📡 RSS Gallery

\n ${\n Object.entries(marketplaceProducts).map(([productKey, product]) => `\n

${escapeHtml(product.title)}

\n ${\n Object.entries(product.feeds).map(([feedKey, feed]) => `\n

\n \n ${escapeHtml(feed.title)}\n RSS\n \n \n ${escapeHtml(feed.title)}\n JSON\n \n

\n `).join(\"\")\n }\n

Luma places

\n ${\n lumaPlaces.map((place) => `\n

\n \n ${escapeHtml(place.name)}\n RSS\n \n \n ${escapeHtml(place.name)}\n JSON\n \n

\n `).join(\"\")\n }\n

Telegram channels

\n ${\n Object.entries(telegramChannels).map(([name, channel]) => `\n

\n \n ${escapeHtml(channel.title)}\n RSS\n \n \n ${escapeHtml(channel.title)}\n JSON\n \n

\n `).join(\"\")\n }\n

Cooper Press (decomposed)

\n Every entry of the latest Cooper Press issue as an individual item,\n tagged with the section it appears under, any sponsorship and every\n single word it links to, closed by an item holding whatever the split of\n the issue did not recognize. Every item links back to the issue it was\n split from. Where the publication serves a page of its own instead of an\n issue, such as a block screen from a firewall in front of the site, the\n feed carries one item noting the failed fetch.\n

\n ${\n Object.entries(cooperPressPublications).map(([slug, publication]) =>\n staticFeedLinks(\n publication.title,\n `/cooperpress/${slug}.rss`,\n `/cooperpress/${slug}.json`,\n )\n ).join(\"\")\n }\n

HN Daily (decomposed)

\n Every article of the recent Hacker News Daily digests as an individual\n item, carrying the day it belongs to and the rank it held that day, and,\n like hnrss.org feeds, a link to the article and to its comments. A\n digest holds the ten highest-rated articles on Hacker News of one day\n which have not appeared on any previous Hacker News Daily, and the day\n of each item links back to the digest it was split from.\n

\n ${staticFeedLinks(\"HN Daily\", \"/hn-daily.rss\", \"/hn-daily.json\")}\n

Wikipedia Current Events (decomposed)

\n Every report of the settled days of the current Wikipedia Current Events\n month as an individual item. A day of the portal is a tree — a bullet\n names the subject of a story, the bullets under it refine that story, and\n a bullet holding no bullet of its own is what happened — so an item\n carries the report as the page nests it, under the day and category it\n stands in, and is named by the story it belongs to. Days reach the feed\n only once 24 hours have passed for the news to settle, and an item is\n named by a hash of its own body rather than by its place in the day, so a\n report the editors revise long after its date reaches a reader as the\n revision it is. Whatever the split of a day did not recognize closes it\n as one last item, and a page that is no month of the portal reaches the\n feed as one item noting the failed fetch.\n

\n ${\n staticFeedLinks(\n \"Wikipedia Current Events\",\n \"/wikipedia-current-events.rss\",\n \"/wikipedia-current-events.json\",\n )\n }\n

This Week in GNOME (decomposed)

\n Every note of the latest This Week in GNOME issue as an individual item,\n tagged with the section it appears under, the project it reports on and\n every single word it links to, and named by the project heading above it\n or by the line it opens with. Every item links back to the issue it was\n split from, and whatever the split of the issue did not recognize —\n the words the issue itself opens and closes with, or a style TWIG adds\n tomorrow — closes it as one last item. Where the site serves a page of\n its own instead of an issue, such as a block screen from a firewall in\n front of it, the feed carries one item noting the failed fetch.\n

\n ${staticFeedLinks(\"This Week in GNOME\", \"/twig.rss\", \"/twig.json\")}\n

Other feeds

\n ${\n staticFeedLinks(\"OpenRouter models\", \"/openrouter.rss\", \"/openrouter.json\")\n }\n ${\n staticFeedLinks(\n \"Artificial Analysis changelog\",\n \"/artificialanalysis.rss\",\n \"/artificialanalysis.json\",\n )\n }\n ${\n staticFeedLinks(\n \"Hugging Face activity\",\n \"/huggingface/DJLougen.rss\",\n \"/huggingface/DJLougen.json\",\n )\n }\n ${\n staticFeedLinks(\n \"LineageOS Devices\",\n \"/lineageos.rss\",\n \"/lineageos.json\",\n )\n }\n ${\n staticFeedLinks(\n \"MorningSave New Arrivals\",\n \"/stores/morningsave.rss\",\n \"/stores/morningsave.json\",\n )\n }\n ${\n staticFeedLinks(\n \"SideDeal New Arrivals\",\n \"/stores/sidedeal.rss\",\n \"/stores/sidedeal.json\",\n )\n }\n ${\n staticFeedLinks(\n \"Rentry (table)\",\n \"/rentry/lmg-news-archive.rss\",\n \"/rentry/lmg-news-archive.json\",\n )\n }\n ${\n staticFeedLinks(\n \"Crowd Supply Prelaunch\",\n \"/crowdsupply/prelaunch.rss\",\n \"/crowdsupply/prelaunch.json\",\n )\n }\n ${\n staticFeedLinks(\n \"Crowd Supply Crowdfunding\",\n \"/crowdsupply/crowdfunding.rss\",\n \"/crowdsupply/crowdfunding.json\",\n )\n }\n ${\n staticFeedLinks(\n \"Rekt news\",\n \"/rekt.rss\",\n \"/rekt.json\",\n )\n }\n ${\n staticFeedLinks(\n \"SQLite release notes\",\n \"/sqlite.rss\",\n \"/sqlite.json\",\n )\n }\n

\n \n \n `);\n});\n\napp.get(\"/openrouter.rss\", openRouterHandler);\napp.get(\"/openrouter.json\", openRouterHandler);\n\napp.get(\"/artificialanalysis.rss\", artificialAnalysisHandler);\napp.get(\"/artificialanalysis.json\", artificialAnalysisHandler);\n\napp.get(\"/vs-marketplace/:product/:feedFile\", vsMarketplaceHandler);\n\napp.get(\"/huggingface/:usernameFile\", huggingFaceHandler);\n\napp.get(\"/lineageos.rss\", lineageOsHandler);\napp.get(\"/lineageos.json\", lineageOsHandler);\n\napp.get(\"/stores/:site\", storesComHandler);\n\napp.get(\"/cooperpress/:publication\", cooperPressHandler);\n\napp.get(\"/hn-daily.rss\", hnDailyHandler);\napp.get(\"/hn-daily.json\", hnDailyHandler);\n\napp.get(\"/wikipedia-current-events.rss\", wikipediaCurrentEventsHandler);\napp.get(\"/wikipedia-current-events.json\", wikipediaCurrentEventsHandler);\n\napp.get(\"/twig.rss\", twigHandler);\napp.get(\"/twig.json\", twigHandler);\n\napp.get(\"/crowdsupply/:page\", crowdSupplyHandler);\n\napp.get(\"/luma/:place\", lumaHandler);\n\napp.get(\"/telegram/:channel\", telegramHandler);\n\napp.get(\"/rentry/:url\", rentryHandler);\n\napp.get(\"/rekt.rss\", rektHandler);\napp.get(\"/rekt.json\", rektHandler);\n\napp.get(\"/sqlite.rss\", sqliteChronologyHandler);\napp.get(\"/sqlite.json\", sqliteChronologyHandler);\n\nfunction staticFeedLinks(title: string, rss: string, json: string): string {\n return `

\n \n ${escapeHtml(title)}\n RSS\n \n \n ${escapeHtml(title)}\n JSON\n \n

`;\n}\n\nfunction escapeHtml(value: string): string {\n return value\n .replace(/&/g, \"&\")\n .replace(//g, \">\")\n .replace(/\"/g, \""\");\n}\n\nexport default app.fetch;\n",[],[194],"74930ff7-5a98-4ec1-81db-b37dcde3440e","feed.ts","11d9fd38-586b-11f1-affc-ee650bb23af1",74,["D",1789538343076],4,"import type { Context, MiddlewareHandler } from \"npm:hono@4.12.16\";\nimport jsonata from \"npm:jsonata@2.1.0\";\nimport jsonfeedToRss from \"npm:jsonfeed-to-rss@3.0.8\";\n\nexport const CACHE_CONTROL = \"max-age=0, s-maxage=3600\";\nconst JSON_FEED_VERSION = \"https://jsonfeed.org/version/1\";\n\ninterface JsonFeedAuthor {\n name: string;\n}\n\nexport interface JsonFeedItem {\n id: string;\n url?: string;\n title: string;\n content_html?: string;\n content_text?: string;\n summary?: string;\n date_published?: string;\n date_modified?: string;\n author?: JsonFeedAuthor;\n tags?: string[];\n attachments?: Array<{ url: string }>;\n}\n\ninterface JsonFeed {\n version: string;\n title: string;\n home_page_url: string;\n feed_url?: string;\n description?: string;\n /** The feed's own icon, which becomes the RSS channel image */\n icon?: string;\n items: JsonFeedItem[];\n}\n\nexport interface FeedSource {\n data: unknown;\n bindings?: Record;\n}\n\ninterface FeedEndpointOptions {\n expression: string;\n loadSource: (c: Context) => Promise;\n}\n\nexport class HttpError extends Error {\n constructor(public status: number, message: string) {\n super(message);\n }\n}\n\nexport function feedEndpoint(options: FeedEndpointOptions): MiddlewareHandler {\n const expression = jsonata(options.expression);\n\n return async (c) => {\n try {\n const { data, bindings = {} } = await options.loadSource(c);\n const feed = await expression.evaluate(data, {\n ...bindings,\n feedUrl: c.req.url,\n jsonFeedVersion: JSON_FEED_VERSION,\n }) as JsonFeed;\n\n feed.items = feedItems(feed.items);\n\n const format = c.req.path.endsWith(\".json\") ? \"json\" : \"rss\";\n if (format === \"json\") {\n return new Response(JSON.stringify(feed, null, 2), {\n headers: {\n \"Content-Type\": \"application/feed+json; charset=utf-8\",\n \"Cache-Control\": CACHE_CONTROL,\n },\n });\n }\n if (format !== \"rss\") {\n throw new HttpError(404, \"Feed format not supported. Use rss or json.\");\n }\n\n return new Response(\n jsonfeedToRss(feed).replaceAll(\n /\\n\\s*Invalid Date<\\/pubDate>/g,\n \"\",\n ),\n {\n headers: {\n \"Content-Type\": \"application/rss+xml; charset=utf-8\",\n \"Cache-Control\": CACHE_CONTROL,\n },\n },\n );\n } catch (error) {\n if (error instanceof HttpError) {\n return new Response(error.message, {\n status: error.status,\n headers: {\n \"Content-Type\": \"text/plain; charset=utf-8\",\n \"Cache-Control\": CACHE_CONTROL,\n },\n });\n }\n throw error;\n }\n };\n}\n\nexport async function fetchJson(url: string): Promise {\n const response = await fetch(url, {\n headers: { \"Accept\": \"application/json\" },\n });\n if (!response.ok) {\n throw new Error(\n `Failed to fetch ${url}: ${response.status} ${response.statusText}`,\n );\n }\n return await response.json();\n}\n\n// The items an expression left for a feed, as the list both feed formats carry.\n// JSONata hands back a sequence of one as the value itself, so a feed holding a\n// single item arrives as that item rather than as a list holding it, and a\n// mapping that matched nothing arrives as nothing at all.\nexport function feedItems(\n items: JsonFeedItem[] | JsonFeedItem | undefined,\n): JsonFeedItem[] {\n if (items === undefined) return [];\n return Array.isArray(items) ? items : [items];\n}\n",[],[203],"1deefc9e-ec71-4a30-93b3-0b69e194f545","deno.json","09936eb6-4506-11f1-ab47-42b51c65c3df",60,["D",1789450241692],3,"{\n \"$schema\": \"https://raw.githubusercontent.com/denoland/deno/98f62cee78e85bfc47c62ed703777c6bc8794f1c/cli/schemas/config-file.v1.json\",\n \"lock\": false,\n \"compilerOptions\": {\n \"noImplicitAny\": false,\n \"strict\": true,\n \"types\": [\n \"https://www.val.town/types/valtown.d.ts\"\n ],\n \"lib\": [\n \"dom\",\n \"dom.iterable\",\n \"dom.asynciterable\",\n \"deno.ns\",\n \"deno.unstable\"\n ]\n },\n \"lint\": {\n \"rules\": {\n \"exclude\": [\n \"no-import-prefix\"\n ]\n }\n },\n \"node_modules_dir\": false,\n \"tasks\": {\n \"test\": \"deno test --allow-net --allow-read --allow-env --allow-import=www.val.town,registry.npmjs.org,jsr.io,esm.sh,esm.town\",\n \"serve\": \"deno serve --watch --port 8123 --allow-net --allow-read --allow-env --allow-import=www.val.town,registry.npmjs.org,jsr.io,esm.sh,esm.town serve.ts\"\n },\n \"experimental\": {\n \"unstable-node-globals\": true,\n \"unstable-temporal\": true,\n \"unstable-worker-options\": true,\n \"unstable-sloppy-imports\": true\n }\n}\n",[],[212],"11e31cba-586b-11f1-9538-ee650bb23af1","AGENTS.md","0985b316-4506-11f1-ac50-42b51c65c3df",13,["D",1779735230326],"You are an advanced assistant specialized in generating Val Town code.\n\n## Core Guidelines\n\n- Ask clarifying questions when requirements are ambiguous\n- Provide complete, functional solutions rather than skeleton implementations\n- Test your logic against edge cases before presenting the final solution\n- Ensure all code follows Val Town's specific platform requirements\n- If a section of code that you're working on is getting too complex, consider\n refactoring it into subcomponents\n\n## Code Standards\n\n- Generate code in TypeScript or TSX\n- Add appropriate TypeScript types and interfaces for all data structures\n- Prefer official SDKs or libraries than writing API calls directly\n- Ask the user to supply API or library documentation if you are at all unsure\n about it\n- **Never bake in secrets into the code** - always use environment variables\n- Include comments explaining complex logic (avoid commenting obvious\n operations)\n- Follow modern ES6+ conventions and functional programming practices if\n possible\n\n## Types of triggers\n\n### 1. HTTP Trigger\n\n- Create web APIs and endpoints\n- Handle HTTP requests and responses\n- Example structure:\n\n```ts\nexport default async function (req: Request) {\n return new Response(\"Hello World\");\n}\n```\n\nFiles that are HTTP triggers have http in their name like `foobar.http.tsx`\n\n### 2. Cron Triggers\n\n- Run on a schedule\n- Use cron expressions for timing\n- Example structure:\n\n```ts\nexport default async function () {\n // Scheduled task code\n}\n```\n\nFiles that are Cron triggers have cron in their name like `foobar.cron.tsx`\n\n### 3. Email Triggers\n\n- Process incoming emails\n- Handle email-based workflows\n- Example structure:\n\n```ts\nexport default async function (email: Email) {\n // Process email\n}\n```\n\nFiles that are Email triggers have email in their name like `foobar.email.tsx`\n\n## Val Town Standard Libraries\n\nVal Town provides several hosted services and utility functions.\n\n### Blob Storage\n\n```ts\nimport { blob } from \"https://esm.town/v/std/blob\";\nawait blob.setJSON(\"myKey\", { hello: \"world\" });\nlet blobDemo = await blob.getJSON(\"myKey\");\nlet appKeys = await blob.list(\"app_\");\nawait blob.delete(\"myKey\");\n```\n\n### SQLite\n\n```ts\nimport { sqlite } from \"https://esm.town/v/stevekrouse/sqlite\";\nconst TABLE_NAME = \"todo_app_users_2\";\n// Create table - do this before usage and change table name when modifying schema\nawait sqlite.execute(`CREATE TABLE IF NOT EXISTS ${TABLE_NAME} (\n id INTEGER PRIMARY KEY AUTOINCREMENT,\n name TEXT NOT NULL\n)`);\n// Query data\nconst result = await sqlite.execute(\n `SELECT * FROM ${TABLE_NAME} WHERE id = ?`,\n [1],\n);\n```\n\nNote: When changing a SQLite table's schema, change the table's name (e.g., add\n_2 or _3) to create a fresh table.\n\n### OpenAI\n\n```ts\nimport { OpenAI } from \"https://esm.town/v/std/openai\";\nconst openai = new OpenAI();\nconst completion = await openai.chat.completions.create({\n messages: [\n { role: \"user\", content: \"Say hello in a creative way\" },\n ],\n model: \"gpt-4o-mini\",\n max_tokens: 30,\n});\n```\n\n### Email\n\n```ts\nimport { email } from \"https://esm.town/v/std/email\";\n// By default emails the owner of the val\nawait email({\n subject: \"Hi\",\n text: \"Hi\",\n html: \"

