@vibe-agent-toolkit/resources 0.1.38 → 0.1.39-rc.2
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 +33 -24
- package/dist/frontmatter-link-validator.d.ts +9 -9
- package/dist/frontmatter-link-validator.d.ts.map +1 -1
- package/dist/frontmatter-link-validator.js +28 -27
- package/dist/frontmatter-link-validator.js.map +1 -1
- package/dist/frontmatter-validator.d.ts +3 -2
- package/dist/frontmatter-validator.d.ts.map +1 -1
- package/dist/frontmatter-validator.js +8 -14
- package/dist/frontmatter-validator.js.map +1 -1
- package/dist/html-link-parser.d.ts +47 -0
- package/dist/html-link-parser.d.ts.map +1 -0
- package/dist/html-link-parser.js +111 -0
- package/dist/html-link-parser.js.map +1 -0
- package/dist/html-transform.d.ts +45 -0
- package/dist/html-transform.d.ts.map +1 -0
- package/dist/html-transform.js +116 -0
- package/dist/html-transform.js.map +1 -0
- package/dist/index.d.ts +4 -1
- package/dist/index.d.ts.map +1 -1
- package/dist/index.js +3 -1
- package/dist/index.js.map +1 -1
- package/dist/link-parser.d.ts +26 -2
- package/dist/link-parser.d.ts.map +1 -1
- package/dist/link-parser.js +40 -6
- package/dist/link-parser.js.map +1 -1
- package/dist/link-validator.d.ts +53 -6
- package/dist/link-validator.d.ts.map +1 -1
- package/dist/link-validator.js +100 -131
- package/dist/link-validator.js.map +1 -1
- package/dist/multi-schema-validator.d.ts.map +1 -1
- package/dist/multi-schema-validator.js +6 -8
- package/dist/multi-schema-validator.js.map +1 -1
- package/dist/resource-registry.d.ts +56 -13
- package/dist/resource-registry.d.ts.map +1 -1
- package/dist/resource-registry.js +165 -66
- package/dist/resource-registry.js.map +1 -1
- package/dist/schemas/link-auth.d.ts +382 -0
- package/dist/schemas/link-auth.d.ts.map +1 -0
- package/dist/schemas/link-auth.js +124 -0
- package/dist/schemas/link-auth.js.map +1 -0
- package/dist/schemas/project-config.d.ts +763 -171
- package/dist/schemas/project-config.d.ts.map +1 -1
- package/dist/schemas/project-config.js +5 -0
- package/dist/schemas/project-config.js.map +1 -1
- package/dist/schemas/resource-metadata.d.ts +46 -10
- package/dist/schemas/resource-metadata.d.ts.map +1 -1
- package/dist/schemas/resource-metadata.js +14 -1
- package/dist/schemas/resource-metadata.js.map +1 -1
- package/dist/schemas/validation-result.d.ts +36 -57
- package/dist/schemas/validation-result.d.ts.map +1 -1
- package/dist/schemas/validation-result.js +5 -27
- package/dist/schemas/validation-result.js.map +1 -1
- package/dist/types/resources.d.ts +1 -1
- package/dist/types/resources.d.ts.map +1 -1
- package/dist/types/resources.js.map +1 -1
- package/dist/utils.d.ts +10 -0
- package/dist/utils.d.ts.map +1 -1
- package/dist/utils.js +14 -0
- package/dist/utils.js.map +1 -1
- package/package.json +4 -3
- package/src/frontmatter-link-validator.ts +28 -30
- package/src/frontmatter-validator.ts +19 -16
- package/src/html-link-parser.ts +140 -0
- package/src/html-transform.ts +157 -0
- package/src/index.ts +6 -1
- package/src/link-parser.ts +60 -11
- package/src/link-validator.ts +175 -145
- package/src/multi-schema-validator.ts +10 -8
- package/src/resource-registry.ts +205 -72
- package/src/schemas/link-auth.ts +146 -0
- package/src/schemas/project-config.ts +6 -0
- package/src/schemas/resource-metadata.ts +17 -1
- package/src/schemas/validation-result.ts +5 -29
- package/src/types/resources.ts +2 -1
- package/src/utils.ts +15 -0
package/package.json
CHANGED
|
@@ -1,6 +1,6 @@
|
|
|
1
1
|
{
|
|
2
2
|
"name": "@vibe-agent-toolkit/resources",
|
|
3
|
-
"version": "0.1.
|
|
3
|
+
"version": "0.1.39-rc.2",
|
|
4
4
|
"description": "Markdown resource parsing, validation, and link integrity checking",
|
|
5
5
|
"type": "module",
|
|
6
6
|
"main": "./dist/index.js",
|
|
@@ -33,12 +33,13 @@
|
|
|
33
33
|
"author": "Jeff Dutton",
|
|
34
34
|
"license": "MIT",
|
|
35
35
|
"dependencies": {
|
|
36
|
-
"@vibe-agent-toolkit/agent-schema": "0.1.
|
|
37
|
-
"@vibe-agent-toolkit/utils": "0.1.
