dsh-git-push 2.20.18 → 3.0.0
This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
- package/README.md +3 -2
- package/cli-tool.mjs +2 -4
- package/cli.mjs +36 -2
- package/lib/app/apply.js +9 -9
- package/lib/app/command-registry.js +44 -20
- package/lib/app/handlers/meta.js +7 -142
- package/lib/app/handlers/npm.js +18 -1
- package/lib/app/handlers/preprocess.js +107 -0
- package/lib/app/handlers/repo-actions.js +2 -2
- package/lib/app/http-handlers.js +25 -12
- package/lib/app/index.js +2 -1
- package/lib/app/inject-text.js +3 -2
- package/lib/app/preprocess-client.js +136 -0
- package/lib/app/repo-ownership.js +1 -1
- package/lib/app/schema.js +3 -3
- package/lib/app/severity.js +14 -0
- package/lib/app/slash-commands.js +5 -164
- package/lib/app/tool-call.js +75 -140
- package/lib/app/tools.js +13 -6
- package/lib/cli/commands-preprocess.mjs +74 -0
- package/lib/cli/commands-tool.mjs +3 -2
- package/lib/cli/commands-vcs.mjs +4 -4
- package/lib/cli/shared.mjs +42 -13
- package/lib/client-parts/git/controller.js +2 -2
- package/lib/client.js +2 -2
- package/lib/commit-push/pre-commit-engine.js +57 -10
- package/lib/commit-push.js +125 -36
- package/lib/{exempt/index.js → git/exempt-markers.js} +8 -1
- package/lib/git/ignore.js +1 -1
- package/lib/git/repos.js +2 -2
- package/lib/git/sensitive.js +1 -1
- package/lib/npm/publish.js +20 -5
- package/lib/plugin/auto-push.js +2 -2
- package/lib/preprocess/providers.js +191 -0
- package/lib/self/index.js +1 -1
- package/lib/skip-dirs.js +14 -55
- package/package.json +29 -37
- package/scripts/build-client.mjs +2 -0
- package/scripts/check.mjs +4 -2
- package/scripts/local-registry.mjs +2 -1
- package/skills/dev/dsh-git-push.md +5 -0
- package/cli-audit.mjs +0 -141
- package/lib/app/audit-api.js +0 -173
- package/lib/app/host-doc-checks.js +0 -25
- package/lib/app/scope-check.js +0 -69
- package/lib/ast/brace.js +0 -120
- package/lib/ast/callgraph.js +0 -135
- package/lib/ast/code-lines.js +0 -175
- package/lib/ast/consts.js +0 -12
- package/lib/ast/control-flow.js +0 -470
- package/lib/ast/credential.js +0 -315
- package/lib/ast/dataflow.js +0 -391
- package/lib/ast/dup-code.js +0 -165
- package/lib/ast/fs-hotspots.js +0 -154
- package/lib/ast/index.js +0 -38
- package/lib/ast/io-risk-const.js +0 -95
- package/lib/ast/io-risk-fn.js +0 -208
- package/lib/ast/io-risk-loop.js +0 -308
- package/lib/ast/io-risk.js +0 -314
- package/lib/ast/lang.js +0 -156
- package/lib/ast/line-count.js +0 -52
- package/lib/ast/magic-number.js +0 -272
- package/lib/ast/naming.js +0 -292
- package/lib/ast/perf.js +0 -488
- package/lib/ast/scope-collisions.js +0 -228
- package/lib/ast/scope.js +0 -241
- package/lib/ast/shell.js +0 -131
- package/lib/ast/size.js +0 -448
- package/lib/ast/symbol-index.js +0 -399
- package/lib/ast/tokenizer.js +0 -421
- package/lib/ast/treesitter.js +0 -413
- package/lib/audit/analysis-coverage.js +0 -98
- package/lib/audit/audit-file.js +0 -205
- package/lib/audit/auditignore-rules.js +0 -148
- package/lib/audit/checks.js +0 -17
- package/lib/audit/client-entry.js +0 -93
- package/lib/audit/collector.js +0 -610
- package/lib/audit/ext-runner.js +0 -80
- package/lib/audit/file-context.js +0 -391
- package/lib/audit/finding.js +0 -101
- package/lib/audit/gitignore-match.js +0 -113
- package/lib/audit/glob.js +0 -63
- package/lib/audit/history-report.js +0 -100
- package/lib/audit/history.js +0 -126
- package/lib/audit/ignore-blind.js +0 -102
- package/lib/audit/index.js +0 -33
- package/lib/audit/orchestrate.js +0 -672
- package/lib/audit/public.js +0 -24
- package/lib/audit/repo-level.js +0 -30
- package/lib/audit/report-yaml.js +0 -90
- package/lib/audit/scope-policy.js +0 -110
- package/lib/audit/slot.js +0 -66
- package/lib/audit-rules/audit-rules-comment.yml +0 -199
- package/lib/audit-rules/audit-rules-cpp.yml +0 -114
- package/lib/audit-rules/audit-rules-csharp.yml +0 -82
- package/lib/audit-rules/audit-rules-docs.yml +0 -32
- package/lib/audit-rules/audit-rules-dsh.yml +0 -177
- package/lib/audit-rules/audit-rules-filehealth.yml +0 -42
- package/lib/audit-rules/audit-rules-folder.yml +0 -99
- package/lib/audit-rules/audit-rules-frontend.yml +0 -335
- package/lib/audit-rules/audit-rules-go.yml +0 -103
- package/lib/audit-rules/audit-rules-i18n.yml +0 -45
- package/lib/audit-rules/audit-rules-java.yml +0 -105
- package/lib/audit-rules/audit-rules-nodejs.yml +0 -662
- package/lib/audit-rules/audit-rules-npm.yml +0 -187
- package/lib/audit-rules/audit-rules-paths.yml +0 -98
- package/lib/audit-rules/audit-rules-performance.yml +0 -245
- package/lib/audit-rules/audit-rules-php.yml +0 -101
- package/lib/audit-rules/audit-rules-private.yml +0 -24
- package/lib/audit-rules/audit-rules-python.yml +0 -106
- package/lib/audit-rules/audit-rules-robustness.yml +0 -31
- package/lib/audit-rules/audit-rules-rust.yml +0 -109
- package/lib/audit-rules/audit-rules-structure.yml +0 -158
- package/lib/audit-rules/audit-rules-swift.yml +0 -99
- package/lib/audit-rules/audit-rules-template.yml +0 -108
- package/lib/audit-ui-sdk/contract.js +0 -117
- package/lib/audit-ui-sdk/index.js +0 -17
- package/lib/checks/button-bind.js +0 -347
- package/lib/checks/common.js +0 -170
- package/lib/checks/credential-file.js +0 -43
- package/lib/checks/dataflow.js +0 -40