Hi

\",\n});\n```\n\n## Val Town Utility Functions\n\nVal Town provides several utility functions to help with common project tasks.\n\n### Importing Utilities\n\nAlways import utilities with version pins to avoid breaking changes:\n\n```ts\nimport {\n parseProject,\n readFile,\n serveFile,\n} from \"https://esm.town/v/std/utils@85-main/index.ts\";\n```\n\n### Available Utilities\n\n#### **serveFile** - Serve project files with proper content types\n\nFor example, in Hono:\n\n```ts\n// serve all files in frontend/ and shared/\napp.get(\"/frontend/*\", (c) => serveFile(c.req.path, import.meta.url));\napp.get(\"/shared/*\", (c) => serveFile(c.req.path, import.meta.url));\n```\n\n#### **readFile** - Read files from within the project:\n\n```ts\n// Read a file from the project\nconst fileContent = await readFile(\"/frontend/index.html\", import.meta.url);\n```\n\n#### **listFiles** - List all files in the project\n\n```ts\nconst files = await listFiles(import.meta.url);\n```\n\n#### **parseProject** - Extract information about the current project from import.meta.url\n\nThis is useful for including info for linking back to a val, ie in \"view source\"\nurls:\n\n```ts\nconst projectVal = parseProject(import.meta.url);\nconsole.log(projectVal.username); // Owner of the project\nconsole.log(projectVal.name); // Project name\nconsole.log(projectVal.version); // Version number\nconsole.log(projectVal.branch); // Branch name\nconsole.log(projectVal.links.self.project); // URL to the project page\n```\n\nHowever, it's _extremely importing_ to note that `parseProject` and other\nStandard Library utilities ONLY RUN ON THE SERVER. If you need access to this\ndata on the client, run it in the server and pass it to the client by splicing\nit into the HTML page or by making an API request for it.\n\n## Val Town Platform Specifics\n\n- **Redirects:** Use\n `return new Response(null, { status: 302, headers: { Location: \"/place/to/redirect\" }})`\n instead of `Response.redirect` which is broken\n- **Images:** Avoid external images or base64 images. Use emojis, unicode\n symbols, or icon fonts/libraries instead\n- **AI Image:** To inline generate an AI image use:\n `

