wandering1
bda7ab2b50
批 2（用户授权改名 + 「新名要一眼看懂」裁定）
- 16 个有歧义模块码单词化，共 1,146 处随 scripts/rename_ids.py 改写：
AI→ADCIN、AH→ADCHEALTH、PS→PROTSET、PE→PROTELEM、PH→PHASOR、
TS→TIMESYNC、VL→VOLLOCK、NP→NANDPORT、CFG→CONFIG、CLK→SYSCLK、
WD→WATCHDOG、STO→STORAGE、SH→SHA256、FA→FTPAUTH、FR→FREQ、104→IEC104
- 不动 NET/GD5F/RTC/SOE/UI/ADC/LOG/RS485/TOOL（已是单词、芯片名或业界通用词，
改了纯属 churn）
- 别名按全局登记：这批字面量已完全限定（TC-<码>-<nnn>），不存在裸形多义
- 映射表存档 scripts/rename_maps/batch2.json 作审计轨迹；§6 模块码表同步
连带修门禁（关键教训）
- 模块码涨到 9 位（ADCHEALTH/WATCHDOG）后，候选正则与族正则的 {0,4} 上界
会让改名后的号**整批从门禁眼里消失**：实测编号字面量从 8271 掉到 7315、
悬空假降 490 处。已放宽到 {0,9} 并写进 §6 注记
- 结论：改名批次必须连带复核识别面，只改数据不改判据 = 静默放行；
这次是靠「可判定比例跌破基线即失败」那条线拦下来的
现状与遗留
- 校验 1 未登记前缀 0 处；可判定比例 82.71%（基线 82.6%，只涨不跌成立）
- 遗留：7 个模块的「单元测试说明.md」仍缺（PROTSET 116、NANDPORT 109、
PROTELEM 65、FREQ 48、FTPAUTH 45、SHA256 25、SOE 14 处悬空）——改名不解决
缺文档，处置方式待议
- 遗留：T-2/T-3/T-4/T-6 历史任务号悬空（别名表未收全，属本工具批次漏项）
- 未改任何 C/H 可编译语义（仅注释与宏名内的编号），故不需 @build
658 lines
34 KiB
Python
658 lines
34 KiB
Python
#!/usr/bin/env python3
|
||
# -*- coding: utf-8 -*-
|
||
"""
|
||
check_ids.py — 编号门禁（docs/编号规范.md §8 的 8 项校验）。
|
||
|
||
两份配置（数据源 = 规范 §4 编号空间总表，逐项对应，不凭印象）：
|
||
scripts/id_registry.json 前缀族 → {形态正则, 属主登记处, 是否冻结, 游标, 作用域, 级别}
|
||
scripts/id_aliases.json 旧字面量 → 现行字面量（照搬 scripts/section_aliases.json 的成功范式）
|
||
|
||
8 项校验（级别按规范 §8 表）：
|
||
1 出现未登记前缀（自造编号） 错误
|
||
2 形态不合规：裸形 S3 / 空格形 Trap 07 / 中文形 陷阱 1 /
|
||
补零错 / 档位词带后缀 P1-8 / 大写后缀 / 缺模块码 错误
|
||
3 引用查无定义且不在别名表（悬空引用） 错误
|
||
4 同一字面量被两个属主重复定义（N-16 类事故） 错误
|
||
5 冻结前缀出现游标之外的新号 错误
|
||
6 文档内编号引用未紧邻文档名 错误
|
||
7 定义了但全库无引用（孤儿条目） 告警
|
||
8 打印「编号引用可判定比例」 每次运行，只许涨不许跌
|
||
|
||
「定义」的机械判据：**字面量出现在该族的属主登记处文档里**（不依赖表格行首这类易漂移的
|
||
形态）。因此 §5.5 的「销项 = 删行 + 清全库引用」一旦只做一半，校验 3 立刻报——这正是
|
||
§1.2 里 N-31/N-32 撞号与悬空的成因。
|
||
|
||
排除（规范 §8 + §4.4）：Lib/、Middlewares/、HAL/CMSIS、docs/芯片手册/、docs/硬件平台/
|
||
（元器件位号 U1/C12/R3/L5/D7 与编号字面同形，是 B1/C1/L1 撞形的根源，必须整体排除）、
|
||
构建产物。裸形盲区显式登记在 id_registry.json 的 `unmanaged_bare`，不是「检查通过」。
|
||
|
||
上线节奏（§8）：默认 **warn-only**（退出码 0）；`--strict` 才让校验 1~6 拦提交。
|
||
第 8 项的比例跌破基线**任何模式都失败**（它就是防回退的那根线）。
|
||
|
||
用法（仓库根）：
|
||
& '.venv\\Scripts\\python.exe' scripts/check_ids.py # warn-only 全库
|
||
& '.venv\\Scripts\\python.exe' scripts/check_ids.py --strict # 错误拦提交
|
||
& '.venv\\Scripts\\python.exe' scripts/check_ids.py --top 30 --verbose # 多打 + 修法
|
||
& '.venv\\Scripts\\python.exe' scripts/check_ids.py --update-baseline # 抬升第 8 项基线
|
||
& '.venv\\Scripts\\python.exe' scripts/check_ids.py --only 1,3,5 # 只跑部分校验
|
||
退出码：0 = 通过 / 仅告警；1 = 错误（--strict）或比例跌破基线；2 = 环境错误。
|
||
"""
|
||
|
||
import argparse
|
||
import fnmatch
|
||
import json
|
||
import math
|
||
import os
|
||
import pathlib
|
||
import re
|
||
import subprocess
|
||
import sys
|
||
from collections import Counter, defaultdict
|
||
|
||
# 控制台编码兜底：pre-commit 钩子环境的代码页是 936/GBK，打印中文或箭头会抛
|
||
# UnicodeEncodeError → 钩子带 traceback 以退出码 1 死亡（同 check_doc_links.py 的教训）。
|
||
for _stream in (sys.stdout, sys.stderr):
|
||
if hasattr(_stream, "reconfigure"):
|
||
_stream.reconfigure(encoding="utf-8", errors="replace")
|
||
|
||
REPO_ROOT = pathlib.Path(__file__).resolve().parent.parent
|
||
REGISTRY_FILE = REPO_ROOT / "scripts" / "id_registry.json"
|
||
ALIASES_FILE = REPO_ROOT / "scripts" / "id_aliases.json"
|
||
TRACE_SCRIPT = REPO_ROOT / "scripts" / "check_test_traceability.py"