- package/lib/checks/dispatch.js +0 -117
- package/lib/checks/dup-code.js +0 -69
- package/lib/checks/dup-const.js +0 -232
- package/lib/checks/file-health.js +0 -150
- package/lib/checks/filter.js +0 -137
- package/lib/checks/folder.js +0 -173
- package/lib/checks/fs-scanner.js +0 -57
- package/lib/checks/hardcoded-path.js +0 -98
- package/lib/checks/index.js +0 -49
- package/lib/checks/io.js +0 -66
- package/lib/checks/magic-number.js +0 -42
- package/lib/checks/npm-json.js +0 -273
- package/lib/checks/private.js +0 -64
- package/lib/checks/regex.js +0 -298
- package/lib/checks/scope-collision.js +0 -68
- package/lib/checks/semantic.js +0 -134
- package/lib/checks/structural.js +0 -281
- package/lib/cli/commands-audit.mjs +0 -142
- package/lib/cli/commands-doc.mjs +0 -127
- package/lib/client-parts/audit/audit.js +0 -301
- package/lib/exempt/tool.js +0 -198
- package/lib/git/module-splitter.js +0 -47
- package/lib/link-check/index.js +0 -187
- package/lib/lint-adapter/detect.js +0 -161
- package/lib/lint-adapter/index.js +0 -50
- package/lib/lint-adapter/runner.js +0 -133
- package/lib/rule/compilers/credential.js +0 -62
- package/lib/rule/compilers/dataflow.js +0 -38
- package/lib/rule/compilers/file-health.js +0 -75
- package/lib/rule/compilers/frontend.js +0 -86
- package/lib/rule/compilers/func.js +0 -35
- package/lib/rule/compilers/helpers.js +0 -71
- package/lib/rule/compilers/index.js +0 -29
- package/lib/rule/compilers/link-semantic.js +0 -45
- package/lib/rule/compilers/list-kinds.js +0 -12
- package/lib/rule/compilers/numeric-kinds.js +0 -73
- package/lib/rule/compilers/regex.js +0 -78
- package/lib/rule/compilers/structure.js +0 -107
- package/lib/rule/compilers.js +0 -17
- package/lib/rule/homoglyph.js +0 -45
- package/lib/rule/loader.js +0 -274
- package/lib/rule/registry.js +0 -122
- package/lib/rule/scope.js +0 -67
- package/lib/score/docs-score.js +0 -166
- package/lib/score/index.js +0 -234
- package/scripts/__pycache__/module-splitter.cpython-38.pyc +0 -0
- package/scripts/audit-ext/_example-readme-present.mjs +0 -34
- package/scripts/audit-ext/dsh-inject-declarations.mjs +0 -209
- package/scripts/audit-ext/variable-min-length.mjs +0 -102
- package/scripts/audit-gate.mjs +0 -235
- package/scripts/audit-runner.mjs +0 -35
- package/scripts/audit-runtime-check.mjs +0 -129
- package/scripts/browser-page-probe.mjs +0 -271
- package/scripts/check-js-scope.mjs +0 -87
- package/scripts/clean-date-comments.py +0 -102
- package/scripts/cpu-sample.mjs +0 -188
- package/scripts/doc-drift.mjs +0 -200
- package/scripts/hot-path-probe-cjs.cjs +0 -71
- package/scripts/hot-path-probe-fs-shim.mjs +0 -28
- package/scripts/hot-path-probe-fsp-shim.mjs +0 -31
- package/scripts/hot-path-probe.mjs +0 -153
- package/scripts/module-splitter.py +0 -800
- package/scripts/picker-repro.mjs +0 -115
- package/scripts/platform-scan.mjs +0 -168
- package/scripts/probe-recheck.mjs +0 -47
- package/scripts/rename-locator.mjs +0 -97
- package/scripts/rule-switch.mjs +0 -97
- package/scripts/rules-solo-audit.mjs +0 -173
- package/scripts/scan-file-io.mjs +0 -717
- package/scripts/scan-fs-hotspots.mjs +0 -107
- package/scripts/symbol-index.mjs +0 -85
- package/scripts/verify-prestep.mjs +0 -100
- package/skills/dev/audit-rule-authoring.md +0 -58
- package/skills/guide/guide-audit.md +0 -270
- /package/lib/{audit → app}/ownership.js +0 -0
- /package/lib/app/{schema-audit.js → schema-preprocess.js} +0 -0
- /package/lib/{audit-defaults.js → git/defaults.js} +0 -0
|
@@ -1,150 +0,0 @@
|
|
|
1
|
-
/**
|
|
2
|
-
* 检查层 · 文件健康度
|
|
3
|
-
*
|
|
4
|
-
* 职责:按行数/字节数/最长行三维给文件打分,识别超大文件与超长行。
|
|
5
|
-
*/
|
|
6
|
-
|
|
7
|
-
import { makeFinding, CODE_EXTS } from '../audit/index.js';
|
|
8
|
-
import { checkFileLines, maxFunctionLength } from '../ast/index.js';
|
|
9
|
-
import { capSeverity } from './common.js';
|
|
10
|
-
import { tokenize } from '../ast/tokenizer.js';
|
|
11
|
-
import { codeLineCount as countCodeLines } from '../ast/line-count.js';
|
|
12
|
-
|
|
13
|
-
/* ───────────────────────── 默认参数(与 lib/rule/compilers/file-health.js 的默认值一致; ─────────────────────────
|
|
14
|
-
* 规则条目可用 snake_case 字段覆盖,未覆盖时用这里的兜底值。两处需同步修改。) */
|
|
15
|
-
|
|
16
|
-
/** 三维权重(合计 1.0):行数 / 大小 / 单行最大长度。 */
|
|
17
|
-
const WEIGHT_LINES = 0.4;
|
|
18
|
-
const WEIGHT_SIZE = 0.35;
|
|
19
|
-
const WEIGHT_LINE_LENGTH = 0.25;
|
|
20
|
-
/** 等级边界(各 4 个分隔点 → 0~4 级)。 */
|
|
21
|
-
const LINE_LEVELS = [200, 400, 800, 1500];
|
|
22
|
-
const SIZE_LEVELS = [30, 100, 300, 1000];
|
|
23
|
-
const LENGTH_LEVELS = [120, 200, 300, 500];
|
|
24
|
-
/** 等级扣分表(level 0~4)。 */
|
|
25
|
-
const LEVEL_PENALTIES = [0, 1.0, 2.5, 4.5, 7.0];
|
|
26
|
-
/** 满分基准、下限分与严重度分界:与 lib/rule/compilers/file-health.js 共用同一来源
|
|
27
|
-
* (lib/audit-defaults.js)——两处必须永远相等,各写一份会出现编译器与检查器
|
|
28
|
-
* 按不同基准判健康度的静默不一致(自审 duplicate-constant 报的真重复)。 */
|
|
29
|
-
import {
|
|
30
|
-
HEALTH_BASE_SCORE as BASE_SCORE, HEALTH_MIN_SCORE as MIN_SCORE,
|
|
31
|
-
HEALTH_WARN_SCORE as WARN_SCORE, HEALTH_BLOCK_SCORE as BLOCK_SCORE,
|
|
32
|
-
} from '../audit-defaults.js';
|
|
33
|
-
|
|
34
|
-
/** 文件健康度矩阵评分(1.0.13):行数×40% + 大小×35% + 行长×25% → 加权等级 → 插值扣分 → score/10。
|
|
35
|
-
* 豁免:generated./.min./locales/ 跳过;constants/ 行数等级强制 0;routes/ 行数阈值放宽(等级-1,最小0)。
|
|
36
|
-