`\n- **Storage:** DO NOT use the Deno KV module for storage\n- **Browser APIs:** DO NOT use the `alert()`, `prompt()`, or `confirm()` methods\n- **Weather Data:** Use open-meteo for weather data (doesn't require API keys)\n unless otherwise specified\n- **View Source:** Add a view source link by importing & using\n `import.meta.url.replace(\"ems.sh\", \"val.town)\"` (or passing this data to the\n client) and include `target=\"_top\"` attribute\n- **Error Debugging:** Add\n `` to HTML to capture\n client-side errors\n- **Error Handling:** Only use try...catch when there's a clear local\n resolution; Avoid catches that merely log or return 500s. Let errors bubble up\n with full context\n- **Environment Variables:** Use `Deno.env.get('keyname')` when you need to, but\n generally prefer APIs that don't require keys\n- **Imports:** Use `https://esm.sh` for npm and Deno dependencies to ensure\n compatibility on server and browser\n- **Storage Strategy:** Only use backend storage if explicitly required; prefer\n simple static client-side sites\n- **React Configuration:** When using React libraries, pin versions with\n `?deps=react@18.2.0,react-dom@18.2.0` and start the file with\n `/** @jsxImportSource https://esm.sh/react@18.2.0 */`\n- Ensure all React dependencies and sub-dependencies are pinned to the same\n version\n- **Styling:** Default to using TailwindCSS via\n `` unless otherwise\n specified\n\n## Project Structure and Design Patterns\n\n### Recommended Directory Structure\n\n```\n├── backend/\n│ ├── database/\n│ │ ├── migrations.ts # Schema definitions\n│ │ ├── queries.ts # DB query functions\n│ │ └── README.md\n│ └── routes/ # Route modules\n│ ├── [route].ts\n│ └── static.ts # Static file serving\n│ ├── index.ts # Main entry point\n│ └── README.md\n├── frontend/\n│ ├── components/\n│ │ ├── App.tsx\n│ │ └── [Component].tsx\n│ ├── favicon.svg\n│ ├── index.html # Main HTML template\n│ ├── index.tsx # Frontend JS entry point\n│ ├── README.md\n│ └── style.css\n├── README.md\n└── shared/\n ├── README.md\n └── utils.ts # Shared types and functions\n```\n\n### Backend (Hono) Best Practices\n\n- Hono is the recommended API framework\n- Main entry point should be `backend/index.ts`\n- **Static asset serving:** Use the utility functions to read and serve project\n files:\n ```ts\n import {\n readFile,\n serveFile,\n } from \"https://esm.town/v/std/utils@85-main/index.ts\";\n\n // serve all files in frontend/ and shared/\n app.get(\"/frontend/*\", (c) => serveFile(c.req.path, import.meta.url));\n app.get(\"/shared/*\", (c) => serveFile(c.req.path, import.meta.url));\n\n // For index.html, often you'll want to bootstrap with initial data\n app.get(\"/\", async (c) => {\n let html = await readFile(\"/frontend/index.html\", import.meta.url);\n\n // Inject data to avoid extra round-trips\n const initialData = await fetchInitialData();\n const dataScript = ``;\n\n html = html.replace(\"\", `${dataScript}`);\n return c.html(html);\n });\n ```\n- Create RESTful API routes for CRUD operations\n- Always include this snippet at the top-level Hono app to re-throwing errors to\n see full stack traces:\n ```ts\n // Unwrap Hono errors to see original error details\n app.onError((err, c) => {\n throw err;\n });\n ```\n\n### Database Patterns\n\n- Run migrations on startup or comment out for performance\n- Change table names when modifying schemas rather than altering\n- Export clear query functions with proper TypeScript typing\n\n## Common Gotchas and Solutions\n\n1. **Environment Limitations:**\n - Val Town runs on Deno in a serverless context, not Node.js\n - Code in `shared/` must work in both frontend and backend environments\n - Cannot use `Deno` keyword in shared code\n - Use `https://esm.sh` for imports that work in both environments\n\n2. **SQLite Peculiarities:**\n - Limited support for ALTER TABLE operations\n - Create new tables with updated schemas and copy data when needed\n - Always run table creation before querying\n\n3. **React Configuration:**\n - All React dependencies must be pinned to 18.2.0\n - Always include `@jsxImportSource https://esm.sh/react@18.2.0` at the top of\n React files\n - Rendering issues often come from mismatched React versions\n\n4. **File Handling:**\n - Val Town only supports text files, not binary\n - Use the provided utilities to read files across branches and forks\n - For files in the project, use `readFile` helpers\n\n5. **API Design:**\n - `fetch` handler is the entry point for HTTP vals\n - Run the Hono app with\n `export default app.fetch // This is the entry point for HTTP vals`\n",[],[221],"082d66a8-cb63-4b34-8b31-43b2e73f63be",".vtignore","0989fcfa-4506-11f1-aa76-42b51c65c3df",["D",1789542072687],5,".git\n.jj\n.vscode\n.cursorrules\n.DS_Store\nnode_modules\nvendor\ninactive-feeds.tsv\n\n# Development-only: nothing at runtime reads these\nfeed.test.ts\nsources/cooperPress.fixtures\nsources/cooperPress.test.ts\nsources/twig.fixtures\nsources/twig.test.ts\nsources/wikipediaCurrentEvents.fixtures\nsources/wikipediaCurrentEvents.test.ts\nsources/hnDaily.fixtures\nsources/hnDaily.test.ts\n",[],[229],"f07c1a6c-a11b-4113-81a2-8f6fcf62f905",".gitignore","ba075c00-a97c-11f1-9aa5-1607ee4eb77e",41,["D",1788648858704],".vt/state.json\n",[],[237],"e7e6c9c5-d3f3-40f3-8ce2-b35f25d960b8","wikipediaCurrentEvents.ts","6704385c-b19c-11f1-b5e9-1607ee4eb77e","11350cd8-586b-11f1-ab2b-ee650bb23af1",75,["D",1789542072642],"import { type CheerioAPI, load } from \"npm:cheerio@1.2.0\";\nimport { type Element, isTag } from \"npm:domhandler@5.0.3\";\nimport { feedEndpoint } from \"../feed.ts\";\nimport {\n chunksOf,\n cutToWords,\n escapeHtml,\n type FetchedPage,\n fetchPage,\n hasContent,\n htmlToText,\n innerHtml,\n joinChunks,\n normalizeWhitespace,\n SENTENCE_END,\n} from \"./pageDrain.ts\";\n\n// Wikipedia's Current events portal keeps one page per month\n// (Portal:Current events/September 2026) holding a day-by-day log of the news.\n// Each day of that page is a tree of entries, the shape the portal's editors\n// write it in: a bullet names the subject of a story (\"Middle Eastern crisis\"),\n// the bullets nested under it refine that story (\"2026 Iran war\" → \"2026 Strait\n// of Hormuz crisis\" → \"September 2026 United States strikes on Iran\"), and a\n// bullet that holds no bullet of its own is the report itself — the sentence a\n// reader comes for, with the sources it rests on linked at its end.\n//\n// These feeds pull a month apart into one item per report: the report as the\n// page nests it, with the bullets that hold no part of it taken away, under a\n// line naming the day and the category it stands in, so that a story read out of\n// context still says where it stands. An item is named by the outermost subject\n// above it, which is the story it belongs to, and by the line the report opens\n// with — \"Russo-Ukrainian war: Russia drone strikes across Ukraine kill at least\n// 15 people…\" — and signs off with the day it was split from, which holds the\n// reports of its own the feed files elsewhere.\n//\n// What the feed carries is the month up to yesterday, never today: a day of\n// Current events is written as it happens and revised for hours after, so a day\n// reaches the feed only once 24 hours have passed for it to settle. Editors keep\n// revising old days for months, though, so an item is not named by its place in\n// a day but by a hash of the body it carries: a revision of the page hands the\n// item a new GUID, and a reader is shown the revision rather than a copy of the\n// day it first met.\n//\n// The page is read through ?action=render, the shortcut that serves its content\n// without the skin, the sidebar and the site's own machinery around it.\nconst CURRENT_EVENTS_URL =\n \"https://en.wikipedia.org/wiki/Portal:Current_events\";\n\n// A month page is named for its month, spelled out: .../September_2026\nconst MONTH_NAMES = [\n \"January\",\n \"February\",\n \"March\",\n \"April\",\n \"May\",\n \"June\",\n \"July\",\n \"August\",\n \"September\",\n \"October\",\n \"November\",\n \"December\",\n];\n\n// The site's own icon, the Wikipedia globe. It rides the feed metadata as the\n// JSON Feed icon, which jsonfeed-to-rss turns into the RSS channel image.\nconst WIKIPEDIA_ICON =\n \"https://en.wikipedia.org/static/images/icons/wikipedia.png\";\n\nconst wikipediaCurrentEventsExpression = `{\n \"version\": $jsonFeedVersion,\n \"title\": \"Wikipedia Current Events (decomposed)\",\n \"home_page_url\": \"${CURRENT_EVENTS_URL}\",\n \"feed_url\": $feedUrl,\n \"icon\": \"${WIKIPEDIA_ICON}\",\n \"description\": \"Every report of the settled days of the current Wikipedia Current Events month as an individual item, each carrying the day and category it stands in and every subject above it as context, named by the story it belongs to, and signed off with the day it was split from, plus a closing item for any content the split did not recognize and a single item noting the failed fetch where the site serves no month page at all\",\n \"items\": items.{\n \"id\": id,\n \"url\": url,\n \"title\": title,\n \"content_html\": content_html,\n \"content_text\": content_text,\n \"date_published\": date_published,\n \"tags\": tags\n }\n}`;\n\nexport const wikipediaCurrentEventsHandler = feedEndpoint({\n expression: wikipediaCurrentEventsExpression,\n loadSource: async () => {\n return { data: { items: await fetchMonthItems(new Date()) } };\n },\n});\n\n/** A month of Current events, as the page the feed reads it from */\nexport interface Month {\n year: number;\n /** The month as a person counts it, 1 for January */\n month: number;\n}\n\n/** One day of a month page, as the page lays it out */\nexport interface DayRegion {\n /** The day the region holds, e.g. 2026-09-01 */\n date: string;\n /** The day as the page names it, e.g. \"September 1, 2026 (Tuesday)\" */\n label: string;\n /** The anchor the day stands under on the month page, e.g. 2026_September_1 */\n anchor: string;\n /** The page the day keeps for itself, e.g. .../Portal:Current_events/2026_September_1 */\n url: string;\n /** The element the day's reports stand in */\n content: Element;\n}\n\n/** One report of a day: the bullet carrying it and the bullets above it */\nexport interface Report {\n /** The bullets above the report, outermost first */\n chain: Element[];\n /** The bullet the report itself is written in */\n leaf: Element;\n}\n\n/** A report, with the category the page files it under */\nexport interface DayReport extends Report {\n /** The category the report stands in, e.g. \"Armed conflicts and attacks\" */\n category?: string;\n}\n\nexport interface FeedItem {\n id: string;\n url: string;\n title: string;\n content_html: string;\n content_text: string;\n date_published?: string;\n tags?: string[];\n}\n\n/** Why no month could be read, and what answered in its place */\nexport type FetchFailure =\n | { kind: \"unreachable\"; url: string; error: string }\n | { kind: \"http-error\"; url: string; status: number; statusText?: string }\n | { kind: \"not-a-month\"; url: string };\n\n/** The page a month page holds, or the failure that stands in for it */\nexport type MonthRead =\n | { kind: \"month\"; html: string }\n | { kind: \"failure\"; failure: FetchFailure };\n\n// --- the month the feed carries --------------------------------------------\n\n/** The page Wikipedia keeps a month of Current events on */\nexport function monthPageUrl(month: Month): string {\n return `${CURRENT_EVENTS_URL}/${MONTH_NAMES[month.month - 1]}_${month.year}`;\n}\n\n// The month the feed carries is the month today falls in. While no day of that\n// month has settled — today is the 1st, so every day of it is either today or\n// ahead of it — the feed carries the month before, which is settled through its\n// last day, so that a reader following the feed is never handed nothing.\nexport function monthOfFeed(now: Date): Month {\n const year = now.getUTCFullYear();\n const month = now.getUTCMonth() + 1;\n if (now.getUTCDate() > 1) return { year, month };\n return month === 1\n ? { year: year - 1, month: 12 }\n : { year, month: month - 1 };\n}\n\n/** The day a moment falls on, e.g. 2026-09-16 */\nexport function isoDay(now: Date): string {\n return now.toISOString().slice(0, 10);\n}\n\n// The page the feed reads: the shortcut that serves the content of a page\n// without the skin, the sidebar and the site's own scripts around it\nfunction renderUrl(pageUrl: string): string {\n return `${pageUrl}?action=render`;\n}\n\n/** The items the feed carries for the month of the moment it was asked for */\nexport async function fetchMonthItems(now: Date): Promise {\n const month = monthOfFeed(now);\n const url = monthPageUrl(month);\n\n const read = readMonth(await fetchPage(renderUrl(url)), url);\n if (read.kind === \"failure\") return [failureItem(read.failure)];\n\n return await decomposeMonth(read.html, url, isoDay(now));\n}\n\n// --- reading what the site served ------------------------------------------\n//\n// The split reads one month of Current events. A change of host, or a firewall\n// put in front of the site, can leave Wikipedia serving something else entirely,\n// and a month that was never written serves nothing at all. None of that is a\n// month of Current events, so none of it is decomposed: the sentences such a\n// page holds are not reports of the portal, and reading them would hand the feed\n// a block screen's prose dressed up as the news.\n//\n// What makes a page a month of Current events is that it lays days out in the\n// regions the portal's own inclusion template writes, the containers a day's\n// reports stand in. That is structure of the portal rather than of a page's\n// layout, so it outlives a change of markup, and no other page carries it. The\n// guard stops there, at the page's shape: whether a day can be decomposed is\n// another matter, and a month whose days the split cannot read is still offered\n// as its items. Only a page that is no month is reported as a failure.\nconst DAY_REGION = \"div.current-events-main\";\n\nexport function readMonth(fetched: FetchedPage, monthUrl: string): MonthRead {\n if (fetched.kind === \"error\") {\n return {\n kind: \"failure\",\n failure: { kind: \"unreachable\", url: monthUrl, error: fetched.error },\n };\n }\n\n if (fetched.status < 200 || fetched.status >= 300) {\n return {\n kind: \"failure\",\n failure: {\n kind: \"http-error\",\n url: monthUrl,\n status: fetched.status,\n statusText: fetched.statusText,\n },\n };\n }\n\n if (load(fetched.html)(DAY_REGION).length === 0) {\n return {\n kind: \"failure\",\n failure: { kind: \"not-a-month\", url: monthUrl },\n };\n }\n\n return { kind: \"month\", html: fetched.html };\n}\n\n// --- month decomposition ---------------------------------------------------\n\n// The items the feed carries for the page the site served for a month: one per\n// report of its days, in order, each day closing with the item that holds\n// whatever its split did not recognize.\nexport async function decomposeMonth(\n html: string,\n monthUrl: string,\n /** The day the feed was asked for, e.g. 2026-09-16 */\n today: string,\n): Promise {\n const $ = load(html);\n\n // The day of today is still being written and revised, so the month is\n // carried up to yesterday, newest day first, as a feed reads\n const days = monthDays($).filter((day) => day.date < today);\n\n const items: FeedItem[] = [];\n for (const day of days.reverse()) {\n items.push(...await dayItems($, day, monthUrl));\n }\n return items;\n}\n\n/** The day regions a month page holds, in page order */\nexport function monthDays($: CheerioAPI): DayRegion[] {\n return $(DAY_REGION).toArray().flatMap((region) => {\n const date = dayDate($, region);\n if (!date) return [];\n\n // A day's reports stand in the element the page lays the day's content out\n // in, and nowhere else: the same region carries the day's own links to the\n // page's edit, history and watch views, which are no part of the news\n const content = $(region).find(\"div.current-events-content\").first().get(0);\n if (!content) return [];\n\n const anchor = anchorOf(date);\n return [{\n date,\n label: dayLabel($, region) ?? dayName(date),\n anchor,\n url: `${CURRENT_EVENTS_URL}/${anchor}`,\n content,\n }];\n });\n}\n\n// The day a region holds. The page writes the day into its own heading, hidden\n// (\"September 1, 2026 (2026-09-01) (Tuesday)\") for a reader on the page and for\n// machines reading it, and names the region after it as well\n// (\"2026_September_1\"); either one places the day.\nconst ISO_DAY = /^\\d{4}-\\d{2}-\\d{2}$/;\nconst NAMED_DAY = /^(\\d{4})_([A-Z][a-z]+)_(\\d{1,2})$/;\n\nfunction dayDate($: CheerioAPI, region: Element): string | undefined {\n const published = normalizeWhitespace($(region).find(\".bday\").first().text());\n if (ISO_DAY.test(published)) return published;\n\n const named = NAMED_DAY.exec($(region).attr(\"id\") ?? \"\");\n if (!named) return undefined;\n const month = MONTH_NAMES.indexOf(named[2]) + 1;\n if (month === 0) return undefined;\n return `${named[1]}-${String(month).padStart(2, \"0\")}-${\n named[3].padStart(2, \"0\")\n }`;\n}\n\n// The day as the page names it, e.g. \"September 1, 2026 (Tuesday)\". The day's\n// own date stands in the heading beside the name, hidden from a reader and\n// written for machines (\"2026-09-01\"), so it is no part of the line.\nfunction dayLabel($: CheerioAPI, region: Element): string | undefined {\n const title = $(region).find(\".current-events-title\").first().clone();\n title.find(\"*\").filter((_, el) =>\n /display\\s*:\\s*none/.test($(el).attr(\"style\") ?? \"\")\n ).remove();\n return normalizeWhitespace(title.text()) || undefined;\n}\n\n// The day as a line of the feed reads it, e.g. \"September 1, 2026\"\nfunction dayName(date: string): string {\n const [year, month, day] = date.split(\"-\").map(Number);\n return `${MONTH_NAMES[month - 1]} ${day}, ${year}`;\n}\n\n// The day as the page names it in an anchor and in a URL, e.g. 2026_September_1\nfunction anchorOf(date: string): string {\n const [year, month, day] = date.split(\"-\").map(Number);\n return `${year}_${MONTH_NAMES[month - 1]}_${day}`;\n}\n\n// The items the feed carries for one day: its reports, in page order, closed by\n// the item that holds whatever the drain of the day did not recognize\nasync function dayItems(\n $: CheerioAPI,\n day: DayRegion,\n monthUrl: string,\n): Promise {\n const items: FeedItem[] = [];\n\n for (const report of reportsOfDay($, day.content)) {\n // The page's own markup for the report, pruned to the report's branch of the\n // tree; a bullet the page left empty is no report, while one holding nothing\n // but an image still is one\n const branch = branchHtml($, report, monthUrl);\n if (!hasContent(branch)) continue;\n\n items.push(\n await dayFeedItem(day, monthUrl, {\n title: reportTitle($, report, day),\n body: `${dayLine(day, report.category)}\\n${branch}`,\n tags: report.category ? [report.category] : undefined,\n }),\n );\n }\n\n const closing = await leftoverItem($, day, monthUrl);\n if (closing) items.push(closing);\n\n return items;\n}\n\n// --- reports ---------------------------------------------------------------\n\n// Markup a day may carry inline that names nothing and holds no report: the\n// styles a template leaves on the page, and the machinery around them\nconst NON_CONTENT_TAGS = new Set([\"style\", \"script\", \"link\", \"meta\"]);\n\n// Every report of a day, in page order, each with the category the page files it\n// under. A day is written as a run of lists, one per category, under a line\n// naming that category — the page writes the line as a bold paragraph today and\n// as a heading div in the eras before it — and the eras before those write a day\n// as one list with no category at all.\nexport function reportsOfDay(\n $: CheerioAPI,\n content: Element,\n): DayReport[] {\n const reports: DayReport[] = [];\n let category: string | undefined;\n\n for (const child of $(content).children().toArray()) {\n if (!isTag(child) || NON_CONTENT_TAGS.has(child.tagName)) continue;\n\n if (isCategoryLine($, child)) {\n category = normalizeWhitespace($(child).text());\n continue;\n }\n\n // Every other element the day carries is a list of reports, or a wrapper\n // the day is laid out in (the div of a day's own container, the dd of a\n // definition list an era indents with) holding the lists that are\n const lists = isList(child) ? [child] : wrappedLists($, child);\n for (const list of lists) reports.push(...bulletsIn($, list, [], category));\n }\n\n return reports;\n}\n\n// The lists an element holds at whatever depth the page lays a day out in,\n// without the lists nested inside a list: those belong to the bullet they stand\n// under, which the walk of that list reads them through\nfunction wrappedLists($: CheerioAPI, container: Element): Element[] {\n return $(container).children().toArray().filter(isTag).flatMap((child) =>\n isList(child) ? [child] : wrappedLists($, child)\n );\n}\n\n// The line naming the category of the lists below it. The page gives a name no\n// link and no image, and gives it no list either: a line carrying a list is the\n// day's own layout, and a line carrying a link or an image is something else an\n// editor left in the day — a stray line of prose, a photograph with its caption\n// — which is content no report of the day holds rather than a name for the\n// lists under it.\nfunction isCategoryLine($: CheerioAPI, el: Element): boolean {\n if (isList(el)) return false;\n if ($(el).find(\"ul, ol, a, img\").length > 0) return false;\n return normalizeWhitespace($(el).text()).length > 0;\n}\n\n// The reports a list holds, each carrying the bullets that stand above it\nfunction bulletsIn(\n $: CheerioAPI,\n list: Element,\n chain: Element[],\n category?: string,\n): DayReport[] {\n return $(list).children(\"li\").toArray().flatMap((bullet) =>\n reportsUnder($, bullet, chain, category)\n );\n}\n\n// The reports a bullet holds: the bullet itself where it holds no bullet of its\n// own, and otherwise the bullets nested under it, in page order\nfunction reportsUnder(\n $: CheerioAPI,\n bullet: Element,\n chain: Element[],\n category?: string,\n): DayReport[] {\n const lists = $(bullet).children().toArray().filter(isTag).filter(isList);\n if (lists.length === 0) return [{ chain, leaf: bullet, category }];\n\n return lists.flatMap((list) =>\n bulletsIn($, list, [...chain, bullet], category)\n );\n}\n\n// A report as the page nests it: the bullet the report is written in, under the\n// bullets naming the subjects above it, with the bullets holding no part of the\n// report taken away. It is the page's own markup for the report, resolved\n// against the page, so a reader sees it as Wikipedia renders it — links, bold,\n// italics and the sources the report rests on alike.\nfunction branchHtml($: CheerioAPI, report: Report, baseUrl: string): string {\n let html = `