|
|
36
|
+
"@vibe-agent-toolkit/agent-schema": "0.1.39-rc.2",
|
|
37
|
+
"@vibe-agent-toolkit/utils": "0.1.39-rc.2",
|
|
38
38
|
"ajv": "^8.17.1",
|
|
39
39
|
"ajv-formats": "^3.0.1",
|
|
40
40
|
"github-slugger": "^2.0.0",
|
|
41
41
|
"markdown-link-check": "^3.14.2",
|
|
42
|
+
"parse5": "^7.2.1",
|
|
42
43
|
"picomatch": "^4.0.3",
|
|
43
44
|
"remark-frontmatter": "^5.0.0",
|
|
44
45
|
"remark-gfm": "^4.0.0",
|
|
@@ -7,21 +7,31 @@
|
|
|
7
7
|
* frontmatter-specific type codes plus a list of external URLs the registry
|
|
8
8
|
* can fold into its existing external URL collection.
|
|
9
9
|
*
|
|
10
|
-
*
|
|
11
|
-
*
|
|
12
|
-
*
|
|
13
|
-
*
|
|
14
|
-
*
|
|
10
|
+
* Code mapping:
|
|
11
|
+
* LINK_BROKEN_FILE -> FRONTMATTER_LINK_BROKEN
|
|
12
|
+
* LINK_BROKEN_ANCHOR -> FRONTMATTER_ANCHOR_MISSING
|
|
13
|
+
* LINK_TO_GITIGNORED -> FRONTMATTER_LINK_TO_GITIGNORED
|
|
14
|
+
* LINK_UNKNOWN -> FRONTMATTER_UNKNOWN_LINK
|
|
15
15
|
*
|
|
16
16
|
* Skipped (no issue, no external):
|
|
17
17
|
* email (mailto:)
|
|
18
18
|
* anchor-only (validated as anchor in current file via validateLink)
|
|
19
19
|
*/
|
|
20
20
|
|
|
21
|
+
import { createRegistryIssue, type IssueCode } from '@vibe-agent-toolkit/agent-schema';
|
|
22
|
+
|
|
21
23
|
import { classifyLink } from './link-parser.js';
|
|
22
|
-
import { validateLink, type ValidateLinkOptions } from './link-validator.js';
|
|
24
|
+
import { validateLink, type FragmentIndex, type ValidateLinkOptions } from './link-validator.js';
|
|
23
25
|
import { walkFrontmatterUriReferences } from './schema-uri-walker.js';
|
|
24
|
-
import type {
|
|
26
|
+
import type { ResourceLink, ValidationIssue } from './types.js';
|
|
27
|
+
|
|
28
|
+
/** Map the link-level code emitted by validateLink to its frontmatter-scoped code. */
|
|
29
|
+
const LINK_CODE_TO_FRONTMATTER_CODE: Partial<Record<IssueCode, IssueCode>> = {
|
|
30
|
+
LINK_BROKEN_FILE: 'FRONTMATTER_LINK_BROKEN',
|
|
31
|
+
LINK_BROKEN_ANCHOR: 'FRONTMATTER_ANCHOR_MISSING',
|
|
32
|
+
LINK_TO_GITIGNORED: 'FRONTMATTER_LINK_TO_GITIGNORED',
|
|
33
|
+
LINK_UNKNOWN: 'FRONTMATTER_UNKNOWN_LINK',
|
|
34
|
+
};
|
|
25
35
|
|
|
26
36
|
/** A frontmatter-sourced external URL captured for downstream health checking. */
|
|
27
37
|
export interface FrontmatterExternalUrl {
|
|
@@ -41,14 +51,14 @@ export interface FrontmatterLinkValidationResult {
|
|
|
41
51
|
* @param frontmatter - Parsed frontmatter (or undefined)
|
|
42
52
|
* @param schema - JSON Schema for the collection
|
|
43
53
|
* @param sourceFilePath - Absolute path to the source file
|
|
44
|
-
* @param
|
|
54
|
+
* @param fragmentsByFile - Fragment index (file path → set of valid fragments) for anchor validation
|
|
45
55
|
* @param options - Same shape as validateLink (projectRoot, gitTracker, ...)
|
|
46
56
|
*/
|
|
47
57
|
export async function validateFrontmatterLinks(
|
|
48
58
|
frontmatter: Record<string, unknown> | undefined,
|
|
49
59
|
schema: object,
|
|
50
60
|
sourceFilePath: string,
|
|
51
|
-
|
|
61
|
+
fragmentsByFile: FragmentIndex,
|
|
52
62
|
options?: ValidateLinkOptions,
|
|
53
63
|
): Promise<FrontmatterLinkValidationResult> {
|
|
54
64
|
if (!frontmatter) return { issues: [], externalUrls: [] };
|
|
@@ -79,7 +89,7 @@ export async function validateFrontmatterLinks(
|
|
|
79
89
|
line: 1, // Frontmatter per-field line numbers are post-v1.
