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
package/lib/ast/io-risk-loop.js
DELETED
|
@@ -1,308 +0,0 @@
|
|
|
1
|
-
/**
|
|
2
|
-
* dsh-git-push — I/O 风险分级:循环判定
|
|
3
|
-
*
|
|
4
|
-
* 职责:判断某处 I/O 是否在循环体内、以及该循环执行次数能否静态确定为很少
|
|
5
|
-
* (小字面量数组 → 判定降级)。这是分级里最易出错的一块:
|
|
6
|
-
* 「迭代器表达式内的 I/O 只执行一次」与「数组方法回调体内的 I/O 逐元素执行」
|
|
7
|
-
* 形态相邻而语义相反,判据与反例都记在各函数注释里。
|
|
8
|
-
*
|
|
9
|
-
* 分层:lib/ast/io-risk.js 的从属模块(由该文件再导出),调用方不应直接 import。
|
|
10
|
-
*/
|
|
11
|
-
import { LOOP_KEYWORDS, LOOP_METHODS, isLoopHead } from './io-risk-const.js';
|
|
12
|
-
import {
|
|
13
|
-
SMALL_FIXED_LOOP_MAX, LOOP_BODY_SEARCH_TOKENS, findLoopRange, matchParen, endLineOf, seekLoopBody,
|
|
14
|
-
isSingleStmtLoopClose, classifyFnHead, classifySyncFnHead,
|
|
15
|
-
} from './io-risk-fn.js';
|
|
16
|
-
import { matchPairBack } from './brace.js';
|
|
17
|
-
|
|
18
|
-
/**
|
|
19
|
-
* 数组方法调用(`.some(` / `.map(` 等)的接收者是否为固定小字面量数组。
|
|
20
|
-
* 形如 `['a','b','c'].some(...)`:向前找到匹配的 `]`,其配对 `[` 即数组字面量,
|
|
21
|
-
* 元素数 ≤ SMALL_FIXED_LOOP_MAX 才算固定小集合。
|
|
22
|
-
* @returns {boolean}
|
|
23
|
-
*/
|
|
24
|
-
export function isSmallLiteralCallee(tokens, methodIdx) {
|
|
25
|
-
// methodIdx 指向方法名,其前应为 `.`,再前应为 `]`
|
|
26
|
-
if (tokens[methodIdx - 1]?.value !== '.') return false;
|
|
27
|
-
const close = methodIdx - 2;
|
|
28
|
-
if (tokens[close]?.value !== ']') return false;
|
|
29
|
-
const open = matchPairBack(tokens, close, '[', ']');
|
|
30
|
-
return open !== -1 && countLiteralElements(tokens, open) <= SMALL_FIXED_LOOP_MAX;
|
|
31
|
-
}
|
|
32
|
-
|
|
33
|
-
/**:字面量数组起手 + 仅少量 push 追加。
|
|
34
|
-
*
|
|
35
|
-
* 识别 `const candidates = ['a']; if (x) candidates.push(b); for (const f of candidates)`
|
|
36
|
-
* 这类**候选路径探测**——它规模确定(≤ SMALL_FIXED_LOOP_MAX)、执行次数极少,
|
|
37
|
-
* 与字面量数组同义;不识别会导致「循环内 I/O 高风险」误报(实测 config.js /
|
|
38
|
-
* 文档生成脚本 / credentials.js 共 8 处候选探测全被误判)。
|
|
39
|
-
*
|
|
40
|
-
* 保守策略:只要出现字面量数组**之外**的写入方式(concat/splice/展开赋值等),
|
|
41
|
-
* 即视为规模未知,返回 false(宁可保守报高风险,也不放过真问题)。
|
|
42
|
-
* @param {Array} tokens tokenize 结果
|
|
43
|
-
* @param {string} name 数组变量名
|
|
44
|
-
* @param {number} ofIdx 循环头 'of'/'in' 的位置(只回看它之前)
|
|
45
|
-
* @returns {boolean}
|
|
46
|
-
*/
|
|
47
|
-
export function isSmallPushedArray(tokens, name, ofIdx) {
|
|
48
|
-
// ① 声明处必须是字面量数组:`const name = [` / `let name = [`
|
|
49
|
-
const declIdx = findLiteralArrayDecl(tokens, name, ofIdx);
|
|
50
|
-
if (declIdx === -1) return false;
|
|
51
|
-
|
|
52
|
-
// ② 统计其后的 push 次数;出现任何让规模不可静态确定的写法即判未知
|
|
53
|
-
const pushes = countUnsafePushes(tokens, name, ofIdx);
|
|
54
|
-
if (pushes === null) return false;
|
|
55
|
-
|
|
56
|
-
// ③ 字面量区元素数 + push 次数 ≤ 上限
|
|
57
|
-
return countLiteralElements(tokens, declIdx) + pushes <= SMALL_FIXED_LOOP_MAX;
|
|
58
|
-
}
|
|
59
|
-
|
|
60
|
-
/**
|
|
61
|
-
* 回看 `name = [` 形态的声明位置;未以字面量数组起手返回 -1。
|
|
62
|
-
* 取**最后一次**匹配(同名变量被重新赋值时,以离循环最近的一次为准)。
|
|
63
|
-
* @returns {number} 数组 `[` 的 token 下标;未找到为 -1
|
|
64
|
-
*/
|
|
65
|
-
export function findLiteralArrayDecl(tokens, name, ofIdx) {
|
|
66
|
-
let declIdx = -1;
|
|
67
|
-
for (let j = 0; j < ofIdx; j++) {
|
|
68
|
-
if (tokens[j].type !== 'ident' || tokens[j].value !== name) continue;
|
|
69
|
-
const eq = tokens[j + 1];
|
|
70
|
-
const open = tokens[j + 2];
|
|
71
|
-
if (eq?.value === '=' && open?.value === '[') declIdx = j + 2;
|
|
72
|
-
}
|
|
73
|
-
return declIdx;
|
|
74
|
-
}
|
|
75
|
-
|
|
76
|
-
/** 让数组规模不可静态确定的写法(合并/改写类方法)。 */
|
|
77
|
-
export const UNSAFE_ARRAY_METHODS = new Set([
|
|
78
|
-