|
||
|
||
# 候选池：任何「1~4 位大写字母 + 连字符 + 数字」先进来，再判是否登记过。
|
||
# 三段式（TC-NET-1501 / TRAP-NET-18）单独一条——中段可能含数字（GD5F），
|
||
# 只用两段式正则会截出 "NET-1501" 这种假前缀。
|
||
CAND_MOD = re.compile(r"(?<![\w./\\-])([A-Z][A-Z0-9]{0,3})-([A-Z][A-Z0-9]{0,9})-(\d{1,4}[A-Za-z]?)(?![\w-])")
|
||
CAND_HYPH = re.compile(r"(?<![\w./\\-])([A-Z][A-Z0-9]{0,3})-(\d{1,4}[A-Za-z]?)(?![\w-])")
|
||
CAND_BARE = re.compile(r"(?<![\w.\-/])([A-Z])(?:\.(\d))?(1[0-9]|[0-9])(?![\w.\-])")
|
||
|
||
# 紧邻窗口（与 check_doc_links.py 的 SEC_MAX_GAP / SEC_GAP_CHARS 同口径）
|
||
ADJ_WINDOW = 60
|
||
# 第 8 项的容差（百分点）：比例的分母随"这一批改了几篇文档"漂移，0.1pp 以内不算回退
|
||
RATIO_TOL = 0.1
|
||
|
||
|
||
def die(msg):
|
||
print("[ERR] " + msg, file=sys.stderr)
|
||
return 2
|
||
|
||
|
||
def load_json(path, what):
|
||
if not path.is_file():
|
||
raise SystemExit(die("缺少 %s（%s）" % (what, path.relative_to(REPO_ROOT))))
|
||
try:
|
||
return json.loads(path.read_text(encoding="utf-8"))
|
||
except (OSError, ValueError) as exc:
|
||
raise SystemExit(die("无法解析 %s：%s" % (path.relative_to(REPO_ROOT), exc)))
|
||
|
||
|
||
class Finding(object):
|
||
__slots__ = ("code", "severity", "rel", "lineno", "literal", "msg", "hint")
|
||
|
||
def __init__(self, code, severity, rel, lineno, literal, msg, hint=""):
|
||
self.code = code
|
||
self.severity = severity
|
||
self.rel = rel
|
||
self.lineno = lineno
|
||
self.literal = literal
|
||
self.msg = msg
|
||
self.hint = hint
|
||
|
||
|
||
def alias_hit(ck, literal, rel):
|
||
"""别名命中判定：带 `scope` 的条目只在所列路径内生效。
|
||
|
||
为什么要限定作用域——同一旧字面量可能同时是**两个族的旧号**：实测 `S3` 在
|
||
scripts/check_coding_standard.py 与 docs/代码规范/ 下是「代码规范条款」旧号，
|
||
在 simulator/tests 与 docs/UI设计/ 下是「UI 像素回归状态」旧号。认成全局别名就会把
|
||
另一族的真悬空引用洗成「已解析」——那等于用工具复制 §1.2 记的 N-31/N-32 撞号事故。
|
||
"""
|
||
for ent in ck.alias_map.get(literal, []):
|
||
scope = ent.get("scope")
|
||
if not scope or any(fnmatch.fnmatch(rel, g) for g in scope):
|
||
return True
|
||
return False
|
||
|
||
|
||
class Checker(object):
|
||
def __init__(self, registry, aliases, verbose=False):
|
||
self.reg = registry
|
||
self.alias_map = {}
|
||
for lit, ent in ((aliases or {}).get("aliases", {})).items():
|
||
# 同一旧字面量可能有**多条 scoped 别名**（实测 `S3` 既是被改名的代码规范条款、
|
||
# 又是被改名的 UI 像素状态）→ 统一成列表，命中判定时按 scope 选条
|
||
self.alias_map[lit] = ent if isinstance(ent, list) else [ent]
|
||
self.verbose = verbose
|
||
self.excl = tuple(registry.get("excluded_paths", []))
|
||
self.excl_suffix = tuple(registry.get("excluded_suffixes", []))
|
||
self.owner_excl = list(registry.get("owner_docs_exclude", []))
|
||
self.noise = [re.compile(p) for p in registry.get("noise_literals", [])]
|
||
self.spec_docs = set(registry.get("spec_docs", []))
|
||
self.bare_letters = set(registry.get("bare_check_letters", []))
|
||
self.bare_forbidden = registry.get("bare_forbidden", {})
|
||
self.bare_exempt_files = set(registry.get("bare_exempt_files", []))
|
||
self.malformed = [(m["name"], re.compile(m["pattern"]), m["severity"], m["why"],
|
||
bool(m.get("docs_only")))
|
||
for m in registry.get("malformed", [])]
|
||
self.families = []
|
||
for fam in registry["families"]:
|
||
f = dict(fam)
|
||
f["_re"] = re.compile(fam["pattern"])
|
||
f["_full"] = re.compile("^(?:%s)$" % fam["pattern"])
|
||
f["_owners"] = self._expand_owners(fam)
|
