From cf1cc568d66410f9b19ad158cba5aadb266a677d Mon Sep 17 00:00:00 2001 From: ThearchyHelios Date: Wed, 21 Oct 2026 16:09:00 +0800 Subject: [PATCH] =?UTF-8?q?=E5=88=86=E6=9E=90=E8=8C=83=E5=9B=B4=E7=9A=84?= =?UTF-8?q?=E5=80=99=E9=80=89=E7=BB=B4=E5=BA=A6=E5=86=99=E6=98=8E=E5=8F=AA?= =?UTF-8?q?=E8=AE=A4=E7=A0=81=E5=80=BC=E8=A1=A8=EF=BC=9A=E5=89=96=E6=9E=90?= =?UTF-8?q?=E7=9B=AE=E5=89=8D=E4=B8=8D=E8=AE=B0=E6=AF=8F=E5=88=97=E7=9A=84?= =?UTF-8?q?=E5=8F=96=E5=80=BC=E6=95=B0=EF=BC=88=E7=A0=81=E5=80=BC=E5=88=97?= =?UTF-8?q?=E5=8F=A6=E6=9C=89=E5=8F=96=E5=80=BC=E5=88=86=E5=B8=83=E7=9A=84?= =?UTF-8?q?=E8=A7=82=E5=AF=9F=E3=80=81=E9=94=AE=E5=88=97=E5=8F=AA=E6=A0=B8?= =?UTF-8?q?=E5=94=AF=E4=B8=80=EF=BC=89=EF=BC=8C=E6=96=B9=E6=A1=88=20=C2=A7?= =?UTF-8?q?3.6=20=E7=9A=84=E3=80=8C=E5=89=96=E6=9E=90=E7=9A=84=E5=8F=96?= =?UTF-8?q?=E5=80=BC=E6=95=B0=20=E2=89=A4=2050=E3=80=8D=E7=AD=89=E5=89=96?= =?UTF-8?q?=E6=9E=90=E8=AE=B0=E4=BA=86=E6=AF=8F=E5=88=97=E5=8F=96=E5=80=BC?= =?UTF-8?q?=E6=95=B0=E5=86=8D=E8=A1=A5=EF=BC=8C=E6=B2=A1=E6=9C=89=E7=A0=81?= =?UTF-8?q?=E5=80=BC=E8=A1=A8=E7=9A=84=E5=B1=9E=E6=80=A7=E5=88=97=E4=B8=8D?= =?UTF-8?q?=E8=BF=9B=E5=80=99=E9=80=89=EF=BC=88=E7=AC=AC=201=20=E6=B3=A2?= =?UTF-8?q?=E5=A4=8D=E6=A0=B8=20R34=EF=BC=8C=E8=AE=B0=E4=B8=BA=E5=81=8F?= =?UTF-8?q?=E7=A6=BB=EF=BC=89=E3=80=82=E5=8F=AA=E6=94=B9=E6=B3=A8=E9=87=8A?= MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit --- backend/app/engine/analysis_scope.py | 4 +++- 1 file changed, 3 insertions(+), 1 deletion(-) diff --git a/backend/app/engine/analysis_scope.py b/backend/app/engine/analysis_scope.py index c5755f54..a982e7c4 100644 --- a/backend/app/engine/analysis_scope.py +++ b/backend/app/engine/analysis_scope.py @@ -11,6 +11,8 @@ - **没拆**(not_decomposed):候选维度里没拆过的。候选维度确定地从数据目录算:拆解所用查询的 SQL 里的事实表(目录标了 fact / log;没标的、查询只读一张表时就是它)上标成属性、状态(attribute / status)、不遮罩、取值不多(码值表不超过 CANDIDATE_MAX 个取值,或者已列全)的列,加上它关联的维度表(目录标了 dimension,或者关系是多对一)上同样的列。 + 「取值不多」只认码值表:剖析目前不记每列的取值数(码值列另有取值分布的观察、键列只核唯一),方案 §3.6 说的「剖析的 + 取值数 ≤ 50」等剖析记了每列取值数再补(第 1 波复核 R34,记为偏离);没有码值表的属性列不进候选。 拆解按的列对不上目录、读不到目录、认不出事实表的,not_decomposed 为 None,note 写一句为什么核对不了——不乱列 - **检验**(tests):这次做了几次差异检验(做成了的);corrected:一次检验比了多对的按次数校正过,几次检验之间没有校正, 做了不止一次时为 False(契约之外加的一项,见交付说明) @@ -138,7 +140,7 @@ def _table_notes(tables: Mapping[str, Any], name: str) -> tuple[str, dict[str, A def _few_values(column: Mapping[str, Any]) -> bool: - """取值不多:码值表列出的取值不超过 CANDIDATE_MAX 个,或者标了已列全。""" + """取值不多:码值表列出的取值不超过 CANDIDATE_MAX 个,或者标了已列全。只认码值表:剖析不记每列的取值数(见模块说明)。""" codes = column.get("codes") if not isinstance(codes, Mapping) or codes.get("status") == "rejected": return False