|
|
80
90
|
};
|
|
81
91
|
|
|
82
|
-
const issue = await validateLink(syntheticLink, sourceFilePath,
|
|
92
|
+
const issue = await validateLink(syntheticLink, sourceFilePath, fragmentsByFile, options);
|
|
83
93
|
if (!issue) continue;
|
|
84
94
|
|
|
85
95
|
issues.push(rewriteIssue(issue, capture.dottedPath));
|
|
@@ -89,24 +99,12 @@ export async function validateFrontmatterLinks(
|
|
|
89
99
|
}
|
|
90
100
|
|
|
91
101
|
function rewriteIssue(issue: ValidationIssue, dottedPath: string): ValidationIssue {
|
|
92
|
-
|
|
93
|
-
|
|
94
|
-
|
|
95
|
-
|
|
96
|
-
|
|
97
|
-
|
|
98
|
-
|
|
99
|
-
|
|
100
|
-
switch (originalType) {
|
|
101
|
-
case 'broken_file':
|
|
102
|
-
return 'frontmatter_link_broken';
|
|
103
|
-
case 'broken_anchor':
|
|
104
|
-
return 'frontmatter_anchor_missing';
|
|
105
|
-
case 'link_to_gitignored':
|
|
106
|
-
return 'frontmatter_link_to_gitignored';
|
|
107
|
-
case 'unknown_link':
|
|
108
|
-
return 'frontmatter_unknown_link';
|
|
109
|
-
default:
|
|
110
|
-
return originalType;
|
|
111
|
-
}
|
|
102
|
+
const mappedCode = LINK_CODE_TO_FRONTMATTER_CODE[issue.code as IssueCode] ?? (issue.code as IssueCode);
|
|
103
|
+
const message = `field \`${dottedPath}\`: ${issue.message}`;
|
|
104
|
+
const extras: Partial<Pick<ValidationIssue, 'location' | 'line' | 'link' | 'suggestion'>> = {};
|
|
105
|
+
if (issue.location !== undefined) extras.location = issue.location;
|
|
106
|
+
if (issue.line !== undefined) extras.line = issue.line;
|
|
107
|
+
if (issue.link !== undefined) extras.link = issue.link;
|
|
108
|
+
if (issue.suggestion !== undefined) extras.suggestion = issue.suggestion;
|
|
109
|
+
return createRegistryIssue(mappedCode, message, extras);
|
|
112
110
|
}
|
|
@@ -13,9 +13,11 @@
|
|
|
13
13
|
* This is the ONLY place in the codebase that should use AJV.
|
|
14
14
|
*/
|
|
15
15
|
|
|
16
|
+
import { createRegistryIssue, type ValidationIssue } from '@vibe-agent-toolkit/agent-schema';
|
|
17
|
+
|
|
16
18
|
import { createAjvWithUriFormats } from './ajv-factory.js';
|
|
17
19
|
import type { ValidationMode } from './schemas/project-config.js';
|
|
18
|
-
import
|
|
20
|
+
import { issueLocation } from './utils.js';
|
|
19
21
|
|
|
20
22
|
/**
|
|
21
23
|
* Validate frontmatter against a JSON Schema.
|
|
@@ -31,6 +33,7 @@ import type { ValidationIssue } from './schemas/validation-result.js';
|
|
|
31
33
|
* @param resourcePath - File path for error reporting
|
|
32
34
|
* @param mode - Validation mode: 'strict' (default) or 'permissive'
|
|
33
35
|
* @param schemaPath - Path to schema file (for error context)
|
|
36
|
+
* @param projectRoot - Project root for computing relative issue locations
|
|
34
37
|
* @returns Array of validation issues (empty if valid)
|
|
35
38
|
*
|
|
36
39
|
* @example
|
|
@@ -54,7 +57,8 @@ export function validateFrontmatter(
|
|
|
54
57
|
schema: object,
|
|
55
58
|
resourcePath: string,
|
|
56
59
|
mode: ValidationMode = 'strict',
|
|
57
|
-
schemaPath?: string
|
|
60
|
+
schemaPath?: string,
|
|
61
|
+
projectRoot?: string,
|
|
58
62
|
): ValidationIssue[] {
|
|
59
63
|
const issues: ValidationIssue[] = [];
|
|
60
64
|
|
|
@@ -87,13 +91,13 @@ export function validateFrontmatter(
|
|
|
87
91
|
const schemaContext = schemaPath ? ` (schema: ${schemaPath}, mode: ${mode})` : '';
|
|
88
92
|
const requiredFields = schemaRequires.join(', ');
|
|
89
93
|
|
|
90
|
-
issues.push(
|
|
91
|
-
|
|
92
|
-
|
|
93
|
-
|
|
94
|
-
|
|
95
|
-
|
|
96
|
-
|
|
94
|
+
issues.push(
|
|
95
|
+
createRegistryIssue(
|
|
96
|
+
'FRONTMATTER_MISSING',
|
|
97
|
+
`No frontmatter found in file. Schema requires: ${requiredFields}${schemaContext}`,
|
|
98
|
+
{ location: issueLocation(resourcePath, projectRoot), line: 1 },
|
|
99
|
+
),
|
|
100
|
+
);
|
|
97
101
|
}
|
|
98
102
|
return issues;
|
|
99
103
|
}
|
|
@@ -108,13 +112,12 @@ export function validateFrontmatter(
|
|
|
108
112
|
// Format validation errors with helpful messages
|
|
109
113
|
for (const error of validate.errors) {
|
|
110
114
|
const message = formatValidationError(error, frontmatter, mode, schemaPath);
|
|
111
|
-
issues.push(
|
|
112
|
-
|
|
113
|
-
|
|
114
|
-
|
|
115
|
-
|
|
116
|
-
|
|
117
|
-
});
|
|
115
|
+
issues.push(
|
|
116
|
+
createRegistryIssue('FRONTMATTER_SCHEMA_ERROR', message, {
|
|
117
|
+
location: issueLocation(resourcePath, projectRoot),
|
|
118
|
+
line: 1,
|
|
119
|
+
}),
|
|
120
|
+
);
|
|
118
121
|
}
|
|
119
122
|
|
|
120
123
|
return issues;
|
|
@@ -0,0 +1,140 @@
|
|
|
1
|
+
/**
|
|
2
|
+
* HTML resource parser.