||
f["_max"] = self._cursor_max(fam)
|
||
self.families.append(f)
|
||
# 长形态优先匹配：TC-UPPER 要先于 TC，避免大写后缀被当成合规
|
||
self.families.sort(key=lambda x: (-len(x["pattern"]), x["name"]))
|
||
self.findings = []
|
||
self.stats = Counter()
|
||
self.form_fired = set()
|
||
self.refs = {} # literal -> set(rel)，属主文档之外的引用处
|
||
self.owner_tokens = defaultdict(set) # family -> 属主文档里出现的字面量（=定义集）
|
||
self.file_cache = {}
|
||
self.owner_docs = set()
|
||
|
||
# ---------- 属主登记处展开 ----------
|
||
def _expand_owners(self, fam):
|
||
paths = list(fam.get("owner_docs", []))
|
||
for g in fam.get("owner_docs_glob", []):
|
||
for hit in REPO_ROOT.glob(g):
|
||
if hit.is_file():
|
||
paths.append(hit.relative_to(REPO_ROOT).as_posix())
|
||
# 属主文档本身不得落在排除目录（位号/芯片手册/生成物不算属主）
|
||
# 另：模板规范类文档里的 `DP-01` / `TC-XXX` 是**写法示例**，不是条目定义
|
||
out = [p for p in paths if not p.startswith(self.excl)]
|
||
for e in self.owner_excl:
|
||
out = [p for p in out if p != e]
|
||
return sorted(set(out))
|
||
|
||
def _cursor_max(self, fam):
|
||
cur = fam.get("cursor")
|
||
if not cur:
|
||
return None
|
||
m = re.search(r"([0-9]+)\s*$", str(cur))
|
||
return int(m.group(1)) if m else None
|
||
|
||
# ---------- 文件枚举（以 git 索引为准，生成物天然挡在外面）----------
|
||
def iter_files(self):
|
||
try:
|
||
out = subprocess.run(["git", "ls-files", "-z"], cwd=str(REPO_ROOT),
|
||
capture_output=True, check=True).stdout
|
||
files = out.decode("utf-8", "replace").split("\0")
|
||
except (OSError, subprocess.CalledProcessError):
|
||
files = [p.relative_to(REPO_ROOT).as_posix() for p in REPO_ROOT.rglob("*")
|
||
if p.is_file()]
|
||
for f in files:
|
||
if not f or f.startswith(self.excl) or f.endswith(self.excl_suffix):
|
||
continue
|
||
if "/__pycache__/" in ("/" + f) or f.endswith(".pyc"):
|
||
continue
|
||
if (REPO_ROOT / f).is_file():
|
||
yield f
|
||
|
||
def read(self, rel):
|
||
if rel not in self.file_cache:
|
||
try:
|
||
self.file_cache[rel] = (REPO_ROOT / rel).read_text(
|
||
encoding="utf-8", errors="replace")
|
||
except OSError as exc:
|
||
print("[WARN] 读取失败 %s：%s" % (rel, exc), file=sys.stderr)
|
||
self.file_cache[rel] = ""
|
||
return self.file_cache[rel]
|
||
|
||
# ---------- 主流程（两遍：先建定义集，再判引用）----------
|
||
def run(self):
|
||
rels = list(self.iter_files())
|
||
if not rels:
|
||
return die("一个文件都没扫到（git ls-files 返回空？）")
|
||
self.stats["files"] = len(rels)
|
||
for rel in rels:
|
||
self.read(rel)
|
||
for fam in self.families:
|
||
for owner in fam["_owners"]:
|
||
self.owner_docs.add(owner)
|
||
self.owner_tokens[fam["name"]].update(
|
||
m.group(0) for m in fam["_re"].finditer(self.read(owner)))
|
||
self.defined_all = set()
|
||
for s in self.owner_tokens.values():
|
||
self.defined_all |= s
|
||
for rel in rels:
|
||
for i, line in enumerate(self.read(rel).splitlines(), 1):
|
||
self.scan_line(rel, i, line)
|
||
self.check_collisions()
|
||
self.check_orphans()
|
||
return 0
|
||
|
||
def _is_noise(self, literal):
|
||
return any(rx.match(literal) for rx in self.noise)
|
||
|
||
def scan_line(self, rel, lineno, line):
|
||
if "[示例]" in line or "[计划]" in line:
|
||
# 报错样例 / 尚未创建的条目：与 check_doc_links.py 共用同一套豁免约定（§5.2）
|
||
self.stats["skip_example"] += len(CAND_HYPH.findall(line))
|
||
self._scan_malformed(rel, lineno, line, skip=True)
|
||