${innerHtml($, report.leaf, baseUrl)}

`;\n\n for (const bullet of [...report.chain].reverse()) {\n const label = innerHtml($, withoutLists($, bullet), baseUrl);\n html = `

${label}\\n

${html}

`;\n }\n\n return `

${html}`;\n}\n\n// A bullet without the bullets nested under it: the words it says of its own,\n// which name the subject it stands for\nfunction withoutLists($: CheerioAPI, bullet: Element): Element {\n const clone = $(bullet).clone();\n clone.children(\"ul, ol\").remove();\n return clone.get(0)!;\n}\n\nfunction bulletLabel($: CheerioAPI, bullet: Element): string {\n return normalizeWhitespace($(withoutLists($, bullet)).text());\n}\n\n// The name a report reaches the feed under: the outermost subject above it names\n// the story the report belongs to, and the report names itself with the line it\n// opens with. A report standing under no subject is named by that line alone,\n// and a report written without words of its own — a photograph, a map — is\n// named by the subjects above it or, where it has none either, by its day.\nfunction reportTitle(\n $: CheerioAPI,\n report: Report,\n day: DayRegion,\n): string {\n const opening = openingLine($, report.leaf);\n const story = report.chain.length > 0 ? bulletLabel($, report.chain[0]) : \"\";\n if (story && opening) return cutToWords(`${story}: ${opening}`);\n return story || opening || dayName(day.date);\n}\n\n// The line a report opens with: its own first sentence, which is the one saying\n// what happened (\"Four people are killed in Israeli strikes during a ground\n// operation in Gaza\"), the sources the page closes the report with no part of it\nfunction openingLine($: CheerioAPI, leaf: Element): string {\n const text = bulletLabel($, leaf);\n return cutToWords(text.split(SENTENCE_END)[0] ?? text);\n}\n\n// --- items -----------------------------------------------------------------\n\n// An item of a day: the body it carries, signed off with the day it was split\n// from and named by a hash of that body\nasync function dayFeedItem(\n day: DayRegion,\n monthUrl: string,\n fields: { title: string; body: string; tags?: string[] },\n): Promise {\n const signOff = daySignOff(day);\n return {\n id: await contentId(monthUrl, fields.body),\n url: `${monthUrl}#${day.anchor}`,\n title: fields.title,\n content_html: `${fields.body}\\n${signOff.html}`,\n content_text: `${htmlToText(fields.body)}\\n${signOff.text}`,\n date_published: `${day.date}T00:00:00.000Z`,\n tags: fields.tags,\n };\n}\n\n// The line an item opens with: the day it stands in, and the category the page\n// files it under where the page names one\nfunction dayLine(day: DayRegion, category?: string): string {\n const name = category ? `${day.label} · ${category}` : day.label;\n return `

${escapeHtml(name)}

`;\n}\n\n// The line every item of a day signs off with: a report is one of many a day\n// holds, and the page the day keeps for itself holds all of them, so a reader\n// who wants the report in its original surroundings has one click to them\nfunction daySignOff(day: DayRegion): { html: string; text: string } {\n const label = `Read the full ${dayName(day.date)} entry on Wikipedia`;\n return {\n html: `

${escapeHtml(label)}

`,\n text: label,\n };\n}\n\n// --- the GUID of an item ---------------------------------------------------\n//\n// A report is not named by where it stands in a day — its day, its category, its\n// place in the list — because the portal's editors keep revising old days for\n// months after their dates, moving reports, renaming subjects and rewriting the\n// sentences under them. An item is named by a hash of the body it carries\n// instead: what a reader has already seen keeps its name, and a report that was\n// rewritten reaches them as the revision it is.\n\nconst GUID_BYTES = 8;\n\nasync function contentId(pageUrl: string, bodyHtml: string): Promise {\n const digest = await crypto.subtle.digest(\n \"SHA-256\",\n new TextEncoder().encode(bodyHtml),\n );\n const hash = [...new Uint8Array(digest).slice(0, GUID_BYTES)]\n .map((byte) => byte.toString(16).padStart(2, \"0\"))\n .join(\"\");\n return `${pageUrl}#${hash}`;\n}\n\n// --- the item that carries what the split did not recognize ----------------\n//\n// The split above reads the day styles the page writes today, so a style the\n// portal adds tomorrow would drop its content from the feed. Whatever no report\n// of a day claimed — prose a day opens with, a block of a kind no era has\n// written yet — is serialized as plain feed HTML into one closing item, or comes\n// back undefined when nothing readable is left. See pageDrain.ts for what that\n// serialization keeps and what it drops.\n\n// Label of the item that carries whatever the split did not recognize\nexport const NON_DECOMPOSED_TAG = \"Non-decomposed\";\n\nasync function leftoverItem(\n $: CheerioAPI,\n day: DayRegion,\n monthUrl: string,\n): Promise {\n const leftover = leftoverHtml($, day.content, monthUrl);\n if (!hasContent(leftover)) return undefined;\n\n return await dayFeedItem(day, monthUrl, {\n title: `Non-decomposed ${dayName(day.date)} content`,\n body: `${dayLine(day)}\\n${leftover}`,\n tags: [NON_DECOMPOSED_TAG],\n });\n}\n\n// What no report of a day held: the lines naming the day's categories and the\n// lists standing under them taken off, everything else the day carries left as\n// it stands\nfunction leftoverHtml(\n $: CheerioAPI,\n content: Element,\n baseUrl: string,\n): string {\n const clone = $(content).clone();\n\n const named = clone.children().toArray().filter((child) =>\n isTag(child) && isCategoryLine($, child)\n );\n for (const line of named) $(line).remove();\n clone.find(\"ul, ol\").remove();\n\n return joinChunks(chunksOf($, clone.get(0)!, baseUrl), true);\n}\n\n// --- the item a failed fetch leaves behind --------------------------------\n\n// The tag of the item that stands in for a month the site did not serve\nexport const UNAVAILABLE_TAG = \"Unavailable\";\n\nconst UNAVAILABLE_TITLE = \"Month page unavailable\";\n\n// The one item the feed carries where no month could be read. It holds no\n// sign-off: there is no day to read on. It names itself and what answered in\n// place of the month, so that a reader who finds no reports learns why.\nfunction failureItem(failure: FetchFailure): FeedItem {\n const url = failure.url;\n const lead = \"Fetching the Wikipedia Current Events month page failed.\";\n const link = `${escapeHtml(url)}`;\n const contentHtml = `

${escapeHtml(lead)} ${link} ${\n escapeHtml(failureAnswer(failure))\n }.