|
|
3
|
+
*
|
|
4
|
+
* Parses local HTML files into the shared `ParseResult` shape:
|
|
5
|
+
* - `<a href>` and `<img src>` links (classified by `classifyLink`)
|
|
6
|
+
* - `id` / `name` attributes as fragment anchors
|
|
7
|
+
* - well-formedness diagnostics from parse5's `onParseError`
|
|
8
|
+
*
|
|
9
|
+
* Uses parse5 (WHATWG-conformant). The parse5 document + element walker are
|
|
10
|
+
* exported so the link rewriter (html-transform.ts) shares one parser path.
|
|
11
|
+
*
|
|
12
|
+
* Non-goal: `<base href>` is not honored — relative links are resolved against
|
|
13
|
+
* the file's own directory, not a document base.
|
|
14
|
+
*/
|
|
15
|
+
|
|
16
|
+
import { readFile, stat } from 'node:fs/promises';
|
|
17
|
+
|
|
18
|
+
import { parse, type DefaultTreeAdapterMap } from 'parse5';
|
|
19
|
+
|
|
20
|
+
import { classifyLink, type ParseResult } from './link-parser.js';
|
|
21
|
+
import type { HtmlParseError } from './schemas/resource-metadata.js';
|
|
22
|
+
import type { ResourceLink } from './types.js';
|
|
23
|
+
|
|
24
|
+
type P5Node = DefaultTreeAdapterMap['node'];
|
|
25
|
+
type P5Element = DefaultTreeAdapterMap['element'];
|
|
26
|
+
type P5ChildNode = DefaultTreeAdapterMap['childNode'];
|
|
27
|
+
|
|
28
|
+
/** A parsed parse5 document together with any well-formedness diagnostics. */
|
|
29
|
+
export interface HtmlDocument {
|
|
30
|
+
document: DefaultTreeAdapterMap['document'];
|
|
31
|
+
parseErrors: HtmlParseError[];
|
|
32
|
+
}
|
|
33
|
+
|
|
34
|
+
/**
|
|
35
|
+
* Parse an HTML source string with source-location info and collect parser
|
|
36
|
+
* errors. Shared by `parseHtml` (link/anchor extraction) and `rewriteHtmlLinks`
|
|
37
|
+
* (attribute offset splicing).
|
|
38
|
+
*/
|
|
39
|
+
export function parseHtmlDocument(source: string): HtmlDocument {
|
|
40
|
+
const parseErrors: HtmlParseError[] = [];
|
|
41
|
+
const document = parse(source, {
|
|
42
|
+
sourceCodeLocationInfo: true,
|
|
43
|
+
onParseError: (err) => {
|
|
44
|
+
// parse5's ParserError extends Location which always has startLine: number.
|
|
45
|
+
// We always include line since it is always present.
|
|
46
|
+
parseErrors.push({ message: err.code, line: err.startLine });
|
|
47
|
+
},
|
|
48
|
+
});
|
|
49
|
+
return { document, parseErrors };
|
|
50
|
+
}
|
|
51
|
+
|
|
52
|
+
/**
|
|
53
|
+
* Depth-first walk yielding every element node in the tree.
|
|
54
|
+
*
|
|
55
|
+
* Deliberate gap: a `<template>` element's content lives in its separate
|
|
56
|
+
* `content` fragment (not `childNodes`), so links inside `<template>` are not
|
|
57
|
+
* walked, and foreign-content (SVG/MathML) subtrees are not special-cased.
|
|
58
|
+
* Both are rare in the content link graph we rewrite.
|
|
59
|
+
*/
|
|
60
|
+
export function* walkElements(node: P5Node): Generator<P5Element> {
|
|
61
|
+
if ('tagName' in node) {
|
|
62
|
+
yield node;
|
|
63
|
+
}
|
|
64
|
+
if ('childNodes' in node) {
|
|
65
|
+
for (const child of node.childNodes as P5ChildNode[]) {
|
|
66
|
+
yield* walkElements(child);
|
|
67
|
+
}
|
|
68
|
+
}
|
|
69
|
+
}
|
|
70
|
+
|
|
71
|
+
function getAttr(element: P5Element, name: string): string | undefined {
|
|
72
|
+
return element.attrs.find((a) => a.name === name)?.value;
|
|
73
|
+
}
|
|
74
|
+
|
|
75
|
+
function makeLink(href: string, line: number | undefined): ResourceLink {
|
|
76
|
+
return { text: '', href, type: classifyLink(href), ...(line !== undefined && { line }) };
|
|
77
|
+
}
|
|
78
|
+
|
|
79
|
+
function visitElement(
|
|
80
|
+
element: P5Element,
|
|
81
|
+
links: ResourceLink[],
|
|
82
|
+
anchors: Set<string>,
|
|
83
|
+
): void {
|
|
84
|
+
const line = element.sourceCodeLocation?.startLine;
|
|
85
|
+
|
|
86
|
+
if (element.tagName === 'a') {
|
|
87
|
+
const href = getAttr(element, 'href');
|
|
88
|
+
if (href !== undefined) {
|
|
89
|
+
links.push(makeLink(href, line));
|
|
90
|
+
}
|
|
91
|
+
const name = getAttr(element, 'name');
|
|
92
|
+
if (name !== undefined && name !== '') {
|
|
93
|
+
anchors.add(name);
|
|
94
|
+
}
|
|
95
|
+
} else if (element.tagName === 'img') {
|
|
96
|
+
const src = getAttr(element, 'src');
|
|
97
|
+
if (src !== undefined) {
|
|
98
|
+
links.push(makeLink(src, line));
|
|
99
|
+
}
|
|
100
|
+
}
|
|
101
|
+
|
|
102
|
+
const id = getAttr(element, 'id');
|
|
103
|
+
if (id !== undefined && id !== '') {
|
|
104
|
+
anchors.add(id);
|
|
105
|
+
}
|
|
106
|
+
}
|
|
107
|
+
|
|
108
|
+
/**
|
|
109
|
+
* Parse an HTML file into a `ParseResult`.