* 输出:score<blockScore → blocker;score<warnScore → warning;否则 pass(不产出 finding)。
|
|
37
|
-
* 诊断 message 含三维等级/得分/verdict/优先处理维度,可直接展示。 */
|
|
38
|
-
export function checkFileHealth({ file, relPath, text, rules }) {
|
|
39
|
-
const findings = [];
|
|
40
|
-
const rule = rules?.[0];
|
|
41
|
-
if (!rule) return findings;
|
|
42
|
-
const x = rule; // ruleOut 把 extra 展开进 rule 顶层(无 rule.extra 属性)
|
|
43
|
-
const rel = relPath || file || '';
|
|
44
|
-
// 豁免判定:相对路径可能是 'constants/table.js' 或 '/constants/table.js',模式统一匹配两种形态
|
|
45
|
-
const relNorm = rel.startsWith('/') ? rel : '/' + rel;
|
|
46
|
-
// 跳过类(generated/min/locales)——先试精确(如 .generated. 可命中文件名),再试带斜杠(目录形态)
|
|
47
|
-
if ((x.exemptSkip || []).some((p) => p.includes('/') ? relNorm.includes(p) : rel.includes(p))) return findings;
|
|
48
|
-
// 行数豁免类(constants):行数等级强制 0
|
|
49
|
-
const constMode = (x.exemptConstants || []).some((p) => p.includes('/') ? relNorm.includes(p) : rel.includes(p));
|
|
50
|
-
// 行数阈值放宽类(routes):行数等级减 1
|
|
51
|
-
const routesMode = (x.exemptRoutes || []).some((p) => p.includes('/') ? relNorm.includes(p) : rel.includes(p));
|
|
52
|
-
|
|
53
|
-
const lines = text.split('\n');
|
|
54
|
-
// 三维(行数/大小/行长)一律按**代码部分**统计(修复):文档注释里内嵌的
|
|
55
|
-
// API 响应示例 JSON(pix-ezviewer response 类)把行数/大小/行长全撑成 L3-L4 假警报——
|
|
56
|
-
// 注释是设计说明不是代码规模。用 tokenizer 标记注释行/注释字节,三个维度都扣掉注释。
|
|
57
|
-
// ①整行注释(该行 token 全是 comment,含跨行块注释覆盖的行)不计行数与行长;
|
|
58
|
-
// ②注释字节(全部 comment token 的文本长度)从文件大小中扣除——代码大小才是规模。
|
|
59
|
-
const commentOnlyLines = new Set();
|
|
60
|
-
let maxLen = 0;
|
|
61
|
-
let codeBytes = 0;
|
|
62
|
-
(function computeLineMetrics() {
|
|
63
|
-
const tokens = tokenize(text);
|
|
64
|
-
let commentBytes = 0;
|
|
65
|
-
for (const t of tokens) {
|
|
66
|
-
if (t.type !== 'comment') continue;
|
|
67
|
-
commentBytes += Buffer.byteLength(String(t.value || ''), 'utf8');
|
|
68
|
-
const start = Number(t.line) || 1;
|
|
69
|
-
const newlines = String(t.value || '').match(/\n/g)?.length || 0;
|
|
70
|
-
for (let lineIdx = start; lineIdx <= start + newlines; lineIdx++) commentOnlyLines.add(lineIdx);
|
|
71
|
-
}
|
|
72
|
-
// 整行代码行长(该行非注释 token 存在 → 代码行;否则纯注释行不计)
|
|
73
|
-
for (let i = 0; i < lines.length; i++) {
|
|
74
|
-
const ln = lines[i];
|
|
75
|
-
if (commentOnlyLines.has(i + 1)) continue; // 整行注释(含示例数据串)不计行长
|
|
76
|
-
if (ln.length > maxLen) maxLen = ln.length;
|
|
77
|
-
}
|
|
78
|
-
codeBytes = Math.max(0, Buffer.byteLength(text, 'utf8') - commentBytes);
|
|
79
|
-
})();
|
|
80
|
-
// 代码行口径走**单一来源**(lib/ast/line-count.js:总行 − 纯注释行 − 空行)——
|
|
81
|
-
// 此前这里只减注释行、**漏减空行**,同一文件与本插件其它规则报出的行数对不上。
|
|
82
|
-
const codeLineCount = countCodeLines(text);
|
|
83
|
-
const codeLineCountSafe = Math.max(1, codeLineCount);
|
|
84
|
-
const sizeKb = codeBytes / 1024;
|
|
85
|
-
const maxCodeLen = maxLen;
|
|
86
|
-
|
|
87
|
-
const levelOf = (val, levels) => {
|
|
88
|
-
for (let i = 0; i < levels.length; i++) if (val <= levels[i]) return i;
|
|
89
|
-
return levels.length;
|
|
90
|
-
};
|
|
91
|
-
let linesLevel = levelOf(codeLineCountSafe, x.lineLevels || LINE_LEVELS);
|
|
92
|
-
const sizeLevel0 = levelOf(sizeKb, x.sizeLevels || SIZE_LEVELS);
|
|
93
|
-
const lenLevel = levelOf(maxCodeLen, x.lengthLevels || LENGTH_LEVELS);
|
|
94
|
-
if (constMode) linesLevel = 0; // constants:行数豁免
|
|
95
|
-
if (routesMode) linesLevel = Math.max(0, linesLevel - 1); // routes:行数阈值放宽
|
|
96
|
-
// 聚合型 vs 臃肿型(用户设计):先罚函数,再考虑文件——函数是执行单元、
|
|
97
|
-
// 文件是组织单元。最大函数 > 50 行(臃肿型:文件大是因为藏着大函数)→ 行数/大小
|
|
98
|
-
// 维度正常计分;全部函数 ≤ 50 行(聚合型:文件大只是函数多/数据多,或纯数据文件
|
|
99
|
-
// 无函数)→ 行数/大小维度豁免(不按文件规模扣分,文件边界只是导航成本)。
|
|
100
|
-
const maxFuncLen = maxFunctionLength(text);
|
|
101
|
-
const FUNC_COMPLIANT_LIMIT = 50; // 与 func-lines 默认 warn 阈值一致
|
|
102
|
-
// 聚合型判定加「有函数」前提:只有文件**含函数且函数都合规**才算聚合型
|
|
103
|
-
// (文件大=函数多/数据多);无函数的纯脚本/数据文件(顶层 1200 行语句)没有「函数
|
|
104
|
-
// 多」作证据,不豁免,仍按行数/大小计分(避免脚本型大文件漏报)。
|
|
105
|
-
const isAggregated = maxFuncLen > 0 && maxFuncLen <= FUNC_COMPLIANT_LIMIT;
|
|
106
|
-
let sizeLevel = sizeLevel0;
|
|
107
|
-
if (isAggregated) { linesLevel = 0; sizeLevel = 0; }
|
|
108
|
-
|
|
109
|
-
const weights = {
|
|
110
|
-
lines: x.weightLines !== undefined ? x.weightLines : WEIGHT_LINES,
|
|
111
|
-
size: x.weightSize !== undefined ? x.weightSize : WEIGHT_SIZE,
|
|
112
|
-
len: x.weightLineLength !== undefined ? x.weightLineLength : WEIGHT_LINE_LENGTH,
|
|
113
|
-
};
|
|
114
|
-
const weighted = linesLevel * weights.lines + sizeLevel * weights.size + lenLevel * weights.len;
|
|
115
|
-
// 扣分插值:penalties[i] 对应等级 i;小数等级按相邻线性插值
|
|
116
|
-
const penalties = x.penalties || LEVEL_PENALTIES;
|
|
117
|
-
const lo = Math.floor(weighted);
|
|
118
|
-
const hi = Math.min(lo + 1, penalties.length - 1);
|
|
119
|
-
const frac = weighted - lo;
|
|
120
|
-
const penalty = penalties[lo] + (penalties[hi] - penalties[lo]) * frac;
|
|
121
|
-
const base = x.baseScore !== undefined ? x.baseScore : BASE_SCORE;
|
|
122
|
-
const score = Math.max(MIN_SCORE, base - penalty);
|
|
123
|
-
|
|
124
|
-
const warnScore = x.warnScore !== undefined ? x.warnScore : WARN_SCORE;
|
|
125
|
-
const blockScore = x.blockScore !== undefined ? x.blockScore : BLOCK_SCORE;
|
|
126
|
-
if (score >= warnScore) return findings; // 健康:不产出
|
|
127
|
-
|
|
128
|
-
// 优先处理维度 = 等级最高的维度(并列取首个)