`;\n\n return {\n // One item per URL, however often the failure is reported\n id: `${url}#unavailable`,\n url,\n title: UNAVAILABLE_TITLE,\n content_html: contentHtml,\n content_text: htmlToText(contentHtml),\n tags: [UNAVAILABLE_TAG],\n };\n}\n\n// What answered in place of the month, e.g. \"answered with HTTP 404 Not Found\"\nfunction failureAnswer(failure: FetchFailure): string {\n switch (failure.kind) {\n case \"unreachable\":\n return `could not be reached (${failure.error})`;\n case \"http-error\":\n return `answered with HTTP ${failure.status}${\n failure.statusText ? ` ${failure.statusText}` : \"\"\n }`;\n case \"not-a-month\":\n return \"answered with something other than a month of Current events\";\n }\n}\n\n// --- helpers --------------------------------------------------------------\n\nfunction isList(el: Element): boolean {\n return el.tagName === \"ul\" || el.tagName === \"ol\";\n}\n",[],[253,245],"sources","11c3fef2-586b-11f1-81b2-ee650bb23af1","vsMarketplace.ts","11c3ffa6-586b-11f1-81b2-ee650bb23af1",11,["D",1779735230121],"import { feedEndpoint, fetchJson, HttpError } from \"../feed.ts\";\n\ninterface MarketplaceFeed {\n title: string;\n category: string;\n}\n\ninterface MarketplaceProduct {\n json: string;\n title: string;\n feeds: Record;\n}\n\nconst vsMarketplaceExpression = `(\n $categoryItems := epc[cn = $category].e;\n {\n \"version\": $jsonFeedVersion,\n \"title\": $feedTitle,\n \"home_page_url\": \"https://marketplace.visualstudio.com/\",\n \"feed_url\": $feedUrl,\n \"description\": $feedTitle,\n \"items\": $categoryItems.{\n \"id\": \"https://marketplace.visualstudio.com/\" & l,\n \"url\": \"https://marketplace.visualstudio.com/\" & l,\n \"title\": t ? t : \"Untitled\",\n \"content_text\": \"(\" & $string(dc ? dc : 0) & \" downloads) \" & (s ? s : \"(no description)\"),\n \"date_published\": $now,\n \"author\": { \"name\": a ? a : \"Unknown Author\" }\n }\n }\n)`;\n\nexport const marketplaceProducts = {\n vscode: {\n json:\n \"https://marketplace.visualstudio.com/getextensionspercategory?Product=vscode&RemoveFirstSetCategories=true\",\n title: \"VS Code Marketplace\",\n feeds: {\n featured: { title: \"Featured\", category: \"Featured\" },\n popular: { title: \"Most Popular\", category: \"Most Popular\" },\n recent: { title: \"Recently Added\", category: \"Recently Added\" },\n },\n },\n vs: {\n json:\n \"https://marketplace.visualstudio.com/getextensionspercategory?Product=vs&RemoveFirstSetCategories=true\",\n title: \"Visual Studio Marketplace\",\n feeds: {\n featured: { title: \"Featured\", category: \"Featured\" },\n popular: { title: \"Most Popular\", category: \"Most Popular\" },\n highest: { title: \"Highest Rated\", category: \"Highest Rated\" },\n },\n },\n vsts: {\n json:\n \"https://marketplace.visualstudio.com/getextensionspercategory?Product=vsts&RemoveFirstSetCategories=true\",\n title: \"Azure DevOps Marketplace\",\n feeds: {\n featured: { title: \"Featured\", category: \"Featured\" },\n popular: { title: \"Most Popular\", category: \"Most Popular\" },\n recent: { title: \"Recently Added\", category: \"Recently Added\" },\n },\n },\n} satisfies Record;\n\nexport const vsMarketplaceHandler = feedEndpoint({\n expression: vsMarketplaceExpression,\n loadSource: async (c) => {\n const productKey = c.req.param(\"product\") ?? \"\";\n const feedKey = c.req.param(\"feedFile\")?.replace(/\\.(rss|json)$/, \"\");\n const product = marketplaceProducts[productKey];\n if (!product) throw new HttpError(404, \"Marketplace product not found\");\n if (!feedKey) throw new HttpError(400, \"Missing marketplace feed\");\n\n const feed = product.feeds[feedKey];\n if (!feed) throw new HttpError(404, `Feed not found for ${product.title}`);\n\n return {\n data: await fetchJson(product.json),\n bindings: {\n category: feed.category,\n feedTitle: `${product.title} - ${feed.title}`,\n now: new Date().toISOString(),\n },\n };\n },\n});\n",[],[253,255],"3cd9448d-8cc6-4aa8-a58c-4fc43f2c6d7a","twig.ts","b803fc64-b193-11f1-ac72-1607ee4eb77e",77,["D",1789542072690],"import {\n type Cheerio,\n type CheerioAPI,\n contains,\n load,\n} from \"npm:cheerio@1.2.0\";\nimport { type Element } from \"npm:domhandler@5.0.3\";\nimport { feedEndpoint } from \"../feed.ts\";\nimport {\n absolutize,\n chunksOf,\n cutToWords,\n escapeHtml,\n type FetchedPage,\n fetchPage,\n hasContent,\n htmlToText,\n joinChunks,\n normalizeWhitespace,\n SENTENCE_END,\n singleWordLinkTags,\n} from \"./pageDrain.ts\";\n\n// This Week in GNOME (thisweek.gnome.org) publishes one issue a week: a headline\n// of its own, a line naming the day it covers, and a run of notes contributed by\n// the community. A note is an attribution (\"Sophie (she/her) reports\") over a\n// blockquote, headed by the project it reports on and a one-line description of\n// that project where the note is written about one in particular, and the issue\n// groups its notes under section headings (\"GNOME Core Apps and Libraries\").\n// These feeds pull the latest issue apart into one item per note, so that a week\n// can be skimmed project by project. Whatever the split does not recognize is\n// offered as one closing item, so that a change of editorial style on their side\n// cannot drop content from the feed. A page that is no issue of the site at all\n// — a firewall's block screen, a parking page, an error document — is never read\n// as one: it reaches the feed as a single item noting the failed fetch.\nconst TWIG_HOME_URL = \"https://thisweek.gnome.org/\";\n\n// The site keeps no /latest: its archive lists every issue it ever published,\n// newest first, and the feed reads the highest issue number off it\nexport const TWIG_ARCHIVE_URL = `${TWIG_HOME_URL}posts/`;\n\n// The only icon the site offers. It rides the feed metadata as the JSON Feed\n// icon, which jsonfeed-to-rss turns into the RSS channel image.\nconst TWIG_ICON = `${TWIG_HOME_URL}favicon.svg`;\n\nconst twigExpression = `{\n \"version\": $jsonFeedVersion,\n \"title\": \"This Week in GNOME (decomposed)\",\n \"home_page_url\": \"${TWIG_HOME_URL}\",\n \"feed_url\": $feedUrl,\n \"icon\": \"${TWIG_ICON}\",\n \"description\": \"Every note of the latest This Week in GNOME issue as an individual item, tagged with the section it appears under, the project it reports on and every single word it links to, each signing off with the issue it was split from, plus a closing item for any content the split did not recognize and a single item noting the failed fetch where the site serves no issue at all\",\n \"items\": items.{\n \"id\": id,\n \"url\": url,\n \"title\": title,\n \"content_html\": content_html,\n \"content_text\": content_text,\n \"date_published\": date_published,\n \"tags\": tags\n }\n}`;\n\nexport const twigHandler = feedEndpoint({\n expression: twigExpression,\n loadSource: async () => {\n return { data: { items: await fetchLatestIssueItems() } };\n },\n});\n\n/** One issue of TWIG, as the feed read it off the page */\nexport interface TwigIssue {\n /** The page the issue keeps for itself, after redirects */\n url: string;\n html: string;\n /** The number the issue was published under, e.g. 260 */\n number?: number;\n /** The headline the issue was given, e.g. \"Freezing\" */\n headline?: string;\n /** The day the issue covers, e.g. 2026-07-31 */\n date?: string;\n /** The number of notes the issue says it holds, e.g. 9 */\n notes?: number;\n}\n\n/** The issue an item was split from, as its content signs off with it */\nexport interface IssueRef {\n url: string;\n number?: number;\n headline?: string;\n date?: string;\n}\n\nexport interface FeedItem {\n id: string;\n url?: string;\n title: string;\n content_html: string;\n content_text: string;\n date_published?: string;\n tags?: string[];\n}\n\n/** Why no issue could be read, and what answered in its place */\nexport type FetchFailure =\n | { kind: \"unreachable\"; url: string; error: string }\n | {\n kind: \"http-error\";\n url: string;\n status: number;\n statusText?: string;\n /** The title of the page that came with the error, when it has one */\n pageTitle?: string;\n }\n | { kind: \"not-an-issue\"; url: string; pageTitle?: string }\n | { kind: \"no-issue-linked\"; url: string; pageTitle?: string };\n\n/** The latest issue of TWIG, or the failure that stands in for it */\nexport type LatestIssue =\n | ({ kind: \"issue\" } & TwigIssue)\n | { kind: \"failure\"; failure: FetchFailure };\n\n/** The latest issue the archive points at, or why it points at none */\nexport type ArchiveRead =\n | { kind: \"issue-url\"; url: string }\n | { kind: \"failure\"; failure: FetchFailure };\n\n// The items the feed carries: the archive and the issue it points at are read\n// over the network, and either fetch can come back as a page that is no part of\n// the site — a firewall's block screen, a parking page, an error document —\n// which reaches the feed as one item noting the failed fetch.\nexport async function fetchLatestIssueItems(): Promise {\n const archive = readArchive(await fetchPage(TWIG_ARCHIVE_URL));\n if (archive.kind === \"failure\") return [fetchFailureItem(archive.failure)];\n\n return latestIssueItems(await fetchPage(archive.url));\n}\n\n// The items the feed carries for the page the site served for an issue: the\n// notes of the issue, or the one item that notes why there are none\nexport function latestIssueItems(fetched: FetchedPage): FeedItem[] {\n const latest = readLatestIssue(fetched);\n return latest.kind === \"issue\"\n ? decomposeIssue(latest.html, latest)\n : [fetchFailureItem(latest.failure)];\n}\n\n// The issue the feed carries is the newest one the archive links, i.e. the one\n// with the highest number, so the feed finds it however the archive is laid out:\n// /posts/2026/09/twig-265 is what it links those pages by. A page that links no\n// issue at all — a block screen, an error document — holds no latest issue.\nconst ISSUE_PATH = /\\/posts\\/\\d{4}\\/\\d{2}\\/twig-(\\d+)\\/?$/;\n\nexport function latestIssueUrl(\n html: string,\n baseUrl = TWIG_ARCHIVE_URL,\n): string | undefined {\n const $ = load(html);\n const issues = $(\"a[href]\").toArray().flatMap((el) => {\n const href = $(el).attr(\"href\");\n if (!href) return [];\n const url = new URL(href, baseUrl);\n const number = Number(ISSUE_PATH.exec(url.pathname)?.[1]);\n return Number.isNaN(number) ? [] : [{ number, url: url.toString() }];\n });\n\n return issues.sort((a, b) => b.number - a.number)[0]?.url;\n}\n\nexport function readArchive(fetched: FetchedPage): ArchiveRead {\n if (fetched.kind === \"error\") {\n return {\n kind: \"failure\",\n failure: { kind: \"unreachable\", url: fetched.url, error: fetched.error },\n };\n }\n\n const pageTitle = pageTitleOf(fetched.html);\n\n if (fetched.status < 200 || fetched.status >= 300) {\n return {\n kind: \"failure\",\n failure: {\n kind: \"http-error\",\n url: fetched.url,\n status: fetched.status,\n statusText: fetched.statusText,\n pageTitle,\n },\n };\n }\n\n const url = latestIssueUrl(fetched.html, fetched.url);\n if (!url) {\n return {\n kind: \"failure\",\n failure: { kind: \"no-issue-linked\", url: fetched.url, pageTitle },\n };\n }\n\n return { kind: \"issue-url\", url };\n}\n\n// --- reading what the site served -----------------------------------------\n//\n// The split reads one issue of the site. A change of host, or a firewall put in\n// front of the site, can leave it serving something else entirely: a Cloudflare\n// block or challenge screen, a parking page, a generic error document. None of\n// that is an issue, so none of it is decomposed. The notes such a page holds are\n// not notes of TWIG, and reading them would hand the feed a block screen's prose\n// dressed up as an issue. The feed asks the page to be an issue first, and notes\n// the failure as one item where it is not.\n//\n// What makes a page an issue of TWIG is that it names itself one: its document\n// title is the issue's own name (\"#260 Freezing\"), which is the headline the\n// page shows at the head of its article too. What the name does not survive is\n// being served another page: a block screen, a parking page or an error document\n// names itself instead, and so is no issue of the site.\n//\n// The guard deliberately stops there, at the page's name: whether an issue can\n// be decomposed is another matter, and an issue the split cannot read is still\n// offered whole as its closing item. Only a page that is no issue at all is\n// reported as a failure.\nconst ISSUE_NAME = /^#(\\d+)\\b\\s*(.*)$/;\n\nexport function issueName(\n value: string,\n): { number: number; headline?: string } | undefined {\n const match = ISSUE_NAME.exec(normalizeWhitespace(value));\n if (!match) return undefined;\n return { number: Number(match[1]), headline: match[2] || undefined };\n}\n\nexport function readLatestIssue(fetched: FetchedPage): LatestIssue {\n if (fetched.kind === \"error\") {\n return {\n kind: \"failure\",\n failure: { kind: \"unreachable\", url: fetched.url, error: fetched.error },\n };\n }\n\n const pageTitle = pageTitleOf(fetched.html);\n\n if (fetched.status < 200 || fetched.status >= 300) {\n return {\n kind: \"failure\",\n failure: {\n kind: \"http-error\",\n url: fetched.url,\n status: fetched.status,\n statusText: fetched.statusText,\n pageTitle,\n },\n };\n }\n\n const name = pageTitle ? issueName(pageTitle) : undefined;\n if (!name) {\n return {\n kind: \"failure\",\n failure: { kind: \"not-an-issue\", url: fetched.url, pageTitle },\n };\n }\n\n return {\n kind: \"issue\",\n url: fetched.url,\n html: fetched.html,\n ...name,\n ...parseIssueMeta(fetched.html),\n };\n}\n\nfunction pageTitleOf(html: string): string | undefined {\n return normalizeWhitespace(load(html)(\"title\").first().text()) || undefined;\n}\n\n// The issue's own date line (\"July 31, 2026 • 9 Notes • Curated by Felix\"), which\n// the page writes below the issue's headline. It names the week the issue covers\n// and how many notes it gathered.\nconst ISSUE_META = /([A-Z][a-z]+ \\d{1,2}, \\d{4})\\s*[•·]\\s*(\\d+)\\s+Notes?\\b/;\n\nexport function parseIssueMeta(\n html: string,\n): { date?: string; notes?: number } {\n const $ = load(html);\n for (const el of $(\"strong\").toArray()) {\n const match = ISSUE_META.exec(normalizeWhitespace($(el).text()));\n if (!match) continue;\n\n const date = new Date(`${match[1]} UTC`);\n return {\n date: Number.isNaN(date.getTime()) ? undefined : date.toISOString(),\n notes: Number(match[2]),\n };\n }\n return {};\n}\n\n// --- issue decomposition --------------------------------------------------\n//\n// The split drains the page of one issue: one pass over its elements that reads\n// each note, hands it to the feed as an item and takes the elements the item\n// holds off the page. Whatever the pass leaves behind is the content of the\n// closing item — which for an issue whose notes were all read is the words the\n// issue opens and closes with, the two of them holding no note of their own.\n//\n// Taking an element off the page is what claiming it means: what an item holds,\n// the closing item does not offer a second time.\n\n// Parts of the page that never hold editorial content: the site's own chrome\nconst FURNITURE_REGIONS = [\"nav\", \"header\", \"footer\", \"script\", \"style\"].join(\n \",\",\n);\n\n// Label of the item that carries whatever the split did not recognize\nexport const NON_DECOMPOSED_TAG = \"Non-decomposed\";\n\n// The tag of the item that stands in for an issue the site did not serve\nexport const UNAVAILABLE_TAG = \"Unavailable\";\n\n// A heading of an issue: the name it carries, the element it is written in drawn\n// on to take it off the page, and the anchor the page links the section by\ninterface Label {\n name: string;\n heading: Element;\n anchor?: string;\n}\n\n// A note of an issue, as the drain reads it off the page\ninterface Note {\n blockquote: Element;\n section?: Label;\n /** The project the note reports on, where the page names one */\n project?: Label;\n /** The line describing that project, where the note was headed by one */\n tagline?: Element;\n /** The line introducing the note, which names its author */\n attribution?: Element;\n}\n\nexport function decomposeIssue(html: string, issue: IssueRef): FeedItem[] {\n const $ = load(html);\n const items = drainIssue($, issueScope($), issue);\n\n // By now an item holds all the content it will carry (its own, the notes the\n // drain attached to it and the issue's sign-off), so the words it links to can\n // join its tags last, and the issue the note was split out of can sign off\n // below everything else\n const signOff = issueSignOff(issue);\n for (const item of items) {\n const tags = [...(item.tags ?? [])];\n for (const word of singleWordLinkTags(item.content_html)) {\n if (!tags.includes(word)) tags.push(word);\n }\n item.tags = tags.length > 0 ? tags : undefined;\n item.content_html += `\\n${signOff.html}`;\n item.content_text = item.content_text\n ? `${item.content_text}\\n${signOff.text}`\n : signOff.text;\n }\n\n return items;\n}\n\n// The container the issue's content lives in: the element holding the name the\n// page gives the issue. A page the site rebuilds differently still has its notes\n// read off it wherever they stand.\nfunction issueScope($: CheerioAPI): Element {\n const named = $(\"h1\").filter((_, el) => Boolean(issueName($(el).text())))\n .first();\n\n return named.parent().get(0) ?? $(\"article\").get(0) ?? $(\"body\").get(0)!;\n}\n\nfunction drainIssue(\n $: CheerioAPI,\n scope: Element,\n issue: IssueRef,\n): FeedItem[] {\n const items: FeedItem[] = [];\n\n // The section currently in effect (\"Third Party Projects\"), which labels and\n // tags every note below it\n let section: Label | undefined;\n // The project heading in effect. It names the note below it and no further: a\n // heading heads the one note written about that project, and the notes that\n // follow report on projects of their own.\n let project: Label | undefined;\n // The line below a project heading, describing the project it heads\n let tagline: Element | undefined;\n // The line above a note, introducing it with the note's author\n let attribution: Element | undefined;\n\n // Taking an element off the page is what claiming it means: what an item\n // holds, the closing item does not offer a second time\n const take = (elements: Array) => {\n for (const el of elements) if (el) $(el).remove();\n };\n\n for (const el of $(scope).find(\"*\").toArray()) {\n // Whatever an item above took off the page is not read a second time\n if (!contains(scope, el)) continue;\n\n const $el = $(el);\n if ($el.parents(FURNITURE_REGIONS).length > 0) continue;\n // Content is read with the block that holds it, so the paragraphs, list\n // items and images inside a note are not read as notes of their own\n if (\n $el.parents(\"h1, h2, h3, p, blockquote, ul, ol, li, figure\").length > 0\n ) {\n continue;\n }\n\n // The issue names itself with a headline and a date line of its own. Every\n // item signs off with the issue instead, so neither is content of one.\n if (el.tagName === \"h1\" && issueName($el.text())) {\n take([el]);\n continue;\n }\n if (el.tagName === \"strong\" && isIssueMetaLine($el)) {\n take([el]);\n continue;\n }\n\n if (el.tagName === \"h2\") {\n section = {\n name: headingName($, el),\n heading: el,\n anchor: $el.attr(\"id\"),\n };\n project = undefined;\n tagline = undefined;\n continue;\n }\n\n if (el.tagName === \"h3\") {\n project = {\n name: headingName($, el),\n heading: el,\n anchor: $el.attr(\"id\"),\n };\n tagline = undefined;\n continue;\n }\n\n if (el.tagName === \"p\") {\n // The line above a note introduces the note it stands above, be it with\n // the note's author or, where the project heading came without one, with\n // the project itself. Every other line of the issue is the issue's own\n // prose, which stays on the page for the closing item to carry.\n if ($el.next().is(\"blockquote\")) {\n attribution = el;\n } else if (project && !tagline) {\n tagline = el;\n }\n continue;\n }\n\n if (el.tagName !== \"blockquote\") continue;\n\n const note: Note = {\n blockquote: el,\n section,\n project,\n tagline,\n attribution,\n };\n // The headings that name the note travel with it; a heading no note ever\n // wears stays on the page and reaches the closing item instead\n take([section?.heading, project?.heading, tagline, attribution, el]);\n items.push(noteItem($, note, issue, items.length));\n project = undefined;\n tagline = undefined;\n attribution = undefined;\n }\n\n // What the drain did not take off the page closes the issue as one item\n const closing = leftoverItem($, scope, issue, items.length);\n if (closing) items.push(closing);\n\n return items;\n}\n\n// A note as an item of the feed: the name the page gives it, the content it\n// holds and the tags it wears\nfunction noteItem(\n $: CheerioAPI,\n note: Note,\n issue: IssueRef,\n index: number,\n): FeedItem {\n // The description of the project, the line introducing the note and the note\n // itself, as the page wrote them and resolved against it\n const held = [note.tagline, note.attribution, note.blockquote]\n .filter((el): el is Element => Boolean(el));\n for (const el of held) absolutize($, el, issue.url);\n\n return feedItem(issue, index, {\n title: noteTitle($, note),\n // A note stands in a section of the issue, and the page links its headings,\n // so the item points at the note's own place in the issue\n url: note.section?.anchor\n ? `${issue.url}#${note.section.anchor}`\n : issue.url,\n contentHtml: held.map((el) => $.html($(el)) ?? \"\").filter(Boolean).join(\n \"\\n\",\n ),\n tags: [note.section?.name, note.project?.name]\n .filter((tag): tag is string => Boolean(tag)),\n });\n}\n\n// The name a note reaches the feed under: the project heading names the note it\n// stands above, and a note the page gives no heading is named by its own opening\n// line. A note written without words of its own is named by the line introducing\n// it, or by the section it stands in.\nfunction noteTitle($: CheerioAPI, note: Note): string {\n return note.project?.name || openingLine($, note.blockquote) ||\n normalizeWhitespace($(note.attribution).text()) || note.section?.name ||\n \"Note\";\n}\n\n// An item of an issue, numbered by its place among the issue's items\nfunction feedItem(\n issue: IssueRef,\n index: number,\n fields: { title: string; url?: string; contentHtml: string; tags?: string[] },\n): FeedItem {\n return {\n id: `${issue.url}#${index}`,\n url: fields.url,\n title: fields.title,\n content_html: fields.contentHtml,\n content_text: htmlToText(fields.contentHtml),\n date_published: issue.date,\n tags: fields.tags,\n };\n}\n\n// The line every item of an issue signs off with: the note belongs to an issue\n// as a whole, and the issue carries notes the split files elsewhere, so a reader\n// is handed the one link that holds all of it\nfunction issueSignOff(issue: IssueRef): { html: string; text: string } {\n const label = `Read the full issue${issue.number ? ` #${issue.number}` : \"\"}${\n issue.headline ? ` ${issue.headline}` : \"\"\n }`;\n return {\n html: `

