app/lib/pages/links.mjs

Functions

Imports

It imports app/lib/content-pages.mjs, app/lib/wordpress-redirects.mjs. Imported by app/lib/pages/save.server.ts.

Source

1// The internal links of a page, judged at save time. check:links is CI's full gate over every link the
2// site publishes, built from the repository's sources; a save through the operator API is live before CI
3// runs, so the part of that judgement a Worker can make from D1 and code is made here, with the same
4// gateway functions the router uses (docs/PAGES.md).
5//
6// It refuses only what it can PROVE broken: a link into a namespace this module can list completely
7// (the Research, Teaching, Software and recipe pages, the procedures, the posts) that names nothing in
8// it, a path the gateway answers 410, a draft post, and an anchor missing from a content page. Any other
9// address passes here and is left to check:links.
10
11import { CONTENT_PAGE_PATHS, contentPageMarkdownPath } from "../content-pages.mjs";
12import { wordpressDisposition } from "../wordpress-redirects.mjs";
13
14/**
15 * @typedef {{
16 *   procedurePaths: ReadonlySet<string>,
17 *   posts: ReadonlyMap<string, boolean>,
18 *   idsOf: (path: string) => ReadonlySet<string> | null,
19 * }} AddressBook
20 * `posts` maps `/writing/<slug>` to whether a reader can open it; `idsOf` is the element ids a content
21 * page carries, or null for a path that is not one.
22 */
23
24const decodeEntities = (/** @type {string} */ value) =>
25  value.replace(/&amp;/g, "&").replace(/&quot;/g, '"').replace(/&#x27;|&#39;/g, "'").replace(/&lt;/g, "<").replace(/&gt;/g, ">");
26
27const safeDecode = (/** @type {string} */ value) => {
28  try {
29    return decodeURIComponent(value);
30  } catch {
31    return value;
32  }
33};
34
35const SITE_HOST = /^https?:\/\/(?:www\.)?dustinedwards\.info(?=[/?#]|$)/i;
36
37/** Every element id in rendered HTML. */
38export function idsIn(/** @type {string} */ html) {
39  return new Set([...html.matchAll(/\sid="([^"]+)"/g)].map((m) => decodeEntities(m[1] ?? "")));
40}
41
42const PAGE_SET = new Set(/** @type {readonly string[]} */ (CONTENT_PAGE_PATHS));
43const PAGE_TWINS = new Set(CONTENT_PAGE_PATHS.map((p) => contentPageMarkdownPath(p)));
44
45/** The namespaces this module can list completely. */
46const OWNED = /^\/(?:research|teaching|software|recipes|writing)(?:\/|$)/;
47/** Listed elsewhere: papers and citation exports, feeds, tag and series pages. */
48const NOT_LISTED = /^\/(?:research\/publications|writing\/(?:tags|series|rss\.xml|feed\.json|atom\.xml))(?:[/.]|$)/;
49
50/**
51 * @param {string} path
52 * @param {AddressBook} book
53 * @param {number} [hops]
54 * @returns {{ kind: "ok" | "unlisted" | "gone" | "draft" | "missing", path: string }}
55 */
56function resolve(path, book, hops = 0) {
57  if (hops > 5) return { kind: "missing", path };
58  const redirect = wordpressDisposition(path);
59  if (redirect?.status === 410) return { kind: "gone", path };
60  if (redirect?.status === 301) return resolve(redirect.location.split("#")[0] ?? "", book, hops + 1);
61  if (!OWNED.test(path) || NOT_LISTED.test(path)) return { kind: "unlisted", path };
62  if (PAGE_SET.has(path) || PAGE_TWINS.has(path)) return { kind: "ok", path };
63  const procedure = path.replace(/(?:\.md|\/sheet)$/, "");
64  if (book.procedurePaths.has(procedure)) return { kind: "ok", path };
65  const visible = book.posts.get(path.replace(/\.md$/, ""));
66  if (visible !== undefined) return { kind: visible ? "ok" : "draft", path };
67  return { kind: "missing", path };
68}
69
70/**
71 * @param {{ html: string }} page
72 * @param {AddressBook} book
73 * @returns {string[]}
74 */
75export function pageLinkErrors({ html }, book) {
76  /** @type {string[]} */
77  const errors = [];
78  const seen = new Set();
79  for (const raw of [...html.matchAll(/\shref="([^"]*)"/g)].map((m) => decodeEntities(m[1] ?? ""))) {
80    const href = SITE_HOST.test(raw) ? raw.replace(SITE_HOST, "") || "/" : raw;
81    if (!href.startsWith("/") || href.startsWith("//") || seen.has(href)) continue;
82    seen.add(href);
83    const [pathAndQuery = "", anchor] = href.split("#");
84    const path = (pathAndQuery.split("?")[0] ?? "").replace(/(.)\/+$/, "$1");
85    const found = resolve(path, book);
86    if (found.kind === "gone") errors.push(`the link ${href} goes to ${path}, which the site answers 410 Gone`);
87    else if (found.kind === "draft") errors.push(`the link ${href} goes to a draft or scheduled post, which every reader gets a 404 for`);
88    else if (found.kind === "missing") errors.push(`the link ${href} goes to ${found.path}, which this site does not serve`);
89    else if (found.kind === "ok" && anchor && found.path === path) {
90      const ids = book.idsOf(found.path);
91      if (ids && !ids.has(safeDecode(anchor))) errors.push(`the link ${href} names an anchor that ${found.path} does not have`);
92    }
93  }
94  return errors;
95}