'concat', 'splice', 'unshift', 'apply', 'flat', 'flatMap',
|
|
79
|
-
]);
|
|
80
|
-
|
|
81
|
-
/**
|
|
82
|
-
* 数「安全 push」的次数。
|
|
83
|
-
* 出现以下任一即返回 null(表示规模未知,调用方按高风险处理):
|
|
84
|
-
* · 展开运算符 `...`
|
|
85
|
-
* · 合并/改写类方法(UNSAFE_ARRAY_METHODS)
|
|
86
|
-
* · push 发生在循环体内(元素来自被遍历集合,次数不可静态确定)
|
|
87
|
-
* · push 次数已超上限(早退)
|
|
88
|
-
* @returns {number|null} push 次数;规模未知为 null
|
|
89
|
-
*/
|
|
90
|
-
export function countUnsafePushes(tokens, name, ofIdx) {
|
|
91
|
-
let pushes = 0;
|
|
92
|
-
for (let j = 0; j < ofIdx; j++) {
|
|
93
|
-
const t = tokens[j];
|
|
94
|
-
// 展开运算符 `[...other]` 是 punct,会让规模不可静态确定
|
|
95
|
-
if (t.type === 'punct' && t.value === '...') return null;
|
|
96
|
-
if (t.type !== 'ident') continue;
|
|
97
|
-
// 合并/改写类方法:仅当是 **name.method** 形态才算(此前不区分方法归属,
|
|
98
|
-
// 扫描 0..ofIdx 会把声明前无关代码里的 concat/splice 误判为本数组的改写,
|
|
99
|
-
// 导致「声明后小数组循环」被错误判为规模未知 → 循环内 I/O 误报中风险)。
|
|
100
|
-
if (UNSAFE_ARRAY_METHODS.has(t.value)) {
|
|
101
|
-
if (tokens[j - 1]?.value === '.' && tokens[j - 2]?.value === name) return null;
|
|
102
|
-
continue;
|
|
103
|
-
}
|
|
104
|
-
if (t.value !== 'push') continue;
|
|
105
|
-
// 形态:name . push (
|
|
106
|
-
if (tokens[j - 1]?.value !== '.' || tokens[j - 2]?.value !== name) continue;
|
|
107
|
-
// push 若发生在循环体内(`for (const x of all) c.push(x)`),元素来自循环变量,
|
|
108
|
-
// 规模取决于被遍历的集合 —— 不可静态确定。
|
|
109
|
-
// 注意从 **push 本身** 的位置起扫:若从参数位置起扫,push 自己的 '(' 会让
|
|
110
|
-
// 深度立刻变负而被误判为「不在循环内」。
|
|
111
|
-
if (insideLoopOver(tokens, j)) return null;
|
|
112
|
-
pushes++;
|
|
113
|
-
if (pushes + 1 > SMALL_FIXED_LOOP_MAX) return null;
|
|
114
|
-
}
|
|
115
|
-
return pushes;
|
|
116
|
-
}
|
|
117
|
-
|
|
118
|
-
/**
|
|
119
|
-
* 循环头 `(...)` 的 token 下标范围 [openIdx, closeIdx];无括号返回 null。
|
|
120
|
-
*/
|
|
121
|
-
export function loopHeadSpan(tokens, i) {
|
|
122
|
-
// 只对关键字循环(for/while)剔除头部表达式:
|
|
123
|
-
// 数组方法(`.forEach(cb)` / `.find(cb)`)的括号里是**回调体**,其中的 I/O
|
|
124
|
-
// 会随每个元素执行,正是「循环内 I/O」,绝不能剔除(实测误判为 low)。
|
|
125
|
-
if (!LOOP_KEYWORDS.has(tokens[i].value)) return null;
|
|
126
|
-
let open = -1;
|
|
127
|
-
for (let j = i + 1; j < Math.min(tokens.length, i + LOOP_BODY_SEARCH_TOKENS); j++) {
|
|
128
|
-
const t = tokens[j];
|
|
129
|
-
if (t.type !== 'punct') continue;
|
|
130
|
-
if (t.value === '(') { open = j; break; }
|
|
131
|
-
if (t.value === '{' || t.value === ';') return null;
|
|
132
|
-
}
|
|
133
|
-
if (open === -1) return null;
|
|
134
|
-
const close = matchParenIn(tokens, open);
|
|
135
|
-
return close === -1 ? null : [open, close];
|
|
136
|
-
}
|
|
137
|
-
|
|
138
|
-
/** 前向圆括号配对:返回右括号下标;不匹配返回 -1。 */
|
|
139
|
-
function matchParenIn(tokens, openIdx) {
|
|
140
|
-
let depth = 0;
|
|
141
|
-
for (let j = openIdx; j < tokens.length; j++) {
|
|
142
|
-
const t = tokens[j];
|
|
143
|
-
if (t.type !== 'punct') continue;
|
|
144
|
-
if (t.value === '(') depth++;
|
|
145
|
-
else if (t.value === ')') { depth--; if (depth === 0) return j; }
|
|
146
|
-
}
|
|
147
|
-
return -1;
|
|
148
|
-
}
|
|
149
|
-
|
|
150
|
-
/**
|
|
151
|
-
* 位置 idx 的 push 是否发生在某个循环体内。
|
|
152
|
-
*
|
|
153
|
-
* 用于识别「从大集合填充」:`for (const x of all) c.push(x)` 静态看只 push 一次,
|
|
154
|
-
* 但实际执行次数 = all 的长度 —— 规模不可静态确定,必须按未知处理。
|
|
155
|
-
*
|
|
156
|
-
* 朴素但够用的判据:向前扫到最近的 `{`,若该 `{` 之前紧跟 `)` 且括号头是
|
|
157
|
-
* for/while/do,或直接是 `do {`,即认为在循环体内。
|
|
158
|
-
* @returns {boolean}
|
|
159
|
-
*/
|
|
160
|
-
export function insideLoopOver(tokens, idx) {
|
|
161
|
-
// 向左扫,遇到 `for`/`while` 关键字即认为在循环体内;
|
|
162
|
-
// 途中若先遇到语句边界(`;`)或块边界(`}`),说明已跨出当前语句,放弃。