return
|
||
spans = []
|
||
for m in CAND_MOD.finditer(line):
|
||
spans.append((m.start(), m.end(), m.group(0), True))
|
||
covered = [(a, b) for a, b, _, _ in spans]
|
||
for m in CAND_HYPH.finditer(line):
|
||
if any(a <= m.start() and m.end() <= b for a, b in covered):
|
||
continue
|
||
spans.append((m.start(), m.end(), m.group(0), True))
|
||
for m in CAND_BARE.finditer(line):
|
||
if m.group(1) not in self.bare_letters:
|
||
continue
|
||
if any(a <= m.start() and m.end() <= b for a, b in covered):
|
||
continue
|
||
lit = m.group(1) + (("." + m.group(2)) if m.group(2) else "") + m.group(3)
|
||
if "." in lit:
|
||
continue
|
||
spans.append((m.start(), m.end(), lit, False))
|
||
spans.sort()
|
||
|
||
seen_here = set()
|
||
for start, _end, literal, hyphenated in spans:
|
||
if (literal, start) in seen_here:
|
||
continue
|
||
seen_here.add((literal, start))
|
||
self.stats["refs_total"] += 1
|
||
fams = [f for f in self.families if f["_full"].match(literal)]
|
||
if not fams:
|
||
if self._is_noise(literal):
|
||
self.stats["noise_allowed"] += 1
|
||
continue
|
||
if not hyphenated:
|
||
# 裸形盲区（§1.2 同字母多义 / §4.4 位号同形）：显式不判，见 unmanaged_bare
|
||
self.stats["skip_bare_unmanaged"] += 1
|
||
continue
|
||
if re.match(r"^P[0-3]-\d", literal):
|
||
self.add(2, "error", rel, lineno, literal,
|
||
"档位词带后缀：P0/P1/P2/P3 是严重度与优先级档位词，"
|
||
"永不带 -<数字>（§4.4）",
|
||
"只写 `P1`，严重度进表格列（§2 P-2）")
|
||
self.stats["undecidable"] += 1
|
||
else:
|
||
self.add(1, "error", rel, lineno, literal,
|
||
"未登记前缀：%s（§4.1 / §4.2 总表里没有这一族）" % literal,
|
||
"归到既有类别；确属新类别先改 docs/编号规范.md §4，再进 registry")
|
||
self.stats["undecidable"] += 1
|
||
continue
|
||
|
||
fam = self._choose_family(literal, fams, rel, line, start)
|
||
if fam["_owners"]:
|
||
self.stats["decidable"] += 1
|
||
else:
|
||
self.stats["undecidable"] += 1
|
||
self._check_form(rel, lineno, literal, fam)
|
||
self._check_frozen(rel, lineno, literal, fam)
|
||
self._check_dangling(rel, lineno, literal, fam)
|
||
self._check_adjacent(rel, lineno, literal, fam, line, start)
|
||
if rel not in fam["_owners"]:
|
||
self.refs.setdefault(literal, set()).add(rel)
|
||
|
||
self._scan_malformed(rel, lineno, line)
|
||
|
||
def _choose_family(self, literal, fams, rel, line, start):
|
||
"""同一字面量可能落进多族（登记簿 N-16 vs 报告 N-16）：属主文档 > 紧邻属主 > 全局族。"""
|
||
if len(fams) == 1:
|
||
return fams[0]
|
||
for f in fams:
|
||
if rel in f["_owners"]:
|
||
return f
|
||
for f in fams:
|
||
if f.get("adjacent_required") and self._has_adjacent_cue(f, line, start):
|
||
return f
|
||
for f in fams:
|
||
if f.get("check_dangling") and f["_owners"]:
|
||
return f
|
||
for f in fams:
|
||
if f["scope"] == "global":
|
||
return f
|
||
return fams[0]
|
||
|
||
def _cues(self, fam):
|
||
cues = list(fam.get("adjacent_cues", []))
|
||
for o in fam.get("owner_docs", []):
|
||
stem = o.rsplit("/", 1)[-1]
|
||
cues += [stem, stem.rsplit(".", 1)[0]]
|
||
for o in fam["_owners"][:6]:
|
||
stem = o.rsplit("/", 1)[-1]
|
||
cues += [stem, stem.rsplit(".", 1)[0]]
|
||
out, seen = [], set()
|
||
for c in cues:
|
||
if c and len(c) >= 2 and c not in seen:
|
||
seen.add(c)
|
||
out.append(c)
|
||
return out
|
||
|
||
def _has_adjacent_cue(self, fam, line, start):
|
||
window = line[max(0, start - ADJ_WINDOW):start]
|
||
