A rename after a join, aggr, or union produces invalid SQL on PostgreSQL

未关闭
#1,676 0 条评论 0 个 reaction 已指派 0 人 在 GitHub 查看

维护者通常 6 天内回复

还没有人认领这个 Issue。

评估

难度
3/5
预计耗时
1-2 天
新手友好度
78/100
Issue 类型
缺陷
描述清晰度
描述清楚
活跃度
活跃
技术栈
mysql, postgresql, python
领域
backend, databases

调研方向

Start in src/datajoint/heading.py at Heading.join() and read how Heading.select() uses table_info to quote renamed attributes. Check the existing regression tests for joins, projections, aggregation, and union, then reproduce the examples on PostgreSQL and MySQL. Done means renames after each listed operation produce valid SQL and return correct rows on both backends.

由索引模型根据 Issue 内容生成。

描述

bug

A rename after a join, a left join, an aggr, or a union produces SQL that mixes backticks into PostgreSQL identifiers, and the query fails with syntax error at or near "as". MySQL is not affected.

This is the follow-up raised in the reviews of #1643. #1643 keeps table_info in Heading.set_primary_key, and #1657 keeps it in Heading.make_subquery_heading. Heading.join() (src/datajoint/heading.py, the return Heading(attrs, lineage_available=...) line) still builds the joined heading without table_info. A later proj() rename therefore reaches Heading.select with no adapter and falls back to a backtick-quoted name. aggr (through join.heading.set_primary_key) and Union (through arg1.heading.join(arg2.heading)) inherit the joined heading, so they fail the same way.

Reproducible example (PostgreSQL)
@schema
class A(dj.Manual):
    definition = """
    a : int32
    ---
    x : int32
    """


@schema
class B(dj.Manual):
    definition = """
    -> A
    k : int32
    ---
    b : int32
    """


A.insert([(1, 10), (2, 20), (3, 30)])
B.insert([(1, 1, 100), (1, 2, 101), (2, 1, 200)])
Query SQL on PostgreSQL
(A * B).proj(c="b") SELECT "a","k",`b` as "c" FROM ... JOIN ... USING ("a")
B.join(A, left=True).proj(c="x") SELECT "a","k",`x` as "c" FROM ... LEFT JOIN ...
A.aggr(B, "x", n="count(*)").proj(xx="x") SELECT "a",`x` as "xx" FROM ... GROUP BY "a"
A.aggr(B, n="count(*)").proj(m="n") SELECT "a",`n` as "m" FROM (SELECT ... GROUP BY "a") as "$c"
((B & "a=1").proj() + (B & "a=2").proj()).proj(kk="k") SELECT "a",`k` as "kk" FROM (... UNION ...)

Every one raises QuerySyntaxError: syntax error at or near "as" on PostgreSQL 15, on master and on master with #1643 and #1657 applied. All return the correct rows on MySQL.

Suggested fix

Carry table_info through the join, as Heading.select already does:

return Heading(attrs, table_info=self.table_info or other.table_info, lineage_available=...)

With this one-line change, all the cases above pass on both backends. A regression test parametrized over both backends should cover a rename after *, a left join, aggr (both a grouped attribute and a computed one), a union, and a subquery over a join.

主要语言
Python
星标
198
派生
103
平均合并
6 天 1 小时
30 天内合并 PR
10

环境准备

  • 提供 Dockerfile 或 Docker Compose 文件
  • 没有 Pull Request 模板
  • 阅读贡献指南

从这里开始

  1. 先读完整个 Issue,再读项目的贡献指南。
  2. 在 Issue 下留言说明你要接手 —— 这能避免两个人做同样的事。
  3. Fork 仓库,在一个分支上完成修改。
  4. 提交 Pull Request,并在描述里引用这个 Issue 编号。

datajoint/datajoint-python 的其他 Issue

查看 datajoint/datajoint-python 的全部 Issue

相似的 Issue

更多 Python Issue

把新 issue 发到你的邮箱

精选适合新手参与的 GitHub issue 摘要。