|
|
163
|
-
// 不区分花括号体与单语句体 —— 两种都算。
|
|
164
|
-
let depth = 0;
|
|
165
|
-
for (let j = idx - 1; j >= 0; j--) {
|
|
166
|
-
const t = tokens[j];
|
|
167
|
-
if (t.type !== 'punct') {
|
|
168
|
-
// 未闭合的深度里遇到循环关键字:说明正处在它的头部/体内
|
|
169
|
-
if (t.type === 'ident' && LOOP_KEYWORDS.has(t.value)) return true;
|
|
170
|
-
continue;
|
|
171
|
-
}
|
|
172
|
-
if (t.value === ')' || t.value === ']' || t.value === '}') { depth++; continue; }
|
|
173
|
-
if (t.value === '(' || t.value === '[') { depth--; if (depth < 0) return false; continue; }
|
|
174
|
-
if (t.value === '{') {
|
|
175
|
-
if (depth > 0) { depth--; continue; }
|
|
176
|
-
return isLoopBodyBrace(tokens, j);
|
|
177
|
-
}
|
|
178
|
-
if (t.value === ';' && depth === 0) return false;
|
|
179
|
-
}
|
|
180
|
-
return false;
|
|
181
|
-
}
|
|
182
|
-
|
|
183
|
-
/** 闭合类括号:向右扫时用于累计深度。 */
|
|
184
|
-
|
|
185
|
-
/** 开启类括号:向左扫时用于消解深度。 */
|
|
186
|
-
|
|
187
|
-
/**
|
|
188
|
-
* 该 `{` 是否开启一个循环体。
|
|
189
|
-
* 块的**开**括号可能是循环体(`for (...) {`)—— 回看它前面是不是循环头;
|
|
190
|
-
* 不是则说明已跨出到别的作用域。
|
|
191
|
-
* @returns {boolean}
|
|
192
|
-
*/
|
|
193
|
-
export function isLoopBodyBrace(tokens, braceIdx) {
|
|
194
|
-
const prev = tokens[braceIdx - 1];
|
|
195
|
-
if (prev?.value === 'do') return true; // do { ... }
|
|
196
|
-
if (prev?.value !== ')') return false;
|
|
197
|
-
const paren = matchPairBack(tokens, braceIdx - 1, '(', ')');
|
|
198
|
-
if (paren === -1) return false;
|
|
199
|
-
const kw = tokens[paren - 1]?.value;
|
|
200
|
-
return kw === 'for' || kw === 'while';
|
|
201
|
-
}
|
|
202
|
-
|
|
203
|
-
/** 数 `[ ... ]` 字面量数组的元素个数(顶层逗号 + 1;空数组为 0)。 */
|
|
204
|
-
|
|
205
|
-
/**
|
|
206
|
-
* 形如 `for (const x of ['a', 'b', 'c'])`:元素数固定且 ≤ SMALL_FIXED_LOOP_MAX。
|
|
207
|
-
* 只认字面量数组字面形态;变量/函数调用一律视为规模未知(保守判高风险)。
|
|
208
|
-
* @returns {boolean}
|
|
209
|
-
*/
|
|
210
|
-
export function isSmallFixedLoop(tokens, i) {
|
|
211
|
-
// ① 数组方法形式(`['a','b'].some((n) => ...)`):受调用的数组若是字面量,
|
|
212
|
-
// 规模同样固定且极小。这一形态不走 for...of,早先完全未覆盖。
|
|
213
|
-
if (LOOP_METHODS.has(tokens[i].value) && isSmallLiteralCallee(tokens, i)) return true;
|
|
214
|
-
|
|
215
|
-
// ② for...of / for...in 形态:看被遍历的表达式
|
|
216
|
-
const ofIdx = findOfKeyword(tokens, i);
|
|
217
|
-
if (ofIdx === -1) return false;
|
|
218
|
-
|
|
219
|
-
const open = tokens[ofIdx + 1];
|
|
220
|
-
if (!open) return false;
|
|
221
|
-
|
|
222
|
-
// 变量数组(`for (const f of candidates)`):回看它的声明,若以字面量数组起手
|
|
223
|
-
// 且仅用少量 push 追加,则规模固定且极小 —— 与字面量数组同义,按小集合处理。
|
|
224
|
-
if (open.value !== '[') {
|
|
225
|
-
return open.type === 'ident' && isSmallPushedArray(tokens, open.value, ofIdx);
|
|
226
|
-
}
|
|
227
|
-
// ③ 字面量数组形态:数顶层元素个数
|
|
228
|
-
return countLiteralElements(tokens, ofIdx + 1) <= SMALL_FIXED_LOOP_MAX;
|
|
229
|
-
}
|
|
230
|
-
|
|
231
|
-
/**
|
|
232
|
-
* 在循环头内定位 `of` / `in` 关键字;找不到返回 -1。
|
|
233
|
-
* 扫到 `{` 或 `;` 即停 —— 说明已越过循环头。
|
|
234
|
-
* @returns {number} 关键字 token 下标
|
|
235
|
-
*/
|
|
236
|
-
export function findOfKeyword(tokens, i) {
|
|
237
|
-
const limit = Math.min(tokens.length, i + LOOP_BODY_SEARCH_TOKENS);
|
|
238
|
-
for (let j = i + 1; j < limit; j++) {
|
|
239
|
-
const tj = tokens[j];
|
|
240
|
-
if (tj.type === 'ident' && (tj.value === 'of' || tj.value === 'in')) return j;
|
|
241
|
-
if (tj.type === 'punct' && (tj.value === '{' || tj.value === ';')) break;
|
|
242
|
-
}
|
|
243
|
-
return -1;
|
|
244
|
-
}
|
|
245
|
-
|
|
246
|
-
/**
|
|
247
|
-
* 从 `[` 起数顶层元素个数,判断是否为「固定小集合」(存在元素且 ≤ 上限)。
|
|
248
|
-
*
|
|
249
|
-
* 计数规则:深度回到 0 时遇 `]` 收尾;深度 1 上的逗号即分隔符。