|
|
110
|
+
*
|
|
111
|
+
* @param filePath - Absolute path to the HTML file.
|
|
112
|
+
*/
|
|
113
|
+
export async function parseHtml(filePath: string): Promise<ParseResult> {
|
|
114
|
+
const [content, stats] = await Promise.all([
|
|
115
|
+
// eslint-disable-next-line security/detect-non-literal-fs-filename -- filePath is a user-provided path parameter
|
|
116
|
+
readFile(filePath, 'utf-8'),
|
|
117
|
+
// eslint-disable-next-line security/detect-non-literal-fs-filename -- filePath is a user-provided path parameter
|
|
118
|
+
stat(filePath),
|
|
119
|
+
]);
|
|
120
|
+
|
|
121
|
+
const { document, parseErrors } = parseHtmlDocument(content);
|
|
122
|
+
|
|
123
|
+
const links: ResourceLink[] = [];
|
|
124
|
+
const anchors = new Set<string>();
|
|
125
|
+
|
|
126
|
+
for (const element of walkElements(document)) {
|
|
127
|
+
visitElement(element, links, anchors);
|
|
128
|
+
}
|
|
129
|
+
|
|
130
|
+
const anchorList = [...anchors];
|
|
131
|
+
return {
|
|
132
|
+
links,
|
|
133
|
+
headings: [],
|
|
134
|
+
content,
|
|
135
|
+
sizeBytes: stats.size,
|
|
136
|
+
estimatedTokenCount: Math.ceil(content.length / 4),
|
|
137
|
+
...(anchorList.length > 0 && { anchors: anchorList }),
|
|
138
|
+
...(parseErrors.length > 0 && { parseErrors }),
|
|
139
|
+
};
|
|
140
|
+
}
|
|
@@ -0,0 +1,157 @@
|
|
|
1
|
+
/**
|
|
2
|
+
* Structure-preserving HTML link rewriter.
|
|
3
|
+
*
|
|
4
|
+
* Rewrites `<a href>` and `<img src>` attribute VALUES in place by splicing the
|
|
5
|
+
* original source string at parse5-reported offsets. The document is never
|
|
6
|
+
* re-serialized (parse5's serializer normalizes whitespace, quotes, void
|
|
7
|
+
* elements, and the doctype), so unchanged input round-trips byte-for-byte.
|
|
8
|
+
*
|
|
9
|
+
* Uses the same `RewriteHref` callback model as `rewriteBodyLinks` — callers
|
|
10
|
+
* supply per-href target resolution; this module owns only the splice mechanics.
|
|
11
|
+
*
|
|
12
|
+
* Non-goal: `<base href>` is not honored. Relative hrefs are resolved by the
|
|
13
|
+
* caller against the file's own directory; a `<base>` element that would
|
|
14
|
+
* override that in a browser is ignored.
|
|
15
|
+
*/
|
|
16
|
+
|
|
17
|
+
import { parseHtmlDocument, walkElements } from './html-link-parser.js';
|
|
18
|
+
import type { RewriteHref } from './rewriter-helpers.js';
|
|
19
|
+
|
|
20
|
+
/** Tag name → the single link-bearing attribute we rewrite. */
|
|
21
|
+
const LINK_ATTR_BY_TAG: Record<string, string | undefined> = { a: 'href', img: 'src' };
|
|
22
|
+
|
|
23
|
+
interface ValueSpan {
|
|
24
|
+
valueStart: number;
|
|
25
|
+
valueEnd: number;
|
|
26
|
+
/** `"` or `'` for quoted attributes; `''` for unquoted. */
|
|
27
|
+
quote: string;
|
|
28
|
+
}
|
|
29
|
+
|
|
30
|
+
interface Edit extends ValueSpan {
|
|
31
|
+
newValue: string;
|
|
32
|
+
}
|
|
33
|
+
|
|
34
|
+
/**
|
|
35
|
+
* Chars that force an unquoted HTML attribute value to be quoted. Follows the
|
|
36
|
+
* WHATWG unquoted-attribute-value rules: whitespace, quotes, `<`, `>`, plus
|
|
37
|
+
* backtick and `=` (a superset of paths we expect, kept strict for safety).
|
|
38
|
+
*/
|
|
39
|
+
const UNQUOTED_UNSAFE = /[\s"'`<>=]/;
|
|
40
|
+
|
|
41
|
+
/**
|
|
42
|
+
* Locate the value sub-range within an attribute's full source span.
|
|
43
|
+
*
|
|
44
|
+
* parse5 reports the whole-attribute span (`href="value"`); this finds the
|
|
45
|
+
* value's absolute offsets and the quote char. Returns undefined for boolean
|
|
46
|
+
* attributes (no `=`).