${escapeHtml(label)}

`,\n text: label,\n };\n}\n\n// --- closing item ---------------------------------------------------------\n//\n// The drain above only reads the note styles the site writes today, so a style\n// they add tomorrow would drop its content from the feed. Whatever no item of an\n// issue took off the page is serialized as plain feed HTML into one closing item\n// — or comes back undefined when nothing readable is left. See pageDrain.ts for\n// what that serialization keeps and what it drops.\n\nfunction leftoverItem(\n $: CheerioAPI,\n scope: Element,\n issue: IssueRef,\n /** The place the closing item takes among the items of its issue */\n index: number,\n): FeedItem | undefined {\n $(scope).find(FURNITURE_REGIONS).remove();\n\n const html = joinChunks(chunksOf($, scope, issue.url), true);\n if (!hasContent(html)) return undefined;\n\n return feedItem(issue, index, {\n title: issue.number\n ? `Non-decomposed issue #${issue.number} content`\n : \"Non-decomposed content\",\n url: issue.url,\n contentHtml: html,\n tags: [NON_DECOMPOSED_TAG],\n });\n}\n\n// --- headings -------------------------------------------------------------\n\n// The name a heading carries. A project heading links the project's own page\n// with an arrow (\"Glycin ↗\"), which is the page's furniture rather than part of\n// the name.\nfunction headingName($: CheerioAPI, el: Element): string {\n return normalizeWhitespace($(el).text()).replace(/\\s*↗\\uFE0F?\\s*$/u, \"\");\n}\n\n// The line an issue names itself with, below its headline\nfunction isIssueMetaLine($el: Cheerio): boolean {\n return ISSUE_META.test(normalizeWhitespace($el.text()));\n}\n\n// --- the headline of a note -----------------------------------------------\n\n// The words a note may greet its reader with before it names anything\nconst PLEASANTRY_WORDS =\n \"hi|hello|hey|greetings|good (morning|afternoon|evening)\";\n// A pleasantry on its own, e.g. \"Hello everyone!\"\nconst PLEASANTRY_SENTENCE = new RegExp(\n `^(?:${PLEASANTRY_WORDS})\\\\b[^.!?…]{0,24}[.!?…]?$`,\n \"i\",\n);\n// A pleasantry that opens the sentence naming the note, together with the\n// address that follows it, e.g. \"Hey people, this week …\"\nconst PLEASANTRY_OPENING = new RegExp(\n `^(?:${PLEASANTRY_WORDS})\\\\b[^.,!?…]{0,24}[.,!?…]\\\\s*`,\n \"i\",\n);\n\n// The line a note opens with, as its headline: the note's own first sentence,\n// which is the one that names what it reports (\"Sudoku v1.8.0 is out!\") — the\n// one behind the pleasantry where the note greets its reader first — and the\n// words that fit of it where that sentence runs long\nfunction openingLine($: CheerioAPI, blockquote: Element): string {\n const paragraphs = $(blockquote).children(\"p\").toArray()\n .map((el) => normalizeWhitespace($(el).text()))\n .filter(Boolean);\n\n for (const paragraph of paragraphs) {\n const headline = headlineOf(paragraph);\n if (headline) return cutToWords(headline);\n }\n return \"\";\n}\n\n// A paragraph of a note, read for the line the note names itself with. A note\n// that greets its reader first names itself in the sentence behind the greeting\n// or, where the greeting is a line of its own, in the paragraph below it.\nfunction headlineOf(paragraph: string): string {\n const opening = paragraph.split(SENTENCE_END)\n .find((sentence) => !PLEASANTRY_SENTENCE.test(sentence.trim())) ??\n paragraph;\n const named = normalizeWhitespace(opening.replace(PLEASANTRY_OPENING, \"\"));\n if (!named) return \"\";\n // The words behind a greeting now open the headline, capitalized as a headline\n // is\n return named === normalizeWhitespace(opening)\n ? named\n : named.charAt(0).toUpperCase() + named.slice(1);\n}\n\n// --- the item a failed fetch leaves behind --------------------------------\n\nconst UNAVAILABLE_TITLE = \"Latest issue unavailable\";\n\n// The one item the feed carries where no issue could be read. It holds no\n// sign-off: there is no issue to read on. It names itself and what answered in\n// place of the issue, so that a reader who finds no notes learns why.\nfunction fetchFailureItem(failure: FetchFailure): FeedItem {\n const url = failure.url;\n const lead = \"Fetching the latest This Week in GNOME issue failed.\";\n const link = `${escapeHtml(url)}`;\n const contentHtml = `

${escapeHtml(lead)} ${link} ${\n escapeHtml(failureAnswer(failure))\n }.