|
|
250
|
-
* 元素数超上限时早退,避免长数组上白扫。
|
|
251
|
-
* @param {number} fromIdx 数组 `[` 的下标
|
|
252
|
-
* @returns {boolean}
|
|
253
|
-
*/
|
|
254
|
-
export function countLiteralElements(tokens, fromIdx) {
|
|
255
|
-
let depth = 0;
|
|
256
|
-
let count = 0;
|
|
257
|
-
let hasElement = false;
|
|
258
|
-
for (let j = fromIdx; j < tokens.length; j++) {
|
|
259
|
-
const tj = tokens[j];
|
|
260
|
-
if (tj.type !== 'punct') { hasElement = true; continue; }
|
|
261
|
-
if (tj.value === '[' || tj.value === '(' || tj.value === '{') { depth++; continue; }
|
|
262
|
-
if (tj.value === ']') {
|
|
263
|
-
depth--;
|
|
264
|
-
if (depth === 0) return hasElement ? count + 1 : 0;
|
|
265
|
-
continue;
|
|
266
|
-
}
|
|
267
|
-
if (tj.value === ')' || tj.value === '}') { depth--; continue; }
|
|
268
|
-
if (tj.value === ',' && depth === 1) { count++; hasElement = true; }
|
|
269
|
-
if (count + 1 > SMALL_FIXED_LOOP_MAX) return count + 1; // 超上限早退(已知大于上限)
|
|
270
|
-
}
|
|
271
|
-
return false;
|
|
272
|
-
}
|
|
273
|
-
|
|
274
|
-
/**
|
|
275
|
-
* 收集所有函数体范围(含 async 标记)与循环体范围,用于上下文归属判断。
|
|
276
|
-
*
|
|
277
|
-
* 刻意不跳过函数体:函数内还嵌着循环与内层函数,跳过会漏收内层范围。
|
|
278
|
-
* @param {Array} tokens tokenize 结果
|
|
279
|
-
* @returns {{asyncFns: Array, loops: Array}}
|
|
280
|
-
*/
|
|
281
|
-
export function collectRanges(tokens) {
|
|
282
|
-
const asyncFns = [];
|
|
283
|
-
const allFns = [];
|
|
284
|
-
const loops = [];
|
|
285
|
-
for (let i = 0; i < tokens.length; i++) {
|
|
286
|
-
const t = tokens[i];
|
|
287
|
-
if (t.type !== 'ident') continue;
|
|
288
|
-
const fnRange = classifyFnHead(tokens, i, t);
|
|
289
|
-
if (fnRange) { asyncFns.push(fnRange); allFns.push(fnRange); continue; }
|
|
290
|
-
// 同步函数/箭头函数/对象方法简写:不参与 inAsync,但要参与「请求路径」判定
|
|
291
|
-
// (否则同步 HTTP handler 里的 I/O 被误判成「启动路径一次性」而低估风险)。
|
|
292
|
-
const syncRange = classifySyncFnHead(tokens, i, t);
|
|
293
|
-
if (syncRange) { allFns.push(syncRange); continue; }
|
|
294
|
-
if (isLoopHead(t, tokens[i - 1])) {
|
|
295
|
-
const range = findLoopRange(tokens, i);
|
|
296
|
-
// 固定小字面量数组(`for (const n of ['a','b','c'])`)迭代次数确定且极少,
|
|
297
|
-
// 循环内同步 I/O 的开销 = 元素数个系统调用,不构成「逐次阻塞放大耗时」——
|
|
298
|
-
// 标出来供判定层降级(实测:3 个 SSH key 名的探测循环被误判为高风险)。
|
|
299
|
-
if (range) {
|
|
300
|
-
// 第 4 项 = 循环头 `(...)` 的 token 下标范围,用于剔除「迭代器表达式内的 I/O」:
|
|
301
|
-
// `for (const line of readFileSync(f).split('\n'))` 里 readFileSync 只在
|
|
302
|
-
// 进入循环前执行一次,把它算作循环内 I/O 是误报(实测 repos.js:81)。
|
|
303
|
-
loops.push([range[0], range[1], isSmallFixedLoop(tokens, i), loopHeadSpan(tokens, i)]);
|
|
304
|
-
}
|
|
305
|
-
}
|
|
306
|
-
}
|
|
307
|
-
return { asyncFns, allFns, loops };
|
|
308
|
-
}
|
package/lib/ast/io-risk.js
DELETED
|
@@ -1,314 +0,0 @@
|
|
|
1
|
-
/**
|
|
2
|
-
* dsh-git-push — I/O 风险分级(AST 层,)
|
|
3
|
-
*
|
|
4
|
-
* 职责:找出文件里的全部 fs 调用(同步 + 异步),判定每个调用的上下文
|
|
5
|
-
* (异步路径 / 循环内 / 请求处理路径 / 启动路径)与操作类别(读 / 写 / 删 / 改名),
|
|
6
|
-
* 据此给出四级风险,供审计 findings 与 scripts/scan-file-io.mjs 共用。
|
|
7
|
-
*
|
|
8
|
-
* 分层约定(README「正则初筛 → AST 精筛」):本文件属 `lib/ast/` 实现层,
|
|
9
|
-
* 只把判定结果交出去;调用方(lib/checks/)只做「转 finding / 转报表」。
|
|
10
|
-
*
|
|
11
|
-
* 四级风险(写类操作加权一档;循环内 I/O 改多维分级,不再一刀切 high):
|
|
12
|
-
* high —— 异步路径中的同步 I/O;或循环内「同步阻塞 + 边界不可控」落在请求路径;
|
|
13
|
-
* 或请求路径上循环内逐个 await(响应时间线性增长)
|
|
14
|
-
* medium —— 请求处理路径上的 I/O;循环内同步 I/O(阻塞事件循环,边界不可控但非关键路径)
|
|
15
|
-
* low —— 启动路径上的同步 I/O(模块顶层 / apply() 等一次性初始化);
|
|
16
|
-
* 循环内异步串行 await(只慢不阻塞);固定小数组循环内的 I/O;
|
|
17
|
-
* 循环内异步 I/O 已 Promise.all 并行
|
|
18
|
-
* safe —— 异步 + 无循环 + 非关键路径
|
|
19