|
|
129
|
-
const dims = [['行数', linesLevel], ['大小', sizeLevel], ['行长', lenLevel]].sort((a, b) => b[1] - a[1]);
|
|
130
|
-
const top = dims[0];
|
|
131
|
-
const sev = score < blockScore ? 'blocker' : 'warning';
|
|
132
|
-
const label = ['优秀', '良好', '警戒', '危险', '严重'];
|
|
133
|
-
// 修 bug:总分标签曾硬编码「🟠 危险」,导致 8.4/10(刚低于 warn_score 9 的
|
|
134
|
-
// 提示阈值)也被写成「危险」,与规则文档「score≥9 健康,<9 提示拆分,<5 blocker」不符。
|
|
135
|
-
// 现按分数区间给档:<block_score 严重 / <(block+warn)/2 危险 / 其余 提示(轻度)。
|
|
136
|
-
const mid = (blockScore + warnScore) / 2;
|
|
137
|
-
const scoreTag = score < blockScore ? '🔴 严重' : (score < mid ? '🟠 危险' : '🟡 提示');
|
|
138
|
-
findings.push(makeFinding({
|
|
139
|
-
file, line: 1, rule: rule.id || 'maintainability/file-health', kind: 'file-health',
|
|
140
|
-
severity: sev,
|
|
141
|
-
message: `文件健康度 ${score.toFixed(1)}/10(${scoreTag}):`
|
|
142
|
-
+ `行数 ${codeLineCountSafe}→L${linesLevel}(${label[linesLevel]})、大小 ${sizeKb.toFixed(1)}KB→L${sizeLevel}(${label[sizeLevel]})、`
|
|
143
|
-
+ `最大代码行长 ${maxCodeLen}→L${lenLevel}(${label[lenLevel]})、最大函数 ${maxFuncLen} 行;优先处理:${top[0]}(L${top[1]})`
|
|
144
|
-
+ `${isAggregated ? '(聚合型:文件大但最大函数 ≤50 行——函数多而非函数长,按组织单元不按规模扣分)' : ''}`
|
|
145
|
-
+ `${lenLevel > 0 ? '(行长按非注释代码行计;注释/文档内嵌示例数据串不计)' : ''}`,
|
|
146
|
-
dimensions: rule.dimensions || ['可维护性', '可读性'],
|
|
147
|
-
exemptHint: 'dsh-skip-quality(文件头=整文件)', scoreImpact: sev === 'blocker' ? 2 : 1,
|
|
148
|
-
}));
|
|
149
|
-
return findings;
|
|
150
|
-
}
|
package/lib/checks/filter.js
DELETED
|
@@ -1,137 +0,0 @@
|
|
|
1
|
-
/**
|
|
2
|
-
* 检查层 · 规则作用域过滤
|
|
3
|
-
*
|
|
4
|
-
* 职责:按文件路径裁剪规则——exts(扩展名)、exclude_paths(目录前缀),
|
|
5
|
-
* 以及新增的 file_patterns(文件内容级初筛,三层审计 L1)。
|
|
6
|
-
* 这是「规则作用域」的唯一落地点:规则声明了作用域就必须在此生效,
|
|
7
|
-
* 否则出现「yml 写了 exts 却被静默忽略」的缺陷。
|
|
8
|
-
*
|
|
9
|
-
* 三层审计管线:
|
|
10
|
-
* L1 正则初筛(本文件 filterRulesByFileText)——每文件一次正则扫描,成本极低,
|
|
11
|
-
* 命中 file_patterns 的文件才进 L2;不命中直接剔除该规则(空数组短路)。
|
|
12
|
-
* L2 AST+YAML 精查(lib/ast/dataflow.js + lib/checks/dataflow.js)——只对 L1
|
|
13
|
-
* 候选文件做同函数内数据流判定(如「清空后访问」)。
|
|
14
|
-
* L3 运行时检测(scripts/audit-runtime-check.mjs)——兜住跨文件/闭包/异步盲区。
|
|
15
|
-
*/
|
|
16
|
-
|
|
17
|
-
/** 全量检查器调度:对单文件跑所有已编译规则 + 内置检查(质量类走 AST 级)。
|
|
18
|
-
* @param {object} opts { level: 'quick'|'standard'|'deep'(缺省 standard 全量) }
|
|
19
|
-
* quick 档跳过 AST/语义重检查(分析成本高):func-lines/complexity/depth/
|
|
20
|
-
* max-lines/repeated-string/semantic/credential-file/min-length,保留正则/
|
|
21
|
-
* 黑名单/路径/内置轻检——适合大仓冒烟、快速门禁。
|
|
22
|
-
* deep 档当前引擎与 standard 等效(全量),为未来追加深度检查预留。 */
|
|
23
|
-
/**
|
|
24
|
-
* 规则 exts 过滤:规则声明 `exts` 时仅对匹配扩展名的文件生效。
|
|
25
|
-
* 编译产物里 exts 存在 `rule.extra.exts`(regex 编译器已透传),此处消费——
|
|
26
|
-
* 修复 dsh/patch-insert-unique-id 等在 README/md 等非目标文件上的误报
|
|
27
|
-
* (例:patch 语义规则只查 yml/yaml,README 里的 insert 示例代码块不再命中)。
|
|
28
|
-
* @param {object} grouped 按 kind 分组的编译规则
|
|
29
|
-
* @param {string} relPath 相对路径(用于取扩展名)
|
|
30
|
-
* @returns {object} 过滤后的 grouped(规则数组已按 exts 裁剪)
|
|
31
|
-
*/
|
|
32
|
-
export function filterRulesByExt(grouped, relPath) {
|
|
33
|
-
// 取扩展名必须**先取 basename 再取最后一个点之后**:
|
|
34
|
-
// 旧实现 String(relPath).split('.').pop() 对含点目录的路径会取错 ✗ ——
|
|
35
|
-
// 例:docs/.vitepress/theme/index.ts ⇒ 得到 "vitepress/theme/index"(不是 ts)
|
|
36
|
-
// ⇒ 该文件所有声明了 exts 的规则**全部匹配不上**(审计实测:.vitepress 下文件只拿到通用规则)。
|
|
37
|
-
// 另:无扩展名文件(Makefile 等)返回空串,声明 exts 的规则自然不匹配(正确语义)。
|
|
38
|
-
const base = String(relPath || '').replace(/\\/g, '/').split('/').pop() || '';
|
|
39
|
-
const ext = base.includes('.') ? base.split('.').pop().toLowerCase() : '';
|
|
40
|
-
const out = {};
|
|
41
|
-
for (const [kind, rules] of Object.entries(grouped || {})) {
|
|
42
|
-
if (!Array.isArray(rules)) { out[kind] = rules; continue; }
|
|
43
|
-
out[kind] = rules.filter((r) => {
|
|
44
|
-
const exts = r?.exts ?? r?.extra?.exts; // ruleOut 展开后 exts 在顶层;兼容 extra 形式
|
|
45
|
-
if (!Array.isArray(exts) || exts.length === 0) return true; // 未声明 exts = 不限文件
|
|
46
|
-
return exts.includes(ext);
|
|
47
|
-
});
|
|
48
|
-
}
|
|
49
|
-
return out;
|
|
50
|
-
}
|
|
51
|
-
|
|
52
|
-
/**
|
|
53
|
-
* 路径级规则过滤——规则声明 exclude_paths 时,按 relPath 前缀排除。
|
|
54
|
-
* 与 filterRulesByExt 分离:exts 只按扩展名,区分不了「同是 .js 的 server/client」;
|
|
55
|
-
* 需要排除整目录(如 client-node-builtin-require 只查浏览器侧,排除 server/ test/ lib/)。
|
|
56
|
-
* 路径匹配:relPath 以任一 exclude 前缀开头即跳过该规则(前缀兼容有无尾斜杠)。
|
|
57
|
-
*/
|
|
58
|
-
/** 路径归一化(去反斜杠、去前导 ./、去尾斜杠)。原先在 filter 回调里逐规则重建闭包,既费分配也难读。 */
|
|
59
|
-
function normRelPath(s) {
|
|
60
|
-
return String(s || '').replace(/\\/g, '/').replace(/^\.?\//, '').replace(/\/+$/, '');
|
|
61
|
-
}
|
|
62
|
-
|
|
63
|
-
/**
|
|
64
|
-
* includePaths 命中判定:含 '/' 的模式按路径相等或后缀匹配;不含 '/' 的按**文件名**匹配(任意层级)。
|
|
65
|
-
* 抽成函数后 filter 回调里只剩一行 some(...),嵌套深度降一层。
|
|
66
|
-
*/
|
|
67
|
-
function includeHit(rel, inc) {
|
|
68
|
-
const pat = normRelPath(inc);
|
|
69
|
-
if (!pat) return false;
|
|
70
|
-
return pat.includes('/')
|
|
71
|
-
? (rel === pat || rel.startsWith(pat + '/') || rel.endsWith('/' + pat))
|
|
72
|
-
: rel.split('/').pop() === pat;
|
|
73
|
-
}
|
|
74
|
-
|
|
75
|
-
/** excludePaths 命中判定(前缀匹配,兼容有无尾斜杠两种写法)。 */
|
|
76
|
-
function excludeHit(rel, ex) {
|
|
77
|
-
const ee = normRelPath(ex);
|
|
78
|
-
return Boolean(ee) && (rel === ee || rel.startsWith(ee + '/') || rel.startsWith(ee));