|
|
47
|
+
*/
|
|
48
|
+
function valueSpan(attrSource: string, base: number): ValueSpan | undefined {
|
|
49
|
+
const eq = attrSource.indexOf('=');
|
|
50
|
+
if (eq === -1) {
|
|
51
|
+
return undefined;
|
|
52
|
+
}
|
|
53
|
+
let i = eq + 1;
|
|
54
|
+
while (i < attrSource.length && /\s/.test(attrSource.charAt(i))) {
|
|
55
|
+
i += 1;
|
|
56
|
+
}
|
|
57
|
+
if (i >= attrSource.length) {
|
|
58
|
+
return undefined;
|
|
59
|
+
}
|
|
60
|
+
const ch = attrSource.charAt(i);
|
|
61
|
+
if (ch === '"' || ch === "'") {
|
|
62
|
+
const close = attrSource.indexOf(ch, i + 1);
|
|
63
|
+
const end = close === -1 ? attrSource.length : close;
|
|
64
|
+
return { valueStart: base + i + 1, valueEnd: base + end, quote: ch };
|
|
65
|
+
}
|
|
66
|
+
return { valueStart: base + i, valueEnd: base + attrSource.length, quote: '' };
|
|
67
|
+
}
|
|
68
|
+
|
|
69
|
+
/**
|
|
70
|
+
* Encode a new value for writing. For quoted attributes the surrounding quotes
|
|
71
|
+
* stay in the source (we only replace the inner value), so we escape `&` and the
|
|
72
|
+
* active quote. For originally-unquoted values we keep them bare when safe, else
|
|
73
|
+
* wrap in double quotes.
|
|
74
|
+
*/
|
|
75
|
+
function encodeValue(newValue: string, quote: string): string {
|
|
76
|
+
if (quote === '"' || quote === "'") {
|
|
77
|
+
const amp = newValue.replaceAll('&', '&');
|
|
78
|
+
return quote === '"' ? amp.replaceAll('"', '"') : amp.replaceAll("'", ''');
|
|
79
|
+
}
|
|
80
|
+
if (newValue.length > 0 && !UNQUOTED_UNSAFE.test(newValue)) {
|
|
81
|
+
return newValue.replaceAll('&', '&');
|
|
82
|
+
}
|
|
83
|
+
const escaped = newValue.replaceAll('&', '&').replaceAll('"', '"');
|
|
84
|
+
return `"${escaped}"`;
|
|
85
|
+
}
|
|
86
|
+
|
|
87
|
+
/** A wanted link rewrite that could not be spliced back into the source. */
|
|
88
|
+
export interface UnappliedRewrite {
|
|
89
|
+
tagName: string;
|
|
90
|
+
attr: string;
|
|
91
|
+
/** The original attribute value the rewrite targeted. */
|
|
92
|
+
from: string;
|
|
93
|
+
/** The value the rewrite wanted to write. */
|
|
94
|
+
to: string;
|
|
95
|
+
/**
|
|
96
|
+
* Why the splice could not be located:
|
|
97
|
+
* - `no-source-location` — parse5 omitted the attribute's offsets (synthesized
|
|
98
|
+
* during error recovery on malformed input).
|
|
99
|
+
* - `unparseable-attribute` — the attribute's source span had no locatable
|
|
100
|
+
* value (e.g. an unterminated quote).
|
|
101
|
+
*/
|
|
102
|
+
reason: 'no-source-location' | 'unparseable-attribute';
|
|
103
|
+
}
|
|
104
|
+
|
|
105
|
+
/**
|
|
106
|
+
* Rewrite `<a href>` / `<img src>` values in `source` using `rewriteHref`.
|
|
107
|
+
* Returns `source` unchanged (byte-for-byte) when no value changes.
|
|
108
|
+
*
|
|
109
|
+
* A rewrite that is wanted (the href resolves to a new value) but cannot be
|
|
110
|
+
* spliced back — because parse5 omitted the source location or the attribute
|
|
111
|
+
* span is unparseable — is reported via `onUnapplied` rather than dropped
|
|
112
|
+
* silently. Malformed pages are exactly the ones that hit this path, so the
|
|
113
|
+
* caller can surface it instead of shipping a stale link.
|
|
114
|
+
*/
|
|
115
|
+
export function rewriteHtmlLinks(
|
|
116
|
+
source: string,
|
|
117
|
+
rewriteHref: RewriteHref,
|
|
118
|
+
onUnapplied?: (info: UnappliedRewrite) => void,
|
|
119
|
+
): string {
|
|
120
|
+
const { document } = parseHtmlDocument(source);
|
|
121
|
+
const edits: Edit[] = [];
|
|
122
|
+
|
|
123
|
+
for (const element of walkElements(document)) {
|
|
124
|
+
const attrName = LINK_ATTR_BY_TAG[element.tagName];
|
|
125
|
+
if (attrName === undefined) {
|
|
126
|
+
continue;
|
|
127
|
+
}
|
|
128
|
+
const attr = element.attrs.find((a) => a.name === attrName);
|
|
129
|
+
if (attr === undefined) {
|
|
130
|
+
continue;
|
|
131
|
+
}
|
|
132
|
+
// Resolve first so we only record drops that actually lose a wanted edit.