|
-
*
|
|
20
|
-
* 模块划分(原单文件按职责拆分,对外出口不变):
|
|
21
|
-
* io-risk-const.js —— 常量与档位工具
|
|
22
|
-
* io-risk-fn.js —— 函数边界识别与 token 配对
|
|
23
|
-
* io-risk-loop.js —— 循环判定
|
|
24
|
-
* 本文件 —— 判定与评分,并再导出上述模块的公共符号
|
|
25
|
-
*/
|
|
26
|
-
import { tokenize } from './tokenizer.js';
|
|
27
|
-
import {
|
|
28
|
-
LOOP_KEYWORDS, LOOP_METHODS, RISK_BADGE, RISK_LABEL,
|
|
29
|
-
SYNC_FS_FNS, ASYNC_FS_FNS, KIND_BY_FN, REQUEST_PATTERNS, LEVELS, raise, META_FS_FNS,
|
|
30
|
-
} from './io-risk-const.js';
|
|
31
|
-
import { buildCallGraph, isInRequestPath } from './callgraph.js';
|
|
32
|
-
import { collectRanges } from './io-risk-loop.js';
|
|
33
|
-
import { enclosingLoop } from './io-risk-fn.js';
|
|
34
|
-
import { isSmallFixedLoop } from './io-risk-loop.js';
|
|
35
|
-
|
|
36
|
-
|
|
37
|
-
/**
|
|
38
|
-
* 扫描文件内全部 fs 调用并分级。
|
|
39
|
-
*
|
|
40
|
-
* @param {string} text 文件全文
|
|
41
|
-
* @returns {Array<{
|
|
42
|
-
* line:number, call:string, via:string, type:'sync'|'async', kind:'read'|'write'|'delete'|'rename',
|
|
43
|
-
* inAsync:boolean, inLoop:boolean, inRequest:boolean, inStartup:boolean,
|
|
44
|
-
* risk:'high'|'medium'|'low'|'safe', reason:string
|
|
45
|
-
* }>}
|
|
46
|
-
*/
|
|
47
|
-
/** 模块顶层/初始化的判定常量(魔数具名化,说明取值理由)。 */
|
|
48
|
-
|
|
49
|
-
/**
|
|
50
|
-
* 判断 token 位置是否是一个真实的 fs 调用。
|
|
51
|
-
* 三个条件:① 标识符是已知 fs 函数名 ② 后一个 token 是 '('(调用形态)
|
|
52
|
-
* ③ 同步/异步归类。拆自 judgeIoTokend(原函数圈复杂度 12,超阈值 10)。
|
|
53
|
-
* @returns {{token: object, isSync: boolean}|null}
|
|
54
|
-
*/
|
|
55
|
-
function classifyIoCall(tokens, idx) {
|
|
56
|
-
const t = tokens[idx];
|
|
57
|
-
if (t.type !== 'ident') return null;
|
|
58
|
-
const isSync = SYNC_FS_FNS.has(t.value);
|
|
59
|
-
const isAsync = !isSync && ASYNC_FS_FNS.has(t.value);
|
|
60
|
-
if (!isSync && !isAsync) return null;
|
|
61
|
-
// 必须是调用形态:名字后紧跟 '('(允许 fs.promises.readFile 的点号形态)
|
|
62
|
-
const next = tokens[idx + 1];
|
|
63
|
-
if (!next || next.type !== 'punct' || next.value !== '(') return null;
|
|
64
|
-
return { token: t, isSync };
|
|
65
|
-
}
|
|
66
|
-
|
|
67
|
-
/**
|
|
68
|
-
* 判定单条 I/O token 的上下文与风险等级。
|
|
69
|
-
* 从 scanIoRiskAst 抽出(原函数圈复杂度 27 / 嵌套 4,超阈值),
|
|
70
|
-
* 使主循环只负责遍历与筛形,判定逻辑可独立测试。
|
|
71
|
-
*
|
|
72
|
-
* @param {object} ctx { tokens, lines, idx, asyncFns, loops, within, enclosingFnText }
|
|
73
|
-
* @returns {object|null} 命中项;非 I/O 调用返回 null
|
|
74
|
-
*/
|
|
75
|
-
function judgeIoTokend(ctx) {
|
|
76
|
-
const { tokens, lines, idx, asyncFns, loops, within, enclosingFnText, enclosingFnName, promiseAllRanges, callGraph } = ctx;
|
|
77
|
-
const callKind = classifyIoCall(tokens, idx);
|
|
78
|
-
if (!callKind) return null;
|
|
79
|
-
const { token: t, isSync } = callKind;
|
|
80
|
-
|
|
81
|
-
const line = t.line;
|
|
82
|
-
const inAsync = within(asyncFns, line);
|
|
83
|
-
const loopHit = enclosingLoop(loops, line);
|
|
84
|
-
// 迭代器表达式内的 I/O 只执行一次(进入循环前),不算「循环内」:
|
|
85
|
-
// `for (const line of readFileSync(f).split('\n'))` —— readFileSync 在头部括号内。
|
|
86
|
-
// 同行代码无法靠行号区分,故按 token 下标判定。
|
|
87
|
-
const inLoopHead = Boolean(loopHit && loopHit[3] && idx > loopHit[3][0] && idx < loopHit[3][1]);
|
|
88
|
-
const inLoop = Boolean(loopHit) && !inLoopHead;
|
|
89
|
-
// 只遍历固定小字面量数组的循环:I/O 次数确定且极少,不按「逐次阻塞放大耗时」计
|
|
90
|
-
const inSmallFixedLoop = Boolean(loopHit && loopHit[2]);
|
|
91
|
-
// Promise.all(...) 表达式内的 I/O = 并行执行(不串行放大),
|
|
92
|
-
// 循环内逐次 await 才是串行放大;并行形态降级为可接受
|
|
93
|
-
const inParallelAll = Boolean(promiseAllRanges && promiseAllRanges.some(([s, e]) => idx > s && idx < e));