|
|
79
|
-
}
|
|
80
|
-
|
|
81
|
-
export function filterRulesByPath(grouped, relPath) {
|
|
82
|
-
const p = String(relPath || '');
|
|
83
|
-
const out = {};
|
|
84
|
-
for (const [kind, rules] of Object.entries(grouped || {})) {
|
|
85
|
-
if (!Array.isArray(rules)) { out[kind] = rules; continue; }
|
|
86
|
-
out[kind] = rules.filter((r) => {
|
|
87
|
-
const rel = normRelPath(p);
|
|
88
|
-
// 白名单:声明 includePaths 时只有命中其一才跑该规则(与 excludePaths 互补)。
|
|
89
|
-
// 为什么需要:有些规则只该作用于特定文件(如「package.json 的 version 必须 SemVer」),
|
|
90
|
-
// 而 exts:["json"] 会把它套到所有 json 上——第三方仓库里成千上万个带 version 字段的 json
|
|
91
|
-
// 被逐条误报(实测 1435 条)。匹配语义见 includeHit。
|
|
92
|
-
const includes = r?.includePaths ?? r?.extra?.includePaths;
|
|
93
|
-
if (Array.isArray(includes) && includes.length && !includes.some((inc) => includeHit(rel, inc))) return false;
|
|
94
|
-
const excludes = r?.excludePaths ?? r?.extra?.excludePaths;
|
|
95
|
-
if (!Array.isArray(excludes) || excludes.length === 0) return true; // 未声明 = 不限路径
|
|
96
|
-
return !excludes.some((e) => excludeHit(rel, e));
|
|
97
|
-
});
|
|
98
|
-
}
|
|
99
|
-
return out;
|
|
100
|
-
}
|
|
101
|
-
|
|
102
|
-
/**
|
|
103
|
-
* 三层审计 L1——文件内容级初筛(file_patterns)。
|
|
104
|
-
*
|
|
105
|
-
* 动机:AST/数据流类规则(第二层)成本中,若对每个文件都跑 token 解析浪费。
|
|
106
|
-
* 声明 `file_patterns` 的规则先做一次**纯正则**快速筛选:文件文本命中任一
|
|
107
|
-
* pattern 才算「候选文件」,才进 L2 精查;未命中 → 该规则剔除(数组为空,
|
|
108
|
-
* 检查器空规则短路,顺带防 rule.severity 崩溃)。
|
|
109
|
-
*
|
|
110
|
-
* 与 exts/exclude_paths 的区别:那是**路径/扩展名**级裁剪,这是**文件内容**级
|
|
111
|
-
* 初筛——解决「正则模式本身命中面太宽、但 AST 精判成本高」的三层分层问题
|
|
112
|
-
* (如 clear-then-access 只在同函数含「清空动作」的文件上才值得做数据流分析)。
|
|
113
|
-
*
|
|
114
|
-
* @param {object} grouped 按 kind 分组的编译规则
|
|
115
|
-
* @param {string} text 文件全文(读一次,多规则共享)
|
|
116
|
-
* @returns {object} 过滤后的 grouped
|
|
117
|
-
*/
|
|
118
|
-
export function filterRulesByFileText(grouped, text) {
|
|
119
|
-
const out = {};
|
|
120
|
-
for (const [kind, rules] of Object.entries(grouped || {})) {
|
|
121
|
-
if (!Array.isArray(rules)) { out[kind] = rules; continue; }
|
|
122
|
-
out[kind] = rules.filter((r) => {
|
|
123
|
-
const fps = r?.filePatterns ?? r?.extra?.filePatterns; // ruleOut 展开后顶层;兼容 extra
|
|
124
|
-
if (!Array.isArray(fps) || fps.length === 0) return true; // 未声明 = 全文都算候选(不做 L1 初筛)
|
|
125
|
-
// 纯文本单次扫描:任一 pattern 命中即候选(L1 只追求召回,不追求精度)
|
|
126
|
-
const body = String(text || '');
|
|
127
|
-
return fps.some((p) => {
|
|
128
|
-
if (p instanceof RegExp) { p.lastIndex = 0; return p.test(body); }
|
|
129
|
-
// 字符串形态 = 正则源串(compiler 从 yml file_patterns 透传,含 \b 等转义),
|
|
130
|
-
// 按 RegExp 编译执行;编译失败(非法正则)退回子串包含(保守:宁放行不漏筛)。
|
|
131
|
-
try { return new RegExp(String(p)).test(body); }
|
|
132
|
-
catch { return body.includes(String(p)); }
|
|
133
|
-
});
|
|
134
|
-
});
|
|
135
|
-
}
|
|
136
|
-
return out;
|
|
137
|
-
}
|
package/lib/checks/folder.js
DELETED
|
@@ -1,173 +0,0 @@
|
|
|
1
|
-
/**
|
|
2
|
-
* 检查层 · 目录级检查
|
|
3
|
-
*
|
|
4
|
-
* 职责:文件夹数量/单目录文件数/解包特征/.gitignore 覆盖等目录级审计。
|
|
5
|
-
* 与文件级检查不同,本模块以目录为输入单位。
|
|
6
|
-
*/
|
|
7
|
-
|
|
8
|
-
import { existsSync, readFileSync, readdirSync, statSync } from 'node:fs';
|
|
9
|
-
import { execFileSync } from 'node:child_process';
|
|
10
|
-
import { join, relative } from 'node:path';
|
|
11
|
-
import { globToRegex } from '../audit/glob.js';
|
|
12
|
-
import { parseGitignore, isIgnoredByRules } from '../audit/gitignore-match.js';
|
|
13
|
-
import { makeFinding } from '../audit/index.js';
|
|
14
|
-
import { capSeverity, HINT_SIZE } from './common.js';
|
|
15
|
-
|
|
16
|
-
/**
|
|
17
|
-
* 目录级审计(folder 槽位,1.0.3)。
|
|
18
|
-
* 统计:源码目录总数(排除 excludeDirs)/ 单目录文件数 / 解包特征目录 / .gitignore 覆盖。
|
|
19
|
-
* 由 auditFull 在文件行级检查之外追加调用(目录层规则无法按文件行跑)。
|
|
20
|
-
* @param {object} opts { root, rules, gitignoreText, excludeDirsBase }
|
|
21
|
-
* @returns {Array} findings(每条 file=根目录,line=1)
|
|
22
|
-
*/
|
|
23
|
-
export function checkFolderRules({ root, rules, gitignoreText = '' }) {
|
|
24
|
-
const findings = [];
|
|
25
|
-
if (!root) return findings;
|
|
26
|
-
|
|
27
|
-
const isDir = (p) => { try { return statSync(p).isDirectory(); } catch { return false; } };
|
|
28
|
-
|
|
29
|
-
// 排除目录(各规则 exclude_dirs 的并集)——**在遍历时剪枝**,不是收集后按名过滤。
|
|
30
|
-
//
|
|
31
|
-
// 修复的缺陷:原实现先递归收集全部目录、再 `dirs.filter(d => !exclude.has(d.name))`
|
|
32
|
-
// 只按目录**名**过滤,不阻止递归 → 被排除目录的**子目录全被计入源码目录**
|
|
33
|
-
// (实测:树内只有 src、lib 两个真源码目录,却因 node_modules/pkg-a、
|
|
34
|
-
// node_modules/pkg-b/node_modules/nested、dist/sub、coverage/x 等子目录被计入而报 9 个)。
|
|
35
|
-
// 后果:任何带 node_modules 的仓库恒定超阈值 30 → folder/total-count 恒误报。
|
|
36
|
-
// 同理 folder/file-count-per-dir 会把 node_modules 里的大目录当"单目录文件过多"报出。
|
|
37
|
-
const excludedNames = new Set(['.git']);
|
|
38
|
-
for (const r of rules || []) for (const n of r.excludeDirs || []) excludedNames.add(n);
|
|
39
|
-
|
|
40
|
-
// .gitignore 感知(上游克隆/忽略目录误报修复):folder 独立遍历此前
|
|
41
|
-
// 不看 .gitignore——被忽略的目录(如 docs/.probe-ktoolbox 上游 KToolBox 完整克隆,
|
|
42
|
-
// .gitignore 的 docs/* 忽略)被计入 total-count 造成误报(Pawchive 76 目录大头即此)。
|
|
43
|
-
// 现用同 collector 的纯 JS 匹配器(lib/audit/gitignore-match.js):被 gitignore 忽略
|
|
44
|
-
// 的目录不入 dirs、不递归(gitignoreText 缺省空 → 规则不变)。
|
|
45
|
-
const gitignoreRules = gitignoreText ? parseGitignore(gitignoreText) : null;
