[Bug] INCLUDE (non-key) index column referenced by a predicate is treated as valid
维护者通常 3 天内回复
还没有人认领这个 Issue。
评估
调研方向
The bug is in GPORCA's construction of index conditions, and the issue suspects CXformUtils::PdrgpcrIndexColumns, which merges key and INCLUDE columns into one array. Reproduce the two SELECT statements with optimizer on and compare against optimizer off, then check which Index Cond the plan builds and where nodeIndexscan.c rejects it. Done means both queries return rows under optimizer=on without either sanity-check error. The issue reports an unmerged local fix but does not name a target design, so confirm the expected behaviour with maintainers first.
由索引模型根据 Issue 内容生成。
描述
Apache Cloudberry version
No response
What happened
A GPORCA bug: an INCLUDE (non-key) index column referenced by a predicate is treated as a valid
index key column when building an Index Cond (likely via CXformUtils::PdrgpcrIndexColumns, which
concatenates key and included columns into one array with no marker distinguishing the two). The
resulting plan carries a scan key against an attribute beyond indnkeyatts, which the executor
rejects. Which sanity check fires depends on the predicate, so one bug produces two error
messages -- both are suppressed below:
CREATE TABLE t(a int, b ); CREATE INDEX i ON t USING btree (a) INCLUDE (b);
SET optimizer = on;
SELECT a FROM t WHERE b; -- b boolean: "bogus index qualification (nodeIndexscan.c)"
SELECT a FROM t WHERE b IS NULL; -- b int: "btree index keys must be ordered by attribute"
Verified scope: optimizer=on only (the Postgres planner never builds such an Index Cond, and
every case above returns the correct rows under optimizer=off); not specific to a distribution
policy (DISTRIBUTED BY and DISTRIBUTED REPLICATED both reproduce); needs no data, an empty table
is enough; and the INCLUDE list does not have to overlap the key columns -- a plain
"INCLUDE (b)" is sufficient. A fix exists locally but has not been merged into the Cloudberry
main branch this environment is built from.
What you think should happen instead
No response
How to reproduce
CREATE TABLE t(a int, b int); CREATE INDEX i ON t USING btree (a) INCLUDE (b);
SET optimizer = on;
SELECT a FROM t WHERE b IS NULL; -- b int: "btree index keys must be ordered by attribute"
Operating System
any
Anything else
No response
Are you willing to submit PR?
- Yes, I am willing to submit a PR!
Code of Conduct
- I agree to follow this project's Code of Conduct.
- 主要语言
- C
- 星标
- 1.4k
- 派生
- 262
- 平均合并
- 4 天 19 小时
- 30 天内合并 PR
- 54
环境准备
- 没有 Dockerfile 或 Docker Compose 文件
- 有 Pull Request 模板
- 阅读贡献指南
从这里开始
- 先读完整个 Issue，再读项目的贡献指南。
- 在 Issue 下留言说明你要接手 —— 这能避免两个人做同样的事。
- Fork 仓库，在一个分支上完成修改。
- 提交 Pull Request，并在描述里引用这个 Issue 编号。
apache/cloudberry 的其他 Issue
-
[Bug] Expand fails to recognize user-defined tablespaces可能已有人在做 @nix-oss 于 59 天前认领。 未关闭type: Bug
难度 2/5 1-3 小时 新手友好度 76/100
apache/cloudberry#1885 · 2 个 reaction ·
维护者通常 3 天内回复
-
gpexpand first stage fails: 'SyncPackages' object has no attribute 'ret' (extension sync never runs; regression)可能已有人在做 @talmacschen-arch 于 120 天前认领。 未关闭
难度 2/5 1-3 小时 新手友好度 86/100
apache/cloudberry#1825 ·
维护者通常 3 天内回复
-
type: Bug
难度 3/5 半天 新手友好度 58/100
apache/cloudberry#2137 ·
维护者通常 3 天内回复
-
type: Proposal
难度 4/5 3-5 天 新手友好度 22/100
apache/cloudberry#2136 · 1 条评论 ·
维护者通常 3 天内回复
-
type: Bug
难度 4/5 3-5 天 新手友好度 30/100
apache/cloudberry#2135 ·
维护者通常 3 天内回复
查看 apache/cloudberry 的全部 Issue
相似的 Issue
-
backlog bug
难度 2/5 1-3 小时 新手友好度 75/100
EchoTools/nevr-runtime#653 ·
维护者通常 1 天内回复
-
难度 2/5 1-3 小时 新手友好度 75/100
InfiniTimeOrg/InfiniTime#2492 ·
-
难度 2/5 1-3 小时 新手友好度 80/100
维护者通常 1 天内回复
-
base-files: 82_config_generate: regenerate /etc/board.json if 0-bytes / truncated after power cut未关闭
难度 2/5 1-3 小时 新手友好度 72/100
openwrt/openwrt#25940 · 5 条评论 ·
维护者通常 1 天内回复
-
难度 2/5 1-3 小时 新手友好度 68/100
维护者通常 1 天内回复