for c in self._cues(fam):
|
||
if c in window:
|
||
return True
|
||
return False
|
||
|
||
# ---------- 单项校验 ----------
|
||
def add(self, code, severity, rel, lineno, literal, msg, hint=""):
|
||
self.findings.append(Finding(code, severity, rel, lineno, literal, msg, hint))
|
||
self.stats["hit_%d" % code] += 1
|
||
if code == 2:
|
||
# 同一处的形态违规不再重复报悬空（同一个毛病报两次，会把 backlog 冲淡）
|
||
self.form_fired.add((rel, lineno, literal))
|
||
|
||
def _check_form(self, rel, lineno, literal, fam):
|
||
if rel in self.spec_docs:
|
||
# 规范正文**必须**引用非法形态才能说明什么叫非法（`Trap 07`、`P1-8`…），
|
||
# 在规范里判"形态不合规"等于让规范判死自己；其余文档照判。
|
||
return
|
||
# 裸形判定按**字面量本身**有无连字符——不能看族正则：`[0-4]` 这类字符类里
|
||
# 就带连字符，按族判断会把 S-REVIEW 误认成"连字符族"（实测踩到过）。
|
||
if "-" not in literal:
|
||
self._check_bare_form(rel, lineno, literal, fam)
|
||
digits = re.findall(r"([0-9]+)$", literal)
|
||
width = fam.get("digit_width")
|
||
if digits and width:
|
||
lo, hi = width
|
||
if not (lo <= len(digits[0]) <= hi):
|
||
self.add(2, "error", rel, lineno, literal,
|
||
"形态不合规：%s 的序号应为 %d~%d 位（§4.1 左补零，P-4）"
|
||
% (literal, lo, hi), "补零后引用，属主登记处一并改")
|
||
if fam.get("suffix") == "lower" and re.search(r"[A-Z]$", literal):
|
||
self.add(2, "error", rel, lineno, literal,
|
||
"形态不合规：%s 后缀必须是小写字母（§4.1）" % literal,
|
||
"改 %s%s" % (literal[:-1], literal[-1].lower()))
|
||
if fam.get("suffix") == "none" and re.search(r"[A-Za-z]$", literal):
|
||
pass
|
||
if fam.get("malformed"):
|
||
self.add(2, "error", rel, lineno, literal,
|
||
"形态不合规：%s —— %s（§4.1 / §6）" % (literal, fam["malformed"]),
|
||
"按规范形态改写")
|
||
|
||
def _check_bare_form(self, rel, lineno, literal, fam):
|
||
"""裸形：S3 之于 S-3（§2 P-4 禁止裸形）。两个条件同时成立才报，避免把
|
||
「历史裸形族在属主文档里被引用」也算成违规：
|
||
① 该裸形字面量**全库无任何属主登记处定义**（在案的历史裸形引用不算违规，
|
||
它们由校验 3 以告警形式处理）；
|
||
② 它的连字符孪生形（S-3）确有在册定义——说明作者是在写一个登记簿条目，
|
||
只是漏了连字符。也不在裸形豁免文件里。"""
|
||
m = re.match(r"^([A-Z])([0-9]{1,2})$", literal)
|
||
if not m:
|
||
return
|
||
letter = m.group(1)
|
||
if letter not in self.bare_forbidden or rel in self.bare_exempt_files:
|
||
return
|
||
hyph = "%s-%s" % (letter, m.group(2))
|
||
hyph_defined = any(hyph in self.owner_tokens[f["name"]] for f in self.families)
|
||
if hyph_defined and literal not in self.defined_all:
|
||
self.add(2, "error", rel, lineno, literal,
|
||
"形态不合规：裸形 %s 是 %s 的非法写法（§2 P-4；%s）"
|
||
% (literal, hyph, self.bare_forbidden[letter]),
|
||
"写全 %s，或改语义标识（§3.3 / P-8）" % hyph)
|
||
|
||
def _check_frozen(self, rel, lineno, literal, fam):
|
||
if not fam.get("frozen") or fam["_max"] is None:
|
||
return
|
||
if fam.get("cursor_owner_only") and rel not in fam["_owners"]:
|
||
# 裸形族（B5/C6/D7/M1/U3/R3/V5…）与位号、档位词、其它文档的同名清单大面积同形
|
||
# （§1.2 五义/四义）→ 越游标只在属主登记处文档内判，文档外不判（见 unmanaged_bare）
|
||
return
|
||
nums = re.findall(r"([0-9]+)", literal.rsplit("-", 1)[-1])
|
||
if not nums:
|
||
return
|
||
if int(nums[-1]) > fam["_max"]:
|
||
self.add(5, "error", rel, lineno, literal,
|
||
"冻结前缀越游标：%s > %s（%s；§4.3 / §8 校验 5）"
|
||
% (literal, fam["cursor"], fam["title"]),
|
||
"该族已封闭，新条目改语义标识（§3.3）；确需新号先改 docs/编号规范.md §4")
|
||
|
||
def _check_dangling(self, rel, lineno, literal, fam):
|
||
if not fam.get("check_dangling") or not fam["_owners"] or rel in fam["_owners"]:
|
||
return
|
||
if rel in self.spec_docs:
|
||