|
|
133
|
+
const newValue = rewriteHref(attr.value);
|
|
134
|
+
if (newValue === attr.value) {
|
|
135
|
+
continue;
|
|
136
|
+
}
|
|
137
|
+
const location = element.sourceCodeLocation?.attrs?.[attrName];
|
|
138
|
+
if (location === undefined) {
|
|
139
|
+
onUnapplied?.({ tagName: element.tagName, attr: attrName, from: attr.value, to: newValue, reason: 'no-source-location' });
|
|
140
|
+
continue;
|
|
141
|
+
}
|
|
142
|
+
const span = valueSpan(source.slice(location.startOffset, location.endOffset), location.startOffset);
|
|
143
|
+
if (span === undefined) {
|
|
144
|
+
onUnapplied?.({ tagName: element.tagName, attr: attrName, from: attr.value, to: newValue, reason: 'unparseable-attribute' });
|
|
145
|
+
continue;
|
|
146
|
+
}
|
|
147
|
+
edits.push({ ...span, newValue });
|
|
148
|
+
}
|
|
149
|
+
|
|
150
|
+
// Apply descending by start offset so earlier edits don't shift later ones.
|
|
151
|
+
edits.sort((a, b) => b.valueStart - a.valueStart);
|
|
152
|
+
let result = source;
|
|
153
|
+
for (const edit of edits) {
|
|
154
|
+
result = result.slice(0, edit.valueStart) + encodeValue(edit.newValue, edit.quote) + result.slice(edit.valueEnd);
|
|
155
|
+
}
|
|
156
|
+
return result;
|
|
157
|
+
}
|
package/src/index.ts
CHANGED
|
@@ -83,7 +83,12 @@ export {
|
|
|
83
83
|
} from './schemas/validation-result.js';
|
|
84
84
|
|
|
85
85
|
// Export parser interface for advanced use cases
|
|
86
|
-
export { parseMarkdown, type ParseResult } from './link-parser.js';
|
|
86
|
+
export { parseMarkdown, classifyLink, isLocalFileLink, type ParseResult } from './link-parser.js';
|
|
87
|
+
|
|
88
|
+
export { parseHtml } from './html-link-parser.js';
|
|
89
|
+
// HtmlParseError is Zod-sourced (single source of truth) — see schemas/resource-metadata.ts.
|
|
90
|
+
export type { HtmlParseError } from './schemas/resource-metadata.js';
|
|
91
|
+
export { rewriteHtmlLinks, type UnappliedRewrite } from './html-transform.js';
|
|
87
92
|
|
|
88
93
|
// Export frontmatter validation
|
|
89
94
|
export { validateFrontmatter } from './frontmatter-validator.js';
|
package/src/link-parser.ts
CHANGED
|
@@ -1,5 +1,5 @@
|
|
|
1
1
|
/**
|
|
2
|
-
* Markdown link parser and
|
|
2
|
+
* Markdown link parser and shared resource-parsing types.
|
|
3
3
|
*
|
|
4
4
|
* Parses markdown files to extract:
|
|
5
5
|
* - Links (regular, reference-style, autolinks)
|
|
@@ -7,6 +7,10 @@
|
|
|
7
7
|
* - File size and token estimates
|
|
8
8
|
*
|
|
9
9
|
* Uses unified/remark for robust markdown parsing with GFM support.
|
|
10
|
+
*
|
|
11
|
+
* Also defines the format-neutral `ParseResult` contract shared with the HTML
|
|
12
|
+
* parser (`html-link-parser.ts`). The `HtmlParseError` shape is Zod-sourced
|
|
13
|
+
* from `schemas/resource-metadata.ts` (single source of truth).
|
|
10
14
|
*/
|
|
11
15
|
|
|
12
16
|
import { readFile, stat } from 'node:fs/promises';
|
|
@@ -20,10 +24,11 @@ import { unified } from 'unified';
|
|
|
20
24
|
import { visit } from 'unist-util-visit';
|
|
21
25
|
import * as yaml from 'yaml';
|
|
22
26
|
|
|
27
|
+
import type { HtmlParseError } from './schemas/resource-metadata.js';
|
|
23
28
|
import type { HeadingNode, LinkType, ResourceLink } from './types.js';
|
|
24
29
|
|
|
25
30
|
/**
|
|
26
|
-
* Result of parsing a markdown
|
|
31
|
+
* Result of parsing a resource file (markdown or HTML).
|
|
27
32
|
*/
|
|
28
33
|
export interface ParseResult {
|
|
29
34
|
links: ResourceLink[];
|
|
@@ -33,6 +38,10 @@ export interface ParseResult {
|
|
|
33
38
|
content: string;
|
|
34
39
|
sizeBytes: number;
|
|
35
40
|
estimatedTokenCount: number;
|
|
41
|
+
/** Fragment targets (HTML `id`/`name` attributes). Markdown leaves this undefined. */
|
|
42
|
+
anchors?: string[];
|
|
43
|
+
/** HTML well-formedness diagnostics. Markdown leaves this undefined. */
|
|
44
|
+
parseErrors?: HtmlParseError[];
|
|
36
45
|
}
|
|
37
46
|
|
|
38
47
|
/**
|
|
@@ -185,6 +194,11 @@ function extractLinkText(node: Link | LinkReference): string {
|
|
|
185
194
|
* classifyLink('#heading') // 'anchor'
|
|
186
195
|
* classifyLink('./file.md') // 'local_file'
|
|
187
196
|
* classifyLink('./file.md#anchor') // 'local_file'
|
|
197
|
+
* classifyLink('docs/') // 'local_directory'
|
|
198
|
+
* classifyLink('./docs/') // 'local_directory'
|
|
199
|
+
* classifyLink('../docs/') // 'local_directory'
|
|
200
|
+
* classifyLink('/docs/') // 'local_directory'
|
|
201
|
+
* classifyLink('https://x.com/docs/') // 'external' (not a local ref)
|
|
188
202
|
* ```
|
|
189
203
|
*/
|
|
190
204
|
export function classifyLink(href: string): LinkType {
|
|
@@ -202,6 +216,12 @@ export function classifyLink(href: string): LinkType {
|
|
|
202
216
|
if (href.includes(':')) {
|
|
203
217
|
return 'unknown';
|
|
204
218
|
}
|
|
219
|
+
// Local directory: path component (before any # or ?) ends in '/'.