|
|
46
|
-
|
|
47
|
-
// ── 全量目录/文件统计(一次遍历供多规则复用)──
|
|
48
|
-
const dirCounts = { total: 0, byDir: new Map() }; // 单目录文件数
|
|
49
|
-
const signatureHits = [];
|
|
50
|
-
const dirs = [];
|
|
51
|
-
(function walk(dir) {
|
|
52
|
-
let entries;
|
|
53
|
-
try { entries = readdirSync(dir, { withFileTypes: true }); } catch { return; }
|
|
54
|
-
let fileCount = 0;
|
|
55
|
-
for (const e of entries) {
|
|
56
|
-
// 排除目录:不入 dirs、不递归、文件也不计数(其内容不属于项目源码)
|
|
57
|
-
if (excludedNames.has(e.name)) continue;
|
|
58
|
-
const full = `${dir}/${e.name}`;
|
|
59
|
-
if (e.isDirectory()) {
|
|
60
|
-
// gitignore 忽略目录:跳过(上游克隆/依赖/内部数据目录不是项目源码)
|
|
61
|
-
if (gitignoreRules) {
|
|
62
|
-
const relDir = relative(root, full).replace(/\\/g, '/');
|
|
63
|
-
if (isIgnoredByRules(gitignoreRules, relDir, true)) continue;
|
|
64
|
-
}
|
|
65
|
-
dirs.push({ dir, name: e.name, full }); walk(full);
|
|
66
|
-
} else fileCount++;
|
|
67
|
-
}
|
|
68
|
-
if (fileCount > 0 || dir !== root) dirCounts.byDir.set(dir, fileCount);
|
|
69
|
-
dirCounts.total++;
|
|
70
|
-
})(root);
|
|
71
|
-
|
|
72
|
-
for (const rule of rules || []) {
|
|
73
|
-
const exclude = new Set(rule.excludeDirs || []);
|
|
74
|
-
// 规则 1:源码目录总数 ≤ threshold(排除 excludeDirs)
|
|
75
|
-
if (rule.threshold !== undefined && dirs.length) {
|
|
76
|
-
const sig = rule.id.includes('total-count');
|
|
77
|
-
if (sig) {
|
|
78
|
-
// 包结构目录排除(Java/Kotlin 误报修复):Maven/Gradle 标准源码根
|
|
79
|
-
// src/main/java、src/main/kotlin、src/test/java 之下的目录是**包层级**(每个
|
|
80
|
-
// 包一个目录,如 com/perol/asdpl/pixivez/**),不是「项目源码目录」——Android/
|
|
81
|
-
// Maven 项目天然几十个包目录,total-count 阈值 30 恒误报(防解包产物混入的
|
|
82
|
-
// 本意对包结构不适用)。判定按路径子串:含 /src/(main|test|androidTest)/(java|kotlin)/
|
|
83
|
-
// 前缀的目录全部不计(含其子目录——包目录逐层都在其下)。
|
|
84
|
-
// 工程/资源目录按**名称**过滤:gradle/docs/res/assets 等是构建系统/
|
|
85
|
-
// 文档/资源/CI 配置目录,不是「项目源码目录」。⚠️ 不能进 exclude_dirs——那会被
|
|
86
|
-
// collector 的 SKIP_DIRS 消费导致全局跳过审计(实测回归);只在计数时过滤。
|
|
87
|
-
// res/ 资源**子目录**(res/layout、res/values-xx、res/drawable-xxhdpi 等)按路径
|
|
88
|
-
// 过滤(/res/ 之下是资源不是源码——Android 多语言 values 目录天然十几个)。
|
|
89
|
-
const isPackageDir = (d) => /\/src\/(?:main|test|androidTest)\/(?:java|kotlin)\//.test(d.full)
|
|
90
|
-
|| /\/res\//.test(d.full);
|
|
91
|
-
const ENGINE_DIRS = new Set(['gradle', 'wrapper', '.github', 'docs', 'res', 'assets', 'resources', 'help', 'preview', 'donation', '.idea', '.gradle']);
|
|
92
|
-
const srcDirs = dirs.filter((d) => !exclude.has(d.name) && !isPackageDir(d) && !ENGINE_DIRS.has(d.name)).length;
|
|
93
|
-
if (srcDirs > rule.threshold) {
|
|
94
|
-
findings.push(makeFinding({
|
|
95
|
-
file: root, line: 1, rule: rule.id, kind: 'folder',
|
|
96
|
-
severity: capSeverity(rule.severity || 'warning', 'warning'),
|
|
97
|
-
message: (rule.message || rule.name || '').replace('{count}', srcDirs).replace('{threshold}', rule.threshold),
|
|
98
|
-
dimensions: rule.dimensions || ['可维护性'],
|
|
99
|
-
exemptHint: HINT_SIZE,
|
|
100
|
-
scoreImpact: 1,
|
|
101
|
-
}));
|
|
102
|
-
}
|
|
103
|
-
}
|
|
104
|
-
}
|
|
105
|
-
// 规则 2:单目录文件数 > threshold
|
|
106
|
-
if (rule.id.includes('file-count-per-dir') && rule.threshold !== undefined) {
|
|
107
|
-
for (const [dir, cnt] of dirCounts.byDir) {
|
|
108
|
-
if (cnt <= rule.threshold) continue;
|
|
109
|
-
findings.push(makeFinding({
|
|
110
|
-
file: `${root}/…`, line: 1, rule: rule.id, kind: 'folder',
|
|
111
|
-
severity: capSeverity(rule.severity || 'warning', 'warning'),
|
|
112
|
-
message: (rule.message || '').replace('{path}', dir.replace(root, '')).replace('{count}', cnt).replace('{threshold}', rule.threshold),
|
|
113
|
-
dimensions: rule.dimensions || ['可维护性'],
|
|
114
|
-
exemptHint: HINT_SIZE,
|
|
115
|
-
scoreImpact: 1,
|
|
116
|
-
}));
|
|
117
|
-
}
|
|
118
|
-
}
|
|
119
|
-
// 规则 3:解包特征目录
|
|
120
|
-
if (Array.isArray(rule.signatures)) {
|
|
121
|
-
for (const sig of rule.signatures) {
|
|
122
|
-
const pat = String(sig.pattern || '').replace(/\*\*/g, '**').replace(/^\*\*\/?/, '');
|
|
123
|
-
const hit = dirs.find((d) => {
|
|
124
|
-
const rel = d.full.replace(root, '').replace(/^\//, '');
|
|
125
|
-
if (pat.includes('package/package.json')) return rel === 'package' && existsSync(`${d.full}/package.json`); // dsh-skip-residue(签名规则条数极少且 find 短路)
|
|
126
|
-
const base = pat.split('/')[0];
|
|
127
|
-
return rel === base || rel.startsWith(`${base}/`);
|
|
128
|
-
});
|
|
129
|
-
if (hit) {
|
|
130
|
-
findings.push(makeFinding({
|
|
131
|
-
file: `${root}/…`, line: 1, rule: rule.id, kind: 'folder',
|
|
132
|
-
severity: capSeverity(rule.severity || 'warning', 'warning'),
|
|
133
|
-
message: sig.message || `${rule.name}: ${sig.pattern}`,
|
|
134
|
-
dimensions: rule.dimensions || ['可维护性'],
|
|
135
|
-
exemptHint: HINT_SIZE,
|
|
136
|
-
scoreImpact: 1,
|
|
137
|
-
}));
|
|
138
|
-
}
|
|
139
|
-
}
|
|
140
|
-
}
|
|
141
|
-
// 规则 4:.gitignore 覆盖检查
|
|
142
|
-
// 修误报:旧实现要求 .gitignore **含全部** required_patterns
|
|
143
|
-
// (node_modules/dist/build/coverage/*.log/.env)——纯 JS 零依赖插件没有构建
|
|
144
|
-
// 产物,dist/build/coverage/.env 根本不存在,报「缺少忽略项」属误报。
|
|
145
|
-
// 正确语义:只报「仓库里**实际存在**的产物目录/文件」还没被忽略的项。
|
|
146
|
-
if (Array.isArray(rule.requiredPatterns) && gitignoreText) {
|
|
147
|
-
const lines = gitignoreText.split('\n').map((l) => l.trim());