# 规范正文里的 `N-99`→`N-100`、`Trap 07` 这类是**写法示例**，不是引用
|
||
self.stats["skip_spec_examples"] += 1
|
||
return
|
||
if (rel, lineno, literal) in self.form_fired:
|
||
return
|
||
if literal in self.defined_all:
|
||
return
|
||
if alias_hit(self, literal, rel) or literal in set(fam.get("allow_missing_defs", [])):
|
||
self.stats["alias_hits"] += 1
|
||
return
|
||
sev = fam.get("dangling_outside_owner", "error") if rel not in fam["_owners"] else "error"
|
||
self.add(3, sev, rel, lineno, literal,
|
||
"悬空引用：%s 在属主登记处（%s）查无定义，也不在别名表里（§8 校验 3）"
|
||
% (literal, " / ".join(fam["_owners"][:2])),
|
||
"补定义行，或删引用（§5.5：销项 = 删行 + 清全库引用）")
|
||
|
||
def _check_adjacent(self, rel, lineno, literal, fam, line, start):
|
||
if not fam.get("adjacent_required") or rel in fam["_owners"] or rel in self.spec_docs:
|
||
return
|
||
if "-" not in literal:
|
||
# 只约束连字符形态的文档内编号：裸形（A2/B5/U3/D7）与位号、档位词大面积同形
|
||
# （§1.2 五义/四义清单），全库要求紧邻文档名属 §9 阶段 1-③ 的改写批次。
|
||
return
|
||
if self._has_adjacent_cue(fam, line, start):
|
||
return
|
||
owner = fam["owner_docs"][0] if fam.get("owner_docs") else (
|
||
fam["_owners"][0] if fam["_owners"] else "docs/<属主文档>.md")
|
||
self.add(6, "error", rel, lineno, literal,
|
||
"文档内编号未紧邻文档名：%s 属「%s」，跨文档引用必须写全属主（§5.2 / §8 校验 6）"
|
||
% (literal, fam["title"]),
|
||
"写成 `%s` %s（或 `%s NN`）" % (owner, literal, literal))
|
||
|
||
def _scan_malformed(self, rel, lineno, line, skip=False):
|
||
is_md = rel.endswith(".md")
|
||
if rel in self.spec_docs and not skip:
|
||
# 同上：规范正文就是靠引用 `Trap 07` / `陷阱 1` / `gen 181` 这些非法形态
|
||
# 才讲得清"什么算非法"，在它自己头上判形态违规没有意义
|
||
return
|
||
for name, rx, sev, why, docs_only in self.malformed:
|
||
if (docs_only and not is_md) or (skip and name != "空格形 Trap"):
|
||
continue
|
||
for m in rx.finditer(line):
|
||
self.add(2, sev, rel, lineno, m.group(0).strip(),
|
||
"形态不合规：%s —— %s" % (name, why), "按 §7.1 迁移表改写")
|
||
self.stats["malformed_%s" % name] += 1
|
||
|
||
# ---------- 校验 4 / 7 ----------
|
||
# 定义位：字面量出现在**表格首格 / 标题行 / 列表项行首**——正文里"提到过"不算定义，
|
||
# 否则任何交叉引用都会被误判成重复定义（§8 校验 4 要抓的是"两处各建一条目"）。
|
||
DEF_ANCHOR = re.compile(r"^\s*(?:\|\s*\*{0,2}|#{1,6}\s+|[-*]\s+\*{0,2})"
|
||
r"([A-Z][A-Z0-9]*(?:-[A-Za-z0-9]+)+)")
|
||
|
||
def check_collisions(self):
|
||
"""同一字面量被两个属主重复定义（N-16 类事故 / §1.2 的 Trap 02 跨文档同号不同义）。
|
||
规范 §4.3 已声明的同号不同义族对由 `collides_with` 登记，不计为事故。"""
|
||
allowed = set()
|
||
for fam in self.families:
|
||
cw = fam.get("collides_with")
|
||
if cw:
|
||
allowed.add(frozenset((fam["class"], cw)))
|
||
defs = defaultdict(list) # literal -> [(class, family_title, owner_doc)]
|
||
for fam in self.families:
|
||
for owner in fam["_owners"]:
|
||
for line in self.read(owner).splitlines():
|
||
m = self.DEF_ANCHOR.match(line)
|
||
if m and fam["_full"].match(m.group(1)):
|
||
defs[m.group(1)].append((fam["class"], fam["title"], owner))
|
||
for lit, hits in sorted(defs.items()):
|
||
classes = {c for c, _, _ in hits}
|
||
docs = {d for _, _, d in hits}
|
||
pairs = [(a, b) for a in classes for b in classes if a < b]
|
||
clash = [p for p in pairs if frozenset(p) not in allowed]
|
||
if clash:
|
||
where = sorted(docs)[:2]
|
||
self.add(4, "error", where[0], 0, lit,
|
||
"重复定义：%s 同时被 %s 两个属主登记处定义（%s；§8 校验 4，"
|
||
"N-16 类事故）" % (lit, " / ".join(sorted(classes)),
|
||
", ".join(d.rsplit("/", 1)[-1] for d in where)),