|
|
94
|
-
// P3:请求路径判定升级——函数体内 HTTP 特征(detectRequestContext)
|
|
95
|
-
// 或「被请求路径函数调用」(调用链 isInRequestPath)——修「请求处理调用的公共函数漏报」
|
|
96
|
-
const inRequest = detectRequestContext(lines, enclosingFnText(line))
|
|
97
|
-
|| (callGraph && enclosingFnName(line) ? isInRequestPath(enclosingFnName(line), callGraph) : false);
|
|
98
|
-
const inStartup = !inAsync && !inLoop && !inRequest;
|
|
99
|
-
const type = isSync ? 'sync' : 'async';
|
|
100
|
-
const kind = META_FS_FNS.has(t.value) ? 'meta' : (KIND_BY_FN[t.value] || 'read'); // 元数据/查询/幂等目录操作独立档
|
|
101
|
-
|
|
102
|
-
const { risk, reason } = gradeRisk({ isSync, inAsync, inLoop, inSmallFixedLoop, inRequest, inParallelAll, kind });
|
|
103
|
-
return { line, call: t.value, via: '', type, kind, inAsync, inLoop, inRequest, inStartup, risk, reason };
|
|
104
|
-
}
|
|
105
|
-
|
|
106
|
-
/** 请求路径判定:所在函数体内出现 HTTP 特征即算(文本回溯,回溯上限见常量)。 */
|
|
107
|
-
function detectRequestContext(lines, fnRange) {
|
|
108
|
-
if (!fnRange) return false;
|
|
109
|
-
const from = Math.max(0, fnRange[0] - 1);
|
|
110
|
-
const body = lines.slice(from, fnRange[1]).join('\n');
|
|
111
|
-
return REQUEST_PATTERNS.some((re) => re.test(body));
|
|
112
|
-
}
|
|
113
|
-
|
|
114
|
-
/**
|
|
115
|
-
* 四级判定:上下文 + 操作类别。
|
|
116
|
-
* 基础档由上下文决定,写/删/改名再按「数据不可逆」加权一档。
|
|
117
|
-
* 循环内 I/O 多维分级(不再一刀切 high):
|
|
118
|
-
* 风险不取决于「在不在循环里」,而取决于「循环次数可不可控 + I/O 同不同步 +
|
|
119
|
-
* 在不在请求路径上」——三者同时满足才是真正高风险;只满足一个或两个降级为提醒。
|
|
120
|
-
* 同步阻塞 + 边界不可控:medium(非关键路径)/ high(请求路径)
|
|
121
|
-
* 异步串行 await:low(只慢不阻塞)/ high(请求路径,响应时间线性增长)
|
|
122
|
-
* 异步并行(Promise.all):low(可接受)
|
|
123
|
-
* @returns {{risk: string, reason: string}}
|
|
124
|
-
*/
|
|
125
|
-
function gradeRisk({ isSync, inAsync, inLoop, inSmallFixedLoop, inRequest, inParallelAll, kind }) {
|
|
126
|
-
let risk = 'safe';
|
|
127
|
-
let reason = '异步 + 无循环 + 非关键路径';
|
|
128
|
-
// 「会重复执行」= 循环内(逐次执行)或请求路径(每次请求执行一次)。
|
|
129
|
-
// 只有这类上下文才存在「反复 + 不可逆」的叠加;异步路径的同步 I/O 本身已是最高档,
|
|
130
|
-
// 且它的风险来自「阻塞事件循环」而非重复,不参与加权。
|
|
131
|
-
const repeated = (inLoop && !inSmallFixedLoop) || inRequest;
|
|
132
|
-
if (isSync && inAsync) { risk = 'high'; reason = '异步路径中的同步 I/O(阻塞事件循环)'; }
|
|
133
|
-
else if (inLoop && inSmallFixedLoop) { risk = 'low'; reason = '固定小数组循环内的 I/O(次数确定且极少,开销可忽略)'; }
|
|
134
|
-
else if (inLoop && isSync) {
|
|
135
|
-
// 循环内同步 I/O(边界不可控,已排除 smallFixed):阻塞事件循环;
|
|
136
|
-
// 落在请求路径再升一档(不可控 + 同步 + 关键路径三条件齐)
|
|
137
|
-
risk = inRequest ? 'high' : 'medium';
|
|
138
|
-
reason = inRequest
|
|
139
|
-
? '请求路径上循环内同步 I/O(阻塞事件循环 + 次数不可控,逐次放大)'
|
|
140
|
-
: '循环内同步 I/O(阻塞事件循环;边界不可控时逐次放大耗时)';
|
|
141
|
-
}
|
|
142
|
-
else if (inLoop && inParallelAll) { risk = 'low'; reason = '循环内异步 I/O 已 Promise.all 并行(不串行放大)'; }
|
|
143
|
-
else if (inLoop && inRequest) { risk = 'high'; reason = '请求处理路径上循环内逐个 await I/O(响应时间线性增长)'; }
|
|
144
|
-
else if (inLoop) { risk = 'low'; reason = '循环内异步 I/O(串行 await 只慢不阻塞事件循环)'; }
|
|
145
|
-
// (io-risk 优化,对照诊断「97% 误报」):请求路径上的**元数据/查询/幂等目录操作**
|
|
146
|
-
// (existsSync/statSync/readdirSync/mkdir(recursive)/rmdir…)单文件微秒级、不搬运数据内容,
|
|
147
|
-
// 不构成请求路径阻塞 → 降 low(不再按「请求路径 I/O」一律 medium)。
|
|
148
|
-
// 注意:循环内的 meta 已被上方循环分级接管(循环内逐次 stat 累积按循环规则判),不受本档豁免。
|
|
149
|
-
else if (inRequest && kind === 'meta') { risk = 'low'; reason = '请求处理路径上的元数据/查询/幂等目录操作(单文件微秒级,无数据内容搬运)'; }
|
|
150
|
-
else if (inRequest) { risk = 'medium'; reason = '请求处理路径上的 I/O'; }
|
|
151
|
-
else if (isSync) { risk = 'low'; reason = '启动路径上的同步 I/O(一次性初始化)'; }
|
|
152
|
-
|
|
153
|
-
// 写/删加权一档(数据不可逆)——**仅当该次 I/O 会重复执行时**。
