indexes.
Java Paimon introduced a V2 container with 64-bit payload ranges and a streaming write interface. Paimon C++ should support the same format while retaining V1 compatibility.
Solution
- Add V2 container reading and writing, with int64 payload offsets and lengths and a footer/trailer layout aligned with Java Paimon.
- Support
file-index.format-versionvalues1and2, keeping V1 as the default. - Add
FileIndexWriter::WriteTo()so plugins can write payloads directly to a caller-owned stream. Keep a default implementation based onSerializedBytes()and implement direct streaming output for Lumina. - Widen plugin reader offsets and lengths to int64, with plugin-specific size limits checked inside the affected plugins.
- Integrate a spillable output stream into the data-file index write path, embedding small containers and spilling larger ones to external files without buffering the complete V2 container.
- Add focused coverage for V1 compatibility, Java-generated V2 golden bytes, large payload ranges, empty indexes, and spill thresholds.
The V2 container's 64-bit ranges do not imply that every index plugin supports payloads larger than 2 GiB; individual plugin serialization formats may retain their own limits.
Anything else?
Java reference: https://github.com/apache/paimon/pull/10101
Related to File Index writing in #173.
Are you willing to submit a PR?
- I'm willing to submit a PR!
- 主要語言
- C++
- 星號
- 68
- 分支
- 34
- 平均合併
- 4 天 2 小時
- 30 天內合併 PR
- 38
環境準備
- 沒有 Dockerfile 或 Docker Compose 檔案
- 有 Pull Request 範本
- 閱讀貢獻指南
從這裡開始
- 先讀完整個 Issue，再讀專案的貢獻指南。
- 在 Issue 下留言說明你要接手 —— 這能避免兩個人做同樣的事。
- Fork 儲存庫，在一個分支上完成修改。
- 送出 Pull Request，並在描述裡引用這個 Issue 編號。
apache/paimon-cpp 的其他 Issue
-
enhancement
難度 5/5 一週以上 新手友好度 35/100
apache/paimon-cpp#452 ·
維護者通常 3 天內回覆
-
難度 5/5 一週以上 新手友好度 8/100
apache/paimon-cpp#435 · 2 則留言 ·
維護者通常 3 天內回覆
-
enhancement
難度 5/5 一週以上 新手友好度 35/100
apache/paimon-cpp#410 ·
維護者通常 3 天內回覆
-
enhancement
難度 5/5 一週以上 新手友好度 25/100
apache/paimon-cpp#409 ·
維護者通常 3 天內回覆
-
enhancement
難度 5/5 一週以上 新手友好度 35/100
apache/paimon-cpp#408 ·
維護者通常 3 天內回覆
查看 apache/paimon-cpp 的全部 Issue
相似的 Issue
-
cudf-polars feature request
難度 2/5 1-3 小時 新手友好度 77/100
維護者通常 2 天內回覆
-
(S1 - Need confirmation)
難度 2/5 1-3 小時 新手友好度 68/100
CleverRaven/Cataclysm-DDA#89212 ·
維護者通常 1 天內回覆
-
Status: Needs confirmation Status: Needs triage
難度 2/5 1-3 小時 新手友好度 70/100
維護者通常 1 天內回覆
-
a byte order mark at the start of a migration makes dbmate miss the up block可能已有人在做 @Dem1241 今天認領。 未關閉
難度 2/5 1-3 小時 新手友好度 70/100
維護者通常 2 天內回覆
-
ai_p2 comp-dictionary
難度 2/5 1-3 小時 新手友好度 78/100
ClickHouse/ClickHouse#127261 ·
維護者通常 1 天內回覆