|
|
148
|
-
const isIgnored = (p) => lines.some((l) => l === p || l === p.replace(/^\*/, '') || l.replace(/\/$/, '') === p);
|
|
149
|
-
const missing = rule.requiredPatterns.filter((p) => {
|
|
150
|
-
if (isIgnored(p)) return false; // 已忽略 → 无需报
|
|
151
|
-
// 未忽略:仅当该项在仓库里实际存在时才报(不存在则本项目本就不需要忽略它)
|
|
152
|
-
const probe = p.replace(/^\*/, '').replace(/^\//, '');
|
|
153
|
-
const hitDir = dirs.some((d) => {
|
|
154
|
-
const rel = d.full.replace(root, '').replace(/^\//, '');
|
|
155
|
-
return rel === probe || rel.startsWith(`${probe}/`);
|
|
156
|
-
});
|
|
157
|
-
const hitFile = existsSync(join(root, probe)); // dsh-skip-residue(required_patterns 条数少,每 pattern 一次判定)
|
|
158
|
-
return hitDir || hitFile;
|
|
159
|
-
});
|
|
160
|
-
if (missing.length) {
|
|
161
|
-
findings.push(makeFinding({
|
|
162
|
-
file: `${root}/.gitignore`, line: 1, rule: rule.id, kind: 'folder',
|
|
163
|
-
severity: capSeverity(rule.severity || 'warning', 'warning'),
|
|
164
|
-
message: (rule.message || rule.name || '').replace('{missing}', missing.join(', ')),
|
|
165
|
-
dimensions: rule.dimensions || ['可维护性'],
|
|
166
|
-
exemptHint: HINT_SIZE,
|
|
167
|
-
scoreImpact: 1,
|
|
168
|
-
}));
|
|
169
|
-
}
|
|
170
|
-
}
|
|
171
|
-
}
|
|
172
|
-
return findings;
|
|
173
|
-
}
|
package/lib/checks/fs-scanner.js
DELETED
|
@@ -1,57 +0,0 @@
|
|
|
1
|
-
/**
|
|
2
|
-
* 仓库级检查:逐文件扫描实现过多(`kind: multi-file-scanners`)
|
|
3
|
-
*
|
|
4
|
-
* 【判据】同一函数体内**同时**出现 `readdir` 与 `stat` ⇒ 一个「逐文件取属性」的扫描实现。
|
|
5
|
-
* 全仓这类实现数 > `min_occurrences`(规则字段,默认 3)⇒ 报 **一条** notice:
|
|
6
|
-
* 「同一职责是否被各处独立实现了一遍」交人工判断(不是「你有 N 个 bug」✗)。
|
|
7
|
-
*
|
|
8
|
-
* 【为什么只报一处】这是**整个仓库**的属性(与 duplicate-code 同性质)。逐文件报会在 N 个
|
|
9
|
-
* 文件上产出 N 条同义提示,把真实问题淹掉 ⇒ 只在代表文件上报一次,其余实现列进 message ✓
|
|
10
|
-
* (代表文件 = 路径字典序最小者,由 orchestrate 的 pickRepoLevelAnchorFile 决定语义)。
|
|
11
|
-
*
|
|
12
|
-
* 【口径单一】判据实现只在 `lib/ast/fs-hotspots.js` 一份 —— 独立 CLI
|
|
13
|
-
* `scripts/scan-fs-hotspots.mjs` 用的是同一个函数,不存在「规则一套、脚本一套」的漂移 ✗
|
|
14
|
-
*
|
|
15
|
-
* @param {Array<{path:string,text:string}>} fileTexts 全仓代码文件文本
|
|
16
|
-
* @param {Array} rules 该 kind 下的规则(取第一条;阈值读 min_occurrences / extra.minOccurrences)
|
|
17
|
-
* @returns {Array} findings
|
|
18
|
-
*/
|
|
19
|
-
import { scanFsHotspots } from '../ast/fs-hotspots.js';
|
|
20
|
-
|
|
21
|
-
export function checkMultiFileScanners(fileTexts = [], rules = null) {
|
|
22
|
-
const findings = [];
|
|
23
|
-
if (!Array.isArray(fileTexts) || fileTexts.length < 2) return findings; // 单文件谈不上「多实现」
|
|
24
|
-
const rule = Array.isArray(rules) && rules.length ? rules[0] : null;
|
|
25
|
-
if (!rule) return findings;
|
|
26
|
-
|
|
27
|
-
const threshold = Number(rule?.extra?.minOccurrences ?? rule?.minOccurrences ?? 3) || 3;
|
|
28
|
-
const severity = 'notice'; // 恒 notice:这是启发式提示(不同用途的扫描器是合法的)⇒ 只提示不警告
|
|
29
|
-
|
|
30
|
-
let agg;
|
|
31
|
-
try {
|
|
32
|
-
agg = scanFsHotspots(fileTexts);
|
|
33
|
-
} catch {
|
|
34
|
-
return findings; // 单个异常不影响整轮审计
|
|
35
|
-
}
|
|
36
|
-
const scanners = agg.scanners || [];
|
|
37
|
-
if (scanners.length <= threshold) return findings;
|
|
38
|
-
|
|
39
|
-
// 代表文件 = 路径字典序最小者(与 repo-level 语义一致,保证同一仓库每次报同一处)
|
|
40
|
-
const sorted = [...scanners].sort((a, b) => String(a.file).localeCompare(String(b.file)));
|
|
41
|
-
const first = sorted[0];
|
|
42
|
-
const rest = sorted.slice(1);
|
|
43
|
-
const list = rest.map((s) => `${s.file}::${s.fn}()`).join('、');
|
|
44
|
-
findings.push({
|
|
45
|
-
rule: rule.id,
|
|
46
|
-
severity,
|
|
47
|
-
file: first.file,
|
|
48
|
-
line: first.line,
|
|
49
|
-
message: `仓库内「同一函数里既 readdir 又 stat」的逐文件扫描实现共 ${scanners.length} 个`
|
|
50
|
-
+ `(阈值 ${threshold}):${first.file}::${first.fn}()`
|
|
51
|
-
+ (list ? ` 等 —— 其余:${list}` : '')
|
|
52
|
-
+ '。提示:同一职责被各处独立实现时,合并成一层遍历/索引可省掉重复 readdir 与重复 stat'
|
|
53
|
-
+ '(不同用途的扫描器属正常,需人工判断)',
|
|
54
|
-
dimensions: rule?.dimensions,
|
|
55
|
-
});
|
|
56
|
-
return findings;
|
|
57
|
-
}
|
|
@@ -1,98 +0,0 @@
|
|
|
1
|
-
// dsh-skip-i18n: 本文件为规则实现,输出文案硬编码中文为产品行为(无 i18n 需求)
|
|
2
|
-
/**
|
|
3
|
-
* 检查层 · 硬编码绝对路径精筛(paths 槽位 / astConfirmKind: "hardcoded-abs-path")
|
|
4
|
-
*
|
|
5
|
-
* 为什么需要精筛(正则做不到的两件事):
|
|
6
|
-
* ① **可访问性判断**——同一个绝对路径在本机可能存在(说明是"这台机器的真实路径",换机必失效)
|
|
7
|
-
* 也可能不存在(别人机器/历史遗留写法),两者的处置不同,只有 fs 才能判断;
|
|
8
|
-
* ② **相对路径换算**——把绝对路径换算成相对该文件所在目录的路径(修复时的替换基准),
|
|
9
|
-
* 供修复时直接替换;这需要文件路径上下文。
|
|
10
|
-
*
|
|
11
|
-
* 与正则的分工(引擎既有形态):规则的 patterns 只做**初筛**(跨系统字面量形状),
|
|
12
|
-
* 本模块给出**允许报出的行号集合**(allow 语义)+ 行号→详情(供 finding message 用)。
|
|
13
|
-
*
|
|
14
|
-
* 唯一实现声明:本模块是"硬编码绝对路径"检测的**唯一实现**——脚本(scripts/scan-hardcoded-paths.mjs)
|
|
15
|
-
* 必须调用它,不得自带第二份 PATTERNS(同一事实两份实现 ⇒ 必然分叉)。
|
|
16
|
-
*/
|
|
17
|
-
import { existsSync } from 'node:fs';
|
|
18
|
-
import { relative, dirname, isAbsolute, join } from 'node:path';
|
|
19
|
-
|
|
20
|
-
/** 跨系统硬编码绝对路径的识别(只取"像绝对路径"的形状,避免把相对路径/URL 误判)。 */
|
|
21
|
-
const PATH_RES = [
|
|
22
|
-
// Linux / macOS:/volume1 /home /Users /var /usr /opt /tmp /mnt /srv /etc /root /Applications /Volumes
|
|
23
|
-
/(?:^|["'`(\s=:])(\/(?:volume\d+|home|Users|var|usr|opt|tmp|mnt|srv|etc|root|Applications|Volumes)\/[^\s"'`)\]},;]*)/g,
|
|
24
|
-
// Windows:C:\ 或 C:/
|
|
25
|
-
/(?:^|["'`(\s=:])([A-Za-z]:[\\/][^\s"'`)\]},;]*)/g,
|
|
26
|
-
// UNC:\\server\share\...