|
|
154
|
-
// 加权要表达的是「反复执行 + 不可逆」的叠加风险:循环内逐次写盘、请求路径并发写。
|
|
155
|
-
// 启动路径的一次性落盘(含 `.tmp` + rename 的标准原子写)执行一次即结束,
|
|
156
|
-
// 不存在重复放大;对它加权会把这套**正确做法**报成「中风险」,
|
|
157
|
-
// 实测 25 条 atomic-json/account-status/scan-repos 等原子写全被误升档。
|
|
158
|
-
// rename 同卷改名 = 元数据级(非数据内容搬运,DELETE 移 .trash 等)→ 不再参与加权。
|
|
159
|
-
if (repeated && (kind === 'write' || kind === 'delete')) {
|
|
160
|
-
const raised = raise(risk, 1);
|
|
161
|
-
if (raised !== risk) { reason += `;${kind} 类操作加权`; risk = raised; }
|
|
162
|
-
}
|
|
163
|
-
return { risk, reason };
|
|
164
|
-
}
|
|
165
|
-
|
|
166
|
-
/** 从 openIdx(应为 `(`)起配对最近的闭合括号,返回闭括号 token 下标(无配对返回 -1)。 */
|
|
167
|
-
function findMatchingClose(tokens, openIdx) {
|
|
168
|
-
let depth = 0;
|
|
169
|
-
for (let j = openIdx; j < tokens.length; j++) {
|
|
170
|
-
if (tokens[j].type === 'punct' && tokens[j].value === '(') depth++;
|
|
171
|
-
else if (tokens[j].type === 'punct' && tokens[j].value === ')') {
|
|
172
|
-
depth--;
|
|
173
|
-
if (depth === 0) return j;
|
|
174
|
-
}
|
|
175
|
-
}
|
|
176
|
-
return -1;
|
|
177
|
-
}
|
|
178
|
-
|
|
179
|
-
/**
|
|
180
|
-
* Promise.all(...) 表达式的 token 区间。
|
|
181
|
-
* 循环内 I/O 若落在 Promise.all 参数里 = 并行执行(不串行放大)→ 降级可接受。
|
|
182
|
-
* 判定依据:`Promise` `.` `all` `(` 四连 token,配对其括号区间 [开, 闭]。
|
|
183
|
-
* @param {Array} tokens tokenize 产物
|
|
184
|
-
* @returns {Array<[number, number]>} token 下标区间(含开闭括号)
|
|
185
|
-
*/
|
|
186
|
-
function findPromiseAllRanges(tokens) {
|
|
187
|
-
const ranges = [];
|
|
188
|
-
for (let i = 0; i < tokens.length - 3; i++) {
|
|
189
|
-
const a = tokens[i], b = tokens[i + 1], c = tokens[i + 2], d = tokens[i + 3];
|
|
190
|
-
const isPromiseAll = a.type === 'ident' && a.value === 'Promise'
|
|
191
|
-
&& b.type === 'punct' && b.value === '.'
|
|
192
|
-
&& c.type === 'ident' && c.value === 'all'
|
|
193
|
-
&& d.type === 'punct' && d.value === '(';
|
|
194
|
-
if (!isPromiseAll) continue;
|
|
195
|
-
const close = findMatchingClose(tokens, i + 3);
|
|
196
|
-
if (close > i + 3) ranges.push([i + 3, close]);
|
|
197
|
-
}
|
|
198
|
-
return ranges;
|
|
199
|
-
}
|
|
200
|
-
|
|
201
|
-
/**
|
|
202
|
-
* 扫描源码里的文件 I/O,按上下文与操作类别给出四级风险。
|
|
203
|
-
* @param {string} text 源码全文
|
|
204
|
-
* @returns {Array<object>} 命中项数组(见文件头字段说明)
|
|
205
|
-
*/
|
|
206
|
-
export function scanIoRiskAst(text = '') {
|
|
207
|
-
const src = String(text);
|
|
208
|
-
const tokens = tokenize(src);
|
|
209
|
-
const { asyncFns, allFns, loops } = collectRanges(tokens);
|
|
210
|
-
const callGraph = buildCallGraph(src); // P3:单文件调用链——被请求路径函数调用的函数也判请求路径(修漏报)
|
|
211
|
-
const promiseAllRanges = findPromiseAllRanges(tokens);
|
|
212
|
-
const lines = src.split('\n');
|
|
213
|
-
const out = [];
|
|
214
|
-
|
|
215
|
-
const within = (ranges, line) => ranges.some(([s, e]) => line >= s && line <= e);
|
|
216
|
-
/**
|
|
217
|
-
* 取最近函数体范围(含区间嵌套时取最内层,用于请求路径特征匹配)。
|
|
218
|
-
* 必须用 allFns(含同步函数/箭头函数/方法简写)—— 请求路径判定看的是
|
|
219
|
-
* 「这段函数体里有没有 HTTP 特征」,与函数是否 async 无关。
|
|
220
|
-
* 早先误用 asyncFns,导致同步 handler(`function h(req,res){ res.end() }`)
|
|
221
|
-
* 识别不到请求上下文,其中的 I/O 被降格成「启动路径一次性」→ 低估风险。
|
|
222
|
-
*/
|
|
223
|
-
const enclosingFnText = (line) => {
|
|
224
|
-
let best = null;
|
|
225
|
-
for (const [s, e] of allFns) if (line >= s && line <= e) best = [s, e];
|
|
226
|
-
return best;
|
|
227
|
-
};
|
|
228
|
-
|
|
229
|
-
const fnNameAt = (line) => {
|
|
230
|
-
let best = null;
|
|
231
|
-
for (const [sf, ef] of allFns) if (line >= sf && line <= ef) { best = [sf, ef]; break; }
|
|
232