`;\n\n return {\n // One note per URL, however often the failure is reported\n id: `${url}#unavailable`,\n url,\n title: UNAVAILABLE_TITLE,\n content_html: contentHtml,\n content_text: htmlToText(contentHtml),\n tags: [UNAVAILABLE_TAG],\n };\n}\n\n// What answered in place of the issue, e.g. \"answered with HTTP 403 Forbidden\"\nfunction failureAnswer(failure: FetchFailure): string {\n switch (failure.kind) {\n case \"unreachable\":\n return `could not be reached (${failure.error})`;\n case \"http-error\":\n return `answered with HTTP ${failure.status}${\n failure.statusText ? ` ${failure.statusText}` : \"\"\n }${pageTitleNote(failure.pageTitle)}`;\n case \"not-an-issue\":\n return failure.pageTitle\n ? `answered with the page \"${failure.pageTitle}\" instead of an issue of the site`\n : \"answered with something other than an issue of the site\";\n case \"no-issue-linked\":\n return failure.pageTitle\n ? `answered with the page \"${failure.pageTitle}\", which links no issue of the site`\n : \"answered with a page that links no issue of the site\";\n }\n}\n\nfunction pageTitleNote(pageTitle?: string): string {\n return pageTitle ? ` (\"${pageTitle}\")` : \"\";\n}\n",[],[253,263],"b0e08402-996f-4064-b12e-66ef677b49bd","telegram.ts","5793e2ec-7ffa-11f1-8432-1607ee4eb77e",33,["D",1784084861296],"import { feedEndpoint, HttpError } from \"../feed.ts\";\n// @deno-types=\"https://esm.sh/gh/akopachov/telegram2rss@be14a9831923d49f1876a9027775943ebb464223/src/telegram-parser.ts?dts\"\nimport {\n getChannelInfoWithPosts,\n type Channel,\n type Media,\n type Poll,\n type Post,\n type Reply,\n} from \"https://esm.sh/gh/akopachov/telegram2rss@be14a9831923d49f1876a9027775943ebb464223/src/telegram-parser.ts\";\nimport { innerTextUntil } from \"https://esm.sh/gh/akopachov/telegram2rss@be14a9831923d49f1876a9027775943ebb464223/src/domutils-extensions.ts\";\nimport render from \"npm:dom-serializer@2.0.0\";\nimport { getChildren, isTag, removeElement } from \"npm:domutils@3.1.0\";\nimport type { AnyNode } from \"npm:domhandler@5.0.3\";\n\nconst WhitelistedAttributes = new Set([\n \"href\",\n \"src\",\n \"alt\",\n \"title\",\n \"target\",\n \"rel\",\n]);\nconst DefaultTitleMaxLength = 100;\n\nexport interface TelegramChannelInfo {\n title: string;\n}\n\nexport const telegramChannels: Record = {\n telegram: { title: \"Telegram News\" },\n techcrimes: { title: \"Tech Crimes\" },\n};\n\nconst telegramExpression = `{\n \"version\": $jsonFeedVersion,\n \"title\": channelTitle,\n \"home_page_url\": homePageUrl,\n \"feed_url\": $feedUrl,\n \"description\": channelDescription,\n \"icon\": logoUrl,\n \"items\": posts.{\n \"id\": id,\n \"url\": link,\n \"title\": title,\n \"content_html\": contentHtml,\n \"date_published\": date,\n \"attachments\": $count(attachments) > 0 ? attachments.{\"url\": url, \"mime_type\": mimeType, \"size_in_bytes\": sizeInBytes} : null\n }\n}`;\n\nexport const telegramHandler = feedEndpoint({\n expression: telegramExpression,\n loadSource: async (c) => {\n const channelName = c.req.param(\"channel\")?.replace(/\\.(rss|json)$/, \"\");\n if (!channelName) throw new HttpError(400, \"Missing Telegram channel name\");\n\n const countRaw = c.req.query(\"count\");\n const count = countRaw ? Math.min(Number(countRaw), 50) : undefined;\n\n const titleMaxLengthRaw = c.req.query(\"titleMaxLength\");\n const titleMaxLength = titleMaxLengthRaw\n ? Number(titleMaxLengthRaw)\n : undefined;\n\n let channel: Channel;\n try {\n channel = await getChannelInfoWithPosts(channelName, { count });\n } catch (e) {\n const message = e instanceof Error ? e.message : String(e);\n if (message.includes(\"Unknown channel\")) {\n throw new HttpError(404, `Telegram channel not found: ${channelName}`);\n }\n throw new HttpError(502, `Failed to fetch Telegram channel: ${message}`);\n }\n\n const posts = await Promise.all(\n (channel.posts || []).map((post) => transformPost(post, titleMaxLength)),\n );\n\n return {\n data: {\n channelTitle: channel.title,\n channelDescription: channel.description,\n homePageUrl: channel.link,\n logoUrl: channel.logoUrl,\n posts,\n },\n };\n },\n});\n\nasync function transformPost(post: Post, titleMaxLength?: number) {\n let title = \"\";\n let description = \"\";\n let pollHtml = \"\";\n let replyHtml = \"\";\n\n if (post.textHtml) {\n const nodes = getChildren(post.textHtml);\n sanitizeHtml(nodes);\n description = renderHtml(nodes);\n title = generateTitle(nodes, titleMaxLength);\n }\n\n if (post.poll) {\n if (!title && post.poll.title) {\n const pollNodes = getChildren(post.poll.title);\n sanitizeHtml(pollNodes);\n title = generateTitle(pollNodes, titleMaxLength);\n }\n pollHtml = generatePollHtml(post.poll);\n }\n\n if (post.reply) {\n replyHtml = generateReplyHtml(post.reply);\n }\n\n const mediaHtml = post.media.map(generateMediaHtml).join(\"
\");\n const combinedDescription = [mediaHtml, description, pollHtml]\n .filter((e) => !!e)\n .join(\"
\");\n\n const attachments = await Promise.all(\n post.media.map(getMediaInfo),\n );\n\n return {\n id: post.id,\n link: post.link,\n date: post.date.toISOString(),\n title: title || `Post #${post.id}`,\n contentHtml: replyHtml + combinedDescription,\n attachments,\n };\n}\n\nfunction renderHtml(nodes: AnyNode[]): string {\n return render(nodes, {\n xmlMode: false,\n selfClosingTags: true,\n encodeEntities: false,\n });\n}\n\nfunction sanitizeHtml(nodes: AnyNode[]): void {\n const queue: AnyNode[] = [...nodes];\n while (queue.length > 0) {\n const node = queue.shift()!;\n if (isTag(node)) {\n const children = getChildren(node);\n queue.push(...children);\n if (children.length === 0 && node.tagName !== \"br\") {\n removeElement(node);\n }\n for (const attr of Object.keys(node.attribs)) {\n if (!WhitelistedAttributes.has(attr.toLowerCase())) {\n delete node.attribs[attr];\n }\n }\n }\n }\n}\n\nfunction generateTitle(nodes: AnyNode[], maxLength: number = DefaultTitleMaxLength): string {\n const title = innerTextUntil(nodes, \"br\").trim();\n if (title.length <= maxLength) return title;\n\n const endOfSentence = /[.!?]+\\s/gi;\n let lastIndexInRange = title.length;\n let match;\n while ((match = endOfSentence.exec(title)) != null) {\n if (match.index > maxLength) {\n if (match.index < lastIndexInRange) {\n lastIndexInRange = match.index;\n }\n break;\n }\n lastIndexInRange = match.index;\n }\n\n if (lastIndexInRange > 0 && lastIndexInRange < title.length) {\n return title.slice(0, lastIndexInRange + 1);\n }\n return title;\n}\n\nfunction generateMediaHtml(media: Media): string {\n switch (media.type) {\n case \"photo\":\n return `

`;\n case \"video\":\n return `

`;\n case \"audio\":\n return `

`;\n default:\n return \"\";\n }\n}\n\nasync function getMediaInfo(\n media: Media,\n): Promise<{ url: string; mimeType?: string; sizeInBytes?: number }> {\n try {\n const response = await fetch(media.url, { method: \"HEAD\" });\n const contentType = response.headers.get(\"Content-Type\") || undefined;\n const contentLength = response.headers.get(\"Content-Length\");\n return {\n url: media.url,\n mimeType: contentType,\n sizeInBytes: contentLength ? Number(contentLength) : undefined,\n };\n } catch {\n return { url: media.url };\n }\n}\n\nfunction generatePollHtml(poll: Poll): string {\n let titleHtml = \"\";\n if (poll.title) {\n const nodes = getChildren(poll.title);\n sanitizeHtml(nodes);\n titleHtml = renderHtml(nodes);\n }\n const parts = [\n `

${titleHtml}

`,\n ];\n for (const option of poll.options) {\n parts.push(\n ``,\n );\n }\n parts.push(\"

${option.percent}%

${option.text}

\");\n return parts.join(\"\");\n}\n\nfunction generateReplyHtml(reply: Reply): string {\n let replyText = \"\";\n if (reply.textHtml) {\n const nodes = getChildren(reply.textHtml);\n sanitizeHtml(nodes);\n replyText = renderHtml(nodes);\n }\n return `

${reply.author}
${replyText}

`;\n}\n",[],[253,271],"24d228be-5881-11f1-ae73-ee650bb23af1","storesCom.ts","e24e3ed8-5880-11f1-bd64-ee650bb23af1",30,["D",1779744711686],"import { load } from \"npm:cheerio@1.2.0\";\nimport { feedEndpoint, HttpError } from \"../feed.ts\";\n\ninterface Store {\n title: string;\n domain: string;\n homeUrl: string;\n}\n\ninterface StoreProduct {\n id: string;\n url: string;\n title: string;\n content_html: string;\n content_text: string;\n}\n\nconst STORES: Record = {\n morningsave: {\n title: \"MorningSave\",\n domain: \"morningsave.com\",\n homeUrl: \"https://morningsave.com\",\n },\n sidedeal: {\n title: \"SideDeal\",\n domain: \"sidedeal.com\",\n homeUrl: \"https://sidedeal.com\",\n },\n};\n\nconst NEW_ARRIVALS_PATH = \"/market/new-arrivals\";\n\nconst storesComExpression = `{\n \"version\": $jsonFeedVersion,\n \"title\": storeTitle & \" New Arrivals\",\n \"home_page_url\": storeHomeUrl & \"/market/new-arrivals\",\n \"feed_url\": $feedUrl,\n \"description\": \"New arrivals from \" & storeTitle,\n \"items\": items.{\n \"id\": id,\n \"url\": url,\n \"title\": title,\n \"content_html\": content_html,\n \"content_text\": content_text,\n \"date_published\": $now()\n }\n}`;\n\nexport const storesComHandler = feedEndpoint({\n expression: storesComExpression,\n loadSource: async (c) => {\n const site = c.req.param(\"site\")?.replace(/\\.(rss|json)$/, \"\");\n if (!site) throw new HttpError(400, \"Missing store site\");\n\n const store = STORES[site];\n if (!store) throw new HttpError(404, \"Store not found. Use: morningsave or sidedeal\");\n\n return {\n data: {\n items: await scrapeNewArrivals(store),\n storeTitle: store.title,\n storeHomeUrl: store.homeUrl,\n },\n };\n },\n});\n\nasync function scrapeNewArrivals(store: Store): Promise {\n const url = `https://${store.domain}${NEW_ARRIVALS_PATH}`;\n const response = await fetch(url, {\n headers: {\n \"User-Agent\": \"Mozilla/5.0 (compatible; RSSGallery/1.0; +https://val.town)\",\n \"Accept\": \"text/html\",\n },\n });\n if (!response.ok) {\n throw new Error(\n `Failed to fetch ${url}: ${response.status} ${response.statusText}`,\n );\n }\n\n const $ = load(await response.text());\n const products: StoreProduct[] = [];\n\n $(\"div.offer\").each((_, offer) => {\n const $offer = $(offer);\n const offerId = $offer.attr(\"id\") ?? \"\";\n\n const titleEl = $offer.find(\"h2 a\");\n const productHref = titleEl.attr(\"href\") ?? \"\";\n const productTitle = titleEl.text().trim();\n if (!productTitle || !productHref) return;\n\n const productUrl = `https://${store.domain}${productHref}`;\n\n const photoEl = $offer.find(\"div.photo a\");\n const imgEl = photoEl.find(\"img\");\n const imgSrc = imgEl.attr(\"src\") ?? \"\";\n const imgAlt = imgEl.attr(\"alt\") ?? productTitle;\n\n const listPrice = $offer.find(\"span.list-price\").text().trim();\n\n const salePriceEl = $offer.find(\"span.sale-price.formatted\");\n const quantifier = salePriceEl.find(\"span.quantifier\").text().trim();\n const salePrice = salePriceEl.find(\"span.amount\").text().trim();\n const salePriceText = quantifier\n ? `${quantifier}${salePrice}`\n : salePrice;\n\n const discountText = $offer.find(\"span.discount\").text().trim();\n const arrivalDateText = $offer.find(\"span.arrival-date\").text().trim();\n const subtext = $offer.find(\"span.subtext\").text().trim();\n\n const contentHtml = [\n imgSrc ? ` $\"${escapeHtml(imgAlt)}\"$ ` : \"\",\n \"

\",\n listPrice ? `~~${escapeHtml(listPrice)}~~ ` : \"\",\n `${escapeHtml(salePriceText)}`,\n discountText ? ` — ${escapeHtml(discountText)}` : \"\",\n arrivalDateText ? `
${escapeHtml(arrivalDateText)}` : \"\",\n subtext ? `
${escapeHtml(subtext)}` : \"\",\n \"

\",\n ].join(\"\");\n\n const contentText = [\n `Price: ${salePriceText}`,\n listPrice ? `Was: ${listPrice}` : \"\",\n discountText,\n arrivalDateText,\n subtext,\n ].filter(Boolean).join(\" | \");\n\n products.push({\n id: offerId || productUrl,\n url: productUrl,\n title: productTitle,\n content_html: contentHtml,\n content_text: contentText,\n });\n });\n\n return products.slice(0, 100);\n}\n\nfunction escapeHtml(value: string): string {\n return value\n .replace(/&/g, \"&\")\n .replace(//g, \">\")\n .replace(/\"/g, \""\");\n}\n",[],[253,279],"b0c7acf5-3ca4-44f8-97dc-c8cff91401bd","sqliteChronology.ts","29b934ce-a991-11f1-bfd2-1607ee4eb77e",53,["D",1788657636013],"import { load } from \"npm:cheerio@1.2.0\";\nimport { feedEndpoint } from \"../feed.ts\";\n\nconst CHRONOLOGY_URL = \"https://www.sqlite.org/chronology.html\";\nconst NOTES_LIMIT = 12; // newest releases; each needs its own notes-page fetch\n\ninterface SqliteRelease {\n version: string;\n url: string;\n datePublished?: string;\n title: string;\n content_html: string;\n content_text: string;\n}\n\nconst sqliteChronologyExpression = `{\n \"version\": $jsonFeedVersion,\n \"title\": \"SQLite Release Notes\",\n \"home_page_url\": \"${CHRONOLOGY_URL}\",\n \"feed_url\": $feedUrl,\n \"description\": \"Newest SQLite releases with full release notes from the chronology page\",\n \"items\": items.{\n \"id\": url,\n \"url\": url,\n \"title\": title,\n \"content_html\": content_html,\n \"content_text\": content_text,\n \"date_published\": datePublished\n }\n}`;\n\nexport const sqliteChronologyHandler = feedEndpoint({\n expression: sqliteChronologyExpression,\n loadSource: async () => ({\n data: { items: await scrapeReleasesWithNotes() },\n }),\n});\n\nasync function scrapeReleasesWithNotes(): Promise {\n const html = await fetchPage(CHRONOLOGY_URL);\n const $ = load(html);\n const releases: SqliteRelease[] = [];\n\n // Rows are newest-first; each has a date link to the timeline and a\n // version link to the release notes page\n const releaseRows = $(\"#chrontab tr\").get().filter((row) =>\n $(row).find(\"a[href*='releaselog']\").length > 0\n );\n for (const row of releaseRows.slice(0, NOTES_LIMIT)) {\n const $row = $(row);\n const $version = $row.find(\"a[href*='releaselog']\").first();\n const href = $version.attr(\"href\") ?? \"\";\n const version = $version.text().trim();\n if (!href || !version) continue;\n\n const url = new URL(href, CHRONOLOGY_URL).href;\n const dateText = $row.find(\"a[href*='timeline']\").first().text().trim();\n const datePublished = parseReleaseDate(dateText);\n\n let notes = { contentHtml: \"\", contentText: \"\" };\n try {\n notes = await fetchReleaseNotes(url);\n } catch (error) {\n // Keep the item title/date even if a single notes page fails\n console.error(`Failed to fetch release notes for ${version}:`, error);\n }\n\n const title = datePublished\n ? `SQLite ${version} (${datePublished.slice(0, 10)})`\n : `SQLite ${version}`;\n\n releases.push({\n version,\n url,\n datePublished,\n title,\n content_html: notes.contentHtml,\n content_text: notes.contentText,\n });\n }\n\n return releases;\n}\n\nasync function fetchReleaseNotes(\n url: string,\n): Promise<{ contentHtml: string; contentText: string }> {\n const html = await fetchPage(url);\n const $ = load(html);\n const $h2 = $(\"h2\").first();\n if ($h2.length === 0) return { contentHtml: \"\", contentText: \"\" };\n\n // Collect siblings from the