|
||
"保留单一属主；另一处改语义标识（§3.3）或引用紧邻文档名（§5.2）")
|
||
elif len(classes) == 1 and len(docs) >= 2:
|
||
self.add(4, "error", sorted(docs)[0], 0, lit,
|
||
"重复定义：%s 在同一族「%s」的 %d 篇属主文档里各定义一次（%s；"
|
||
"§8 校验 4，§1.2 的 Trap 02 跨文档同号不同义类事故）"
|
||
% (lit, hits[0][1], len(docs),
|
||
", ".join(sorted(d.rsplit("/", 1)[-1] for d in docs)[:3])),
|
||
"只在一处定义；另一处改引用，或按 §6 补模块码把两族拆开")
|
||
else:
|
||
self.stats["collision_allowed"] += 1
|
||
|
||
def check_orphans(self):
|
||
for fam in self.families:
|
||
if not fam.get("check_dangling") or not fam["_owners"]:
|
||
continue
|
||
for lit in sorted(self.owner_tokens[fam["name"]]):
|
||
outside = {r for r in self.refs.get(lit, set()) if r not in fam["_owners"]}
|
||
if not outside:
|
||
self.add(7, "warn", fam["_owners"][0], 0, lit,
|
||
"孤儿条目：%s 定义在 %s，全库无引用（§8 校验 7）"
|
||
% (lit, fam["_owners"][0]),
|
||
"确认已闭环可销项（§5.5），或补上引用")
|
||
|
||
# ---------- 报告 ----------
|
||
TITLES = {
|
||
1: "校验 1 未登记前缀（自造编号）",
|
||
2: "校验 2 形态不合规（裸形/空格形/中文形/补零/档位词带后缀）",
|
||
3: "校验 3 悬空引用（查无定义且不在别名表）",
|
||
4: "校验 4 重复定义（同一字面量两个属主）",
|
||
5: "校验 5 冻结前缀越游标",
|
||
6: "校验 6 文档内编号未紧邻文档名",
|
||
7: "校验 7 孤儿条目（定义了但无引用）",
|
||
}
|
||
|
||
def report(self, top, only):
|
||
by_code = defaultdict(list)
|
||
for f in self.findings:
|
||
by_code[f.code].append(f)
|
||
print("=== 编号门禁 check_ids.py（docs/编号规范.md §8）===")
|
||
print("扫描 %d 个文件；编号字面量 %d 处（豁免 [示例]/[计划] 行 %d 处、"
|
||
"非编号噪声 %d 处、裸形盲区 %d 处、命中别名表 %d 处）"
|
||
% (self.stats["files"], self.stats["refs_total"], self.stats["skip_example"],
|
||
self.stats["noise_allowed"], self.stats["skip_bare_unmanaged"],
|
||
self.stats["alias_hits"]))
|
||
print("已声明的同号不同义族对（§4.3）：%d 个字面量，不计入校验 4"
|
||
% self.stats["collision_allowed"])
|
||
|
||
for code in sorted(self.TITLES):
|
||
if only and code not in only:
|
||
continue
|
||
items = [f for f in by_code.get(code, [])]
|
||
n_err = sum(1 for f in items if f.severity == "error")
|
||
n_warn = len(items) - n_err
|
||
print("\n[%s] %s：%d 处（错误 %d / 告警 %d）"
|
||
% ("WARN" if not n_err else "FAIL", self.TITLES[code],
|
||
len(items), n_err, n_warn))
|
||
for f in items[:top]:
|
||
loc = "%s:%d" % (f.rel, f.lineno) if f.lineno else f.rel
|
||
# 带上校验号 + 级别：--top 打出来的每一行都能独立追溯回规范 §8 的表
|
||
print(" [%d-%s] %s %s" % (f.code, "E" if f.severity == "error" else "W",
|
||
loc, f.msg))
|
||
if self.verbose and f.hint:
|
||
print(" → %s" % f.hint)
|
||
if len(items) > top:
|
||
print(" … 其余 %d 处略（--top N 多打；--verbose 带修法）" % (len(items) - top))
|
||
|
||
total = self.stats["refs_total"]
|
||
# errs 恒按**全部** findings 统计：--only 是"局部排查"用的，不能因为它少扫两类
|
||
# 就把门禁结论报成"无错误"（真踩过的坑：--only 1,3 让 --strict 误报绿）
|
||
errs = sum(1 for f in self.findings
|
||
if f.severity == "error" and f.code != 8)
|
||
if only:
|
||
errs = sum(1 for f in self.findings
|
||
if f.severity == "error" and f.code in only)
|
||
print("\n⚠ --only %s：本次只跑部分校验，**不构成门禁结论**"
|
||
% ",".join(str(c) for c in sorted(only)))
|
||
dec = self.stats["decidable"]
|
||
ratio = (100.0 * dec / total) if total else 100.0
|
||
base = self.reg.get("decidable_ratio_baseline")
|
||
print("\n[比例] 校验 8 编号引用可判定 %d/%d = %.2f%%（基线 %s，容差 %.1fpp）"
|
||
% (dec, total, ratio,
|
||
"未建档" if base is None else "%.1f%%" % base, RATIO_TOL))
|
||
print(" 可判定 = 字面量落进已登记前缀族**且该族属主登记处已存在**；"