|
-
if (!best) return '';
|
|
233
|
-
const head = src.split('\n')[best[0] - 1] || '';
|
|
234
|
-
const match = head.match(/(?:function\s+)?([A-Za-z_$][\w$]*)\s*(?:=\s*)?(?:\(|=>)/);
|
|
235
|
-
return match ? match[1] : '';
|
|
236
|
-
};
|
|
237
|
-
const ctx = { tokens, lines, asyncFns, loops, within, enclosingFnText, enclosingFnName: fnNameAt, promiseAllRanges, callGraph };
|
|
238
|
-
for (let i = 0; i < tokens.length; i++) {
|
|
239
|
-
const hit = judgeIoTokend({ ...ctx, idx: i });
|
|
240
|
-
if (!hit) continue;
|
|
241
|
-
out.push(hit);
|
|
242
|
-
i += 1;
|
|
243
|
-
}
|
|
244
|
-
return out;
|
|
245
|
-
}
|
|
246
|
-
|
|
247
|
-
/**
|
|
248
|
-
* 统计 I/O 分布(供报表与审计摘要用)。
|
|
249
|
-
* @param {Array} hits scanIoRiskAst 的结果
|
|
250
|
-
* @returns {{total:number, sync:number, async:number, syncRatio:number, byRisk:object, byKind:object, byFile:Array}}
|
|
251
|
-
*/
|
|
252
|
-
export function summarizeIoRisk(hits = []) {
|
|
253
|
-
const list = Array.isArray(hits) ? hits : [];
|
|
254
|
-
const sync = list.filter((h) => h.type === 'sync').length;
|
|
255
|
-
const byRisk = { high: 0, medium: 0, low: 0, safe: 0 };
|
|
256
|
-
const byKind = { read: 0, write: 0, delete: 0, rename: 0 };
|
|
257
|
-
const fileMap = new Map();
|
|
258
|
-
for (const hit of list) {
|
|
259
|
-
byRisk[hit.risk] = (byRisk[hit.risk] || 0) + 1;
|
|
260
|
-
byKind[hit.kind] = (byKind[hit.kind] || 0) + 1;
|
|
261
|
-
if (hit.file) fileMap.set(hit.file, (fileMap.get(hit.file) || 0) + 1);
|
|
262
|
-
}
|
|
263
|
-
const byFile = [...fileMap.entries()]
|
|
264
|
-
.map(([file, count]) => ({ file, count }))
|
|
265
|
-
.sort((a, b) => b.count - a.count || a.file.localeCompare(b.file));
|
|
266
|
-
return {
|
|
267
|
-
total: list.length,
|
|
268
|
-
sync,
|
|
269
|
-
async: list.length - sync,
|
|
270
|
-
syncRatio: list.length ? sync / list.length : 0,
|
|
271
|
-
byRisk, byKind, byFile,
|
|
272
|
-
};
|
|
273
|
-
}
|
|
274
|
-
|
|
275
|
-
/**
|
|
276
|
-
* 生成改造优先级清单:按风险级别排序(同级内写/删优先、同步优先)。
|
|
277
|
-
* @param {Array} hits 每条需带 file 字段
|
|
278
|
-
* @returns {Array} 排序后的清单(含 rank 序号)
|
|
279
|
-
*/
|
|
280
|
-
export function rankIoFixList(hits = []) {
|
|
281
|
-
const list = Array.isArray(hits) ? hits : [];
|
|
282
|
-
const sorted = [...list].sort(compareFixPriority);
|
|
283
|
-
return sorted.map((hit, idx) => ({ ...hit, rank: idx + 1 }));
|
|
284
|
-
}
|
|
285
|
-
|
|
286
|
-
/**
|
|
287
|
-
* 改造优先级排序:风险 > 写类 > 同步 > 文件 > 行号。
|
|
288
|
-
* 从 rankIoFixList 的比较器抽出(原为内联多级比较,变量名过短且难读)。
|
|
289
|
-
* @returns {number} 负数 = a 优先
|
|
290
|
-
*/
|
|
291
|
-
function compareFixPriority(a, b) {
|
|
292
|
-
const byRisk = LEVELS.indexOf(b.risk) - LEVELS.indexOf(a.risk);
|
|
293
|
-
if (byRisk) return byRisk;
|
|
294
|
-
const byKind = isWriteLike(b.kind) - isWriteLike(a.kind);
|
|
295
|
-
if (byKind) return byKind;
|
|
296
|
-
const byType = isSync(b.type) - isSync(a.type);
|
|
297
|
-
if (byType) return byType;
|
|
298
|
-
const byFile = String(a.file || '').localeCompare(String(b.file || ''));
|
|
299
|
-
if (byFile) return byFile;
|
|
300
|
-
return (a.line || 0) - (b.line || 0);
|
|
301
|
-
}
|
|
302
|
-
|
|
303
|
-
/** 写/删/改名优先于读(数据不可逆,改造收益更高)。 */
|
|
304
|
-
function isWriteLike(kind) {
|
|
305
|
-
return kind === 'read' ? 0 : 1;
|
|
306
|
-
}
|
|
307
|
-
|
|
308
|
-
/** 同步调用优先于异步(阻塞事件循环,收益直接)。 */
|
|
309
|
-
function isSync(type) {
|
|
310
|
-
return type === 'sync' ? 1 : 0;
|
|
311
|
-
}
|
|
312
|
-
|
|
313
|
-
// ── 对外出口:保持与原单文件一致,调用方无需改动 ──
|
|
314
|
-
export { RISK_BADGE, RISK_LABEL };
|