until the standard closing paragraph\n // (\"A complete list of SQLite releases ...\"), which ends the notes body\n const parts: string[] = [];\n let el = $h2.next();\n while (el.length > 0) {\n if (el.is(\"p\") && el.text().includes(\"complete list of SQLite releases\")) {\n break;\n }\n parts.push($.html(el));\n el = el.next();\n }\n\n // Make relative links (../c3ref/..., ../wal.html) absolute\n const $notes = load(parts.join(\"\"));\n $notes(\"a\").each((_, anchor) => {\n const href = $notes(anchor).attr(\"href\");\n if (href && /^https?:\\/\\//.test(href) === false) {\n $notes(anchor).attr(\"href\", new URL(href, url).href);\n }\n });\n\n const contentHtml = ($notes(\"body\").html() ?? \"\").replace(\n /
\\s<\\/p>/g,\n \"\",\n );\n const contentText = normalizeWhitespace($notes(\"body\").text());\n return { contentHtml, contentText };\n}\n\nfunction parseReleaseDate(value: string): string | undefined {\n const date = new Date(value);\n if (Number.isNaN(date.getTime())) return undefined;\n return date.toISOString();\n}\n\nfunction normalizeWhitespace(value: string): string {\n return value.replace(/\\s+/g, \" \").trim();\n}\n\nasync function fetchPage(url: string): Promise {\n const response = await fetch(url, {\n headers: {\n \"User-Agent\": \"Mozilla/5.0 (compatible; RSSGallery/1.0; +https://val.town)\",\n \"Accept\": \"text/html\",\n },\n });\n if (!response.ok) {\n throw new Error(\n `Failed to fetch ${url}: ${response.status} ${response.statusText}`,\n );\n }\n return await response.text();\n}",[],[253,287],"40102055-3506-4d16-a6bf-a5555aa9e767","rentry.ts","2761fb9e-95ef-11f1-98f9-1607ee4eb77e",39,["D",1786501493586],2,"import { load } from \"npm:cheerio@1.2.0\";\nimport { feedEndpoint, HttpError } from \"../feed.ts\";\n\nconst RENTRY_PAGE_URL = \"https://rentry.org\";\nconst RENTRY_ITEMS_LIMIT = 10;\n\ninterface RentryTableRow {\n id: string;\n title: string;\n content_html: string;\n content_text: string;\n}\n\nexport const rentryHandler = feedEndpoint({\n expression: `{\n \"version\": $jsonFeedVersion,\n \"title\": pageTitle,\n \"home_page_url\": pageUrl,\n \"feed_url\": $feedUrl,\n \"description\": \"Table from \" & pageUrl,\n \"items\": items\n}`,\n loadSource: async (c) => {\n const pageSlug = c.req.param(\"url\")?.replace(/\\.(rss|json)$/, \"\");\n\n if (!pageSlug || !/^[a-zA-Z0-9-]+$/.test(pageSlug)) {\n throw new HttpError(400, \"Invalid Rentry page name\");\n }\n\n const pageUrl = `${RENTRY_PAGE_URL}/${pageSlug}`;\n\n const response = await fetch(pageUrl, {\n headers: {\n \"User-Agent\":\n \"Mozilla/5.0 (compatible; RSSGallery/1.0; +https://val.town)\",\n \"Accept\": \"text/html\",\n },\n });\n if (response.status === 404) {\n throw new HttpError(404, `Rentry page not found: ${pageSlug}`);\n }\n if (!response.ok) {\n throw new Error(\n `Failed to fetch ${pageUrl}: ${response.status} ${response.statusText}`,\n );\n }\n\n const html = await response.text();\n const tableItems = scrapeFirstTable(html);\n if (tableItems.length === 0) {\n throw new HttpError(\n 502,\n `No table found on ${pageUrl}. Rentry pages need at least one table.`,\n );\n }\n\n const from = c.req.query(\"from\") ?? \"top\";\n const items = from !== \"bottom\"\n ? sliceItems(tableItems, \"top\")\n : sliceItems(tableItems, \"bottom\");\n\n return {\n data: {\n items,\n pageTitle: parseTitle(html) ?? `Rentry: ${pageSlug}`,\n pageUrl,\n },\n };\n },\n});\n\nfunction sliceItems(items: RentryTableRow[], from: \"top\" | \"bottom\") {\n return from === \"top\"\n ? items.slice(0, RENTRY_ITEMS_LIMIT)\n : items.slice(-RENTRY_ITEMS_LIMIT);\n}\n\nfunction scrapeFirstTable(html: string): RentryTableRow[] {\n const $ = load(html);\n const items: RentryTableRow[] = [];\n const $table = $(\"table\").first();\n if ($table.length === 0) return items;\n\n const headers = $table.find(\"thead th\").map((_, th) => $(th).text().trim())\n .get();\n\n $table.find(\"tbody tr\").each((_, row) => {\n const $row = $(row);\n const cells = $row.find(\"td\");\n if (cells.length === 0) return;\n\n const cellsHtml = cells.map((_, cell) => $.html(cell)).get().join(\"\");\n const contentHtml = buildRowTable(headers, cellsHtml);\n const contentText = cells\n .map((_, cell) => $(cell).text())\n .get()\n .map((text) => normalizeWhitespace(text))\n .join(\" | \");\n\n const title = cells.first().text().trim();\n\n items.push({\n id: hashRow(contentHtml),\n title,\n content_html: contentHtml,\n content_text: contentText,\n });\n });\n\n return items;\n}\n\nfunction buildRowTable(headers: string[], cellsHtml: string): string {\n const headerHtml = headers.length > 0\n ? `${\n headers.map((header) => `${escapeHtml(header)}`).join(\"\")\n }`\n : \"\";\n\n return `
${headerHtml}${cellsHtml}
`;\n}\n\nfunction escapeHtml(value: string): string {\n return value\n .replace(/&/g, \"&\")\n .replace(//g, \">\")\n .replace(/\"/g, \""\");\n}\n\nfunction parseTitle(html: string): string | undefined {\n const match = /]>([^<]*)<\\/title>/i.exec(html);\n if (!match) return undefined;\n const title = normalizeWhitespace(match[1]);\n return title || undefined;\n}\n\nfunction normalizeWhitespace(value: string): string {\n return value.replace(/\\s+/g, \" \").trim();\n}\n\nfunction hashRow(contentHtml: string): string {\n let h1 = 0x811c9dc5;\n let h2 = 0x01000193;\n for (let i = 0; i < contentHtml.length; i++) {\n const code = contentHtml.charCodeAt(i);\n h1 = Math.imul(h1 ^ code, 16777619);\n h2 = Math.imul(h2 ^ code, 2) + 0x41347f;\n }\n const x = (h1 ^ h2) >>> 0;\n return x.toString(36);\n}\n",[],[253,295],"2c99e9f7-7d13-4883-a66f-d66dbf9cc998","rekt.ts","2e32da20-a98f-11f1-9111-1607ee4eb77e",51,["D",1788656784507],"import { load } from \"npm:cheerio@1.2.0\";\nimport { feedEndpoint, HttpError } from \"../feed.ts\";\n\nconst REKT_URL = \"https://rekt.news/\";\n\ninterface RektPost {\n id: string;\n url: string;\n title: string;\n excerpt: string;\n tags: string[];\n datePublished?: string;\n content_html: string;\n content_text: string;\n}\n\nconst rektExpression = `{\n \"version\": $jsonFeedVersion,\n \"title\": \"Rekt\",\n \"home_page_url\": \"${REKT_URL}\",\n \"feed_url\": $feedUrl,\n \"description\": \"Crypto hacks, exploits, and security news from REKT\",\n \"items\": items.{\n \"id\": id,\n \"url\": url,\n \"title\": title,\n \"content_html\": content_html,\n \"content_text\": content_text,\n \"date_published\": datePublished,\n \"tags\": tags\n }\n}`;\n\nexport const rektHandler = feedEndpoint({\n expression: rektExpression,\n loadSource: async () => {\n const items = await scrapePosts();\n if (items.length === 0) {\n throw new HttpError(502, \"No posts found on rekt.news\");\n }\n return { data: { items } };\n },\n});\n\nasync function scrapePosts(): Promise {\n const response = await fetch(REKT_URL, {\n headers: {\n \"User-Agent\": \"Mozilla/5.0 (compatible; RSSGallery/1.0; +https://val.town)\",\n \"Accept\": \"text/html\",\n },\n });\n if (!response.ok) {\n throw new Error(\n `Failed to fetch ${REKT_URL}: ${response.status} ${response.statusText}`,\n );\n }\n\n const $ = load(await response.text());\n const posts: RektPost[] = [];\n\n $(\"article.post\").each((_, post) => {\n const $post = $(post);\n const $title = $post.find(\"h5.post-title a\").first();\n const title = $title.text().trim();\n const href = $title.attr(\"href\") ?? \"\";\n if (!title || !href) return;\n\n const url = new URL(href, REKT_URL).href;\n const excerpt = $post.find(\"section.post-excerpt p\").first().text().trim();\n const tags = $post.find(\"span.post-meta p a\").map((_, a) => $(a).text().trim()).get();\n const datePublished = parseDate($post.find(\"time\").first().text());\n\n posts.push({\n id: url,\n url,\n title,\n excerpt,\n tags,\n datePublished,\n content_html: buildContentHtml({ title, excerpt, tags, url }),\n content_text: buildContentText({ excerpt, tags }),\n });\n });\n\n return posts;\n}\n\n// \"Wednesday, September 2, 2026\" -> ISO date; weekday prefix is stripped for Date parsing\nfunction parseDate(value: string): string | undefined {\n const cleaned = value.replace(/^[A-Za-z]+,/, \"\").trim();\n const date = new Date(cleaned);\n if (Number.isNaN(date.getTime())) return undefined;\n return date.toISOString();\n}\n\nfunction buildContentHtml(post: {\n title: string;\n excerpt: string;\n tags: string[];\n url: string;\n}): string {\n const parts: string[] = [];\n if (post.excerpt) {\n parts.push(`
${escapeHtml(post.excerpt)}
`);\n }\n if (post.tags.length > 0) {\n parts.push(\n `
${post.tags.map((tag) =>\n `${escapeHtml(tag)}`\n ).join(\" · \")}
`,\n );\n }\n parts.push(`
Read the full report
`);\n return parts.join(\"\");\n}\n\nfunction buildContentText(post: { excerpt: string; tags: string[] }): string {\n return [post.excerpt, post.tags.join(\", \")].filter(Boolean).join(\" | \");\n}\n\nfunction escapeHtml(value: string): string {\n return value\n .replace(/&/g, \"&\")\n .replace(//g, \">\")\n .replace(/\"/g, \""\");\n}",[],[253,304],"dd97ae99-32fe-4bc2-af1c-1976a6bfb8d3","pageDrain.ts","b7d75c7c-b193-11f1-825a-1607ee4eb77e",78,["D",1789542072686],"import { type CheerioAPI, load } from \"npm:cheerio@1.2.0\";\nimport {\n type AnyNode,\n type Element,\n isTag,\n isText,\n} from \"npm:domhandler@5.0.3\";\n\n// The toolkit the page-draining feeds share: the ones that read a page and hand\n// it to the feed as one item per entry it holds (the Cooper Press newsletters\n// and This Week in GNOME). It fetches a page as the feed reads it off the wire,\n// resolves what an item links to and shows against the page it came from, and\n// serializes whatever an item leaves behind into feed HTML.\n//\n// Serializing is what drops the scaffolding of a page: wrapper tables, spacer\n// cells and the