|
||
"未登记前缀、形态违规、属主登记处还是 [计划] 的都算不可判定 → 只许涨不许跌")
|
||
return ratio, errs
|
||
|
||
|
||
def parse_only(s):
|
||
if not s:
|
||
return set()
|
||
return {int(x) for x in re.split(r"[,\s]+", s) if x.strip().isdigit()}
|
||
|
||
|
||
def main():
|
||
ap = argparse.ArgumentParser(
|
||
description="编号门禁：docs/编号规范.md §8 的 8 项校验（默认 warn-only）")
|
||
ap.add_argument("--strict", action="store_true",
|
||
help="校验 1~6 的错误改为拦提交（默认 warn-only，不拦）")
|
||
ap.add_argument("--verbose", action="store_true", help="每条附带修法")
|
||
ap.add_argument("--top", type=int, default=12, help="每类校验打印的条数（默认 12）")
|
||
ap.add_argument("--only", default="", help="只跑指定校验，如 --only 1,3,5")
|
||
ap.add_argument("--update-baseline", action="store_true",
|
||
help="把本次可判定比例写进 id_registry.json 作为基线（只许涨）")
|
||
ap.add_argument("--no-trace", action="store_true",
|
||
help="跳过顺带收编的 TC 双向追溯（check_test_traceability.py）")
|
||
args = ap.parse_args()
|
||
|
||
reg = load_json(REGISTRY_FILE, "编号注册表")
|
||
aliases = load_json(ALIASES_FILE, "编号别名表")
|
||
|
||
ck = Checker(reg, aliases, verbose=args.verbose)
|
||
rc = ck.run()
|
||
if rc:
|
||
return rc
|
||
ratio, errs = ck.report(args.top, parse_only(args.only))
|
||
|
||
if args.update_baseline:
|
||
base = reg.get("decidable_ratio_baseline")
|
||
if base is not None and ratio + RATIO_TOL < base:
|
||
print("[ERR] 拒绝下调基线：本次 %.2f%% < 基线 %.2f%%（§8 校验 8「只许涨不许跌」）"
|
||
% (ratio, base), file=sys.stderr)
|
||
return 1
|
||
# 基线**下取整到 0.1 个百分点**：小数点第 2 位随"这一批改了哪几篇文档"抖动，
|
||
# 用 round() 会把 82.34 进位成 82.3 再反过来把自己判死（实测踩过），故 floor。
|
||
reg["decidable_ratio_baseline"] = math.floor(ratio * 10) / 10.0
|
||
REGISTRY_FILE.write_text(json.dumps(reg, ensure_ascii=False, indent=2) + "\n",
|
||
encoding="utf-8", newline="\n")
|
||
print("[OK] 基线已写进 scripts/id_registry.json：%.1f%%（本次实测 %.2f%%）"
|
||
% (reg["decidable_ratio_baseline"], ratio))
|
||
|
||
if not args.no_trace and TRACE_SCRIPT.is_file():
|
||
print("\n=== 校验 8 附带：TC 用例双向追溯（顺带收编 check_test_traceability.py）===")
|
||
try:
|
||
# 子进程的 stdout 在管道下默认按本地代码页（936/GBK）编码 → 必须显式压成
|
||
# UTF-8，否则这里按 utf-8 解出来全是乱码（实测踩到：[DOC-ONLY] 行打印成 mojibake）
|
||
env = dict(os.environ, PYTHONIOENCODING="utf-8")
|
||
p = subprocess.run([sys.executable, str(TRACE_SCRIPT)], cwd=str(REPO_ROOT),
|
||
capture_output=True, text=True, encoding="utf-8",
|
||
errors="replace", env=env)
|
||
for ln in (p.stdout or "").splitlines():
|
||
if ln.startswith(("代码侧", "文档侧", "交集", "[DOC-ONLY]", "[CODE-ONLY]",
|
||
"[OK]", "[WARN]", "=== ")):
|
||
print(" " + ln)
|
||
if p.returncode == 2:
|
||
print("[WARN] 追溯脚本环境错误（退出码 2），不影响本门禁结论", file=sys.stderr)
|
||
except OSError as exc:
|
||
print("[WARN] 无法运行追溯脚本：%s" % exc, file=sys.stderr)
|
||
|
||
base = reg.get("decidable_ratio_baseline")
|
||
if base is not None and ratio + RATIO_TOL < base and not args.update_baseline:
|
||
print("\n[FAIL] 校验 8：可判定比例跌破基线 %.2f%% < %.2f%%（只许涨不许跌，容差 %.1fpp）"
|
||
% (ratio, base, RATIO_TOL))
|
||
return 1
|
||
if errs:
|
||
if args.strict:
|
||
print("\n[FAIL] 编号门禁：%d 处错误级问题（--strict 拦提交）" % errs)
|
||
return 1
|
||
print("\n[WARN] 编号 backlog：%d 处错误级问题，warn-only 不拦提交"
|
||
"（§8 上线节奏：清账后加 --strict）" % errs)
|
||
else:
|
||
print("\n[OK] 校验 1~6 无错误")
|
||
return 0
|
||
|
||
|
||
if __name__ == "__main__":
|
||
sys.exit(main())
|