|
|
220
|
+
// Must come after all protocol guards so external URLs are never reclassified.
|
|
221
|
+
const pathPart = href.split(/[#?]/u)[0] ?? href;
|
|
222
|
+
if (pathPart.endsWith('/')) {
|
|
223
|
+
return 'local_directory';
|
|
224
|
+
}
|
|
205
225
|
// Links with anchors are still local file links
|
|
206
226
|
if (href.includes('#')) {
|
|
207
227
|
return 'local_file';
|
|
@@ -242,6 +262,19 @@ export function classifyLink(href: string): LinkType {
|
|
|
242
262
|
return 'unknown';
|
|
243
263
|
}
|
|
244
264
|
|
|
265
|
+
/**
|
|
266
|
+
* Returns true for link types that represent local filesystem targets — both
|
|
267
|
+
* regular files and directories. Other packages (e.g. agent-skills walker)
|
|
268
|
+
* import this predicate as the single source of truth for "should we treat
|
|
269
|
+
* this link like a file link during validation/traversal?"
|
|
270
|
+
*
|
|
271
|
+
* @param type - The classified link type
|
|
272
|
+
* @returns `true` for `'local_file'` and `'local_directory'`
|
|
273
|
+
*/
|
|
274
|
+
export function isLocalFileLink(type: LinkType): boolean {
|
|
275
|
+
return type === 'local_file' || type === 'local_directory';
|
|
276
|
+
}
|
|
277
|
+
|
|
245
278
|
/**
|
|
246
279
|
* Extract headings from the markdown AST and build a nested tree structure.
|
|
247
280
|
*
|
|
@@ -310,29 +343,45 @@ function extractHeadingText(node: Heading): string {
|
|
|
310
343
|
return extractTextFromChildren(node.children);
|
|
311
344
|
}
|
|
312
345
|
|
|
346
|
+
/**
|
|
347
|
+
* Inline node shape for heading text extraction. Leaf inline nodes (`text`,
|
|
348
|
+
* `inlineCode`) carry their content in `value`; container inline nodes
|
|
349
|
+
* (`strong`, `emphasis`, `link`, `delete`) carry it in `children`.
|
|
350
|
+
*/
|
|
351
|
+
interface InlineNode {
|
|
352
|
+
type: string;
|
|
353
|
+
value?: unknown;
|
|
354
|
+
children?: InlineNode[];
|
|
355
|
+
}
|
|
356
|
+
|
|
313
357
|
/**
|
|
314
358
|
* Extract text content from inline children nodes.
|
|
315
359
|
*
|
|
316
|
-
* Handles
|
|
360
|
+
* Handles leaf nodes (`text`, `inlineCode`) and recurses into container inline
|
|
361
|
+
* elements (`strong`, `emphasis`, `link`, `delete`) so that styled headings —
|
|
362
|
+
* e.g. `### **CRITICAL: ...**` — produce the same text (and therefore the same
|
|
363
|
+
* GitHub slug) as their plain-text equivalents. Without recursion, bold/italic
|
|
364
|
+
* headings yielded empty text and bogus slugs, causing false LINK_BROKEN_ANCHOR
|
|
365
|
+
* errors for links targeting them.
|
|
317
366
|
*
|
|
318
367
|
* @param children - Array of child nodes or undefined
|
|
319
368
|
* @returns Concatenated text content
|
|
320
369
|
*/
|
|
321
|
-
function extractTextFromChildren(
|
|
322
|
-
children: Array<{ type: string; value?: unknown }> | undefined
|
|
323
|
-
): string {
|
|
370
|
+
function extractTextFromChildren(children: InlineNode[] | undefined): string {
|
|
324
371
|
if (!children || children.length === 0) {
|
|
325
372
|
return '';
|
|
326
373
|
}
|
|
327
374
|
|
|
328
375
|
return children
|
|
329
376
|
.map((child) => {
|
|
330
|
-
|
|
331
|
-
|
|
377
|
+
// Leaf inline nodes (text, inlineCode) hold their content in `value`.
|
|
378
|
+
if (typeof child.value === 'string') {
|
|
379
|
+
return child.value;
|
|
332
380
|
}
|
|
333
|
-
//
|
|
334
|
-
|
|
335
|
-
|
|
381
|
+
// Container inline nodes (strong, emphasis, link, delete) hold text in
|
|
382
|
+
// their children — recurse to gather it.
|
|
383
|
+
if (child.children) {
|
|
384
|
+
return extractTextFromChildren(child.children);
|
|
336
385
|
}
|
|
337
386
|
return '';
|
|
338
387
|
})
|