|
|
27
|
-
/(?:^|["'`(\s=:])(\\\\[A-Za-z0-9_.-]+\\[^\s"'`)\]},;]*)/g,
|
|
28
|
-
// file:///...
|
|
29
|
-
/(file:\/\/\/[^\s"'`)\]},;]*)/g,
|
|
30
|
-
];
|
|
31
|
-
|
|
32
|
-
/**
|
|
33
|
-
* 命中里混入中文 / 全角标点 ⇒ 这是**散文**(说明文字、规则描述、注释里的举例),不是路径。
|
|
34
|
-
* 实测两类误报都由它兜掉:
|
|
35
|
-
* ① 规则自己的描述文本被夹具引用:`/volume1/...,换机即失效)`(后面跟了中文逗号与全角括号);
|
|
36
|
-
* ② 截图输出路径后紧跟全角括号:`/tmp/iwara-login-shots)`。
|
|
37
|
-
* 真实路径不会含 CJK 或全角标点 ⇒ 直接丢弃该命中(不是"只警告",是根本不报)。
|
|
38
|
-
*/
|
|
39
|
-
const PROSE_MIXED = /[\u3000-\u303f\u4e00-\u9fff\uff00-\uffef]/;
|
|
40
|
-
|
|
41
|
-
/**
|
|
42
|
-
* 扫描文本里的硬编码绝对路径(引擎侧唯一实现)。
|
|
43
|
-
* @param {string} text 文件全文
|
|
44
|
-
* @param {string} filePath 文件路径(用于换算相对路径;缺省时不做换算)
|
|
45
|
-
* @returns {Array<{line:number, raw:string, accessible:boolean, relative:string, os:string}>}
|
|
46
|
-
*/
|
|
47
|
-
export function hardcodedAbsPathHits(text, filePath = '', baseDir = '') {
|
|
48
|
-
const hits = [];
|
|
49
|
-
// 审计传进来的 file 可能是**相对审计根**的路径(如 src/sample.mjs)⇒ 必须用 repoPath 补成绝对路径,
|
|
50
|
-
// 否则 relative('src', '/tmp/x') 会算出一串 ../../(实测踩到:message 里出现 ../../../../../../tmp/…)。
|
|
51
|
-
const absFile = filePath ? (isAbsolute(filePath) ? filePath : join(String(baseDir || ''), filePath)) : '';
|
|
52
|
-
const lines = String(text || '').split('\n');
|
|
53
|
-
lines.forEach((line, i) => {
|
|
54
|
-
PATH_RES.forEach((re, idx) => {
|
|
55
|
-
re.lastIndex = 0;
|
|
56
|
-
let m;
|
|
57
|
-
while ((m = re.exec(line))) {
|
|
58
|
-
const raw = m[1];
|
|
59
|
-
if (!raw) continue;
|
|
60
|
-
if (PROSE_MIXED.test(raw)) continue; // 混入中文/全角标点 = 散文举例,不是路径(见 PROSE_MIXED 注释)
|
|
61
|
-
const accessible = existsSync(raw);
|
|
62
|
-
hits.push({
|
|
63
|
-
line: i + 1,
|
|
64
|
-
raw,
|
|
65
|
-
accessible,
|
|
66
|
-
// 可访问才换算:不可访问的路径(别人机器/已删除)无从确定基准,只提示不改写
|
|
67
|
-
relative: accessible && absFile ? relative(dirname(absFile), raw).split('\\').join('/') : '',
|
|
68
|
-
os: ['posix', 'windows', 'unc', 'file-url'][idx] || 'unknown',
|
|
69
|
-
});
|
|
70
|
-
}
|
|
71
|
-
});
|
|
72
|
-
});
|
|
73
|
-
return hits;
|
|
74
|
-
}
|
|
75
|
-
|
|
76
|
-
/** astConfirmKind: "hardcoded-abs-path" 的集合取值——报出的行号(allow 语义:全部命中都报)。 */
|
|
77
|
-
export function hardcodedAbsPathLines(text, filePath = '', baseDir = '') {
|
|
78
|
-
return new Set(hardcodedAbsPathHits(text, filePath, baseDir).map((h) => h.line));
|
|
79
|
-
}
|
|
80
|
-
|
|
81
|
-
/** astConfirmKind: "hardcoded-abs-path" 的行号→详情(供 finding message:可访问者附相对路径)。 */
|
|
82
|
-
export function hardcodedAbsPathDetail(text, filePath = '', baseDir = '') {
|
|
83
|
-
const out = new Map();
|
|
84
|
-
for (const h of hardcodedAbsPathHits(text, filePath, baseDir)) {
|
|
85
|
-
const prev = out.get(h.line);
|
|
86
|
-
// 同一行多个路径时优先展示"可访问"的那条(可操作:能直接换成相对路径)
|
|
87
|
-
if (!prev || (!prev.accessible && h.accessible)) out.set(h.line, h);
|
|
88
|
-
}
|
|
89
|
-
return out;
|
|
90
|
-
}
|
|
91
|
-
|
|
92
|
-
/** 人类可读的一句话说明(可访问 ⇒ 附相对路径;不可访问 ⇒ 仅警告)。 */
|
|
93
|
-
export function describeHardcodedAbsPath(hit) {
|
|
94
|
-
if (!hit) return '硬编码绝对路径';
|
|
95
|
-
return hit.accessible && hit.relative
|
|
96
|
-
? `硬编码绝对路径 ${hit.raw}(本机可访问)⇒ 建议改为相对路径 ${hit.relative}`
|
|
97
|
-
: `硬编码绝对路径 ${hit.raw}(本机不可访问)⇒ 仅警告,无法换算相对路径`;
|
|
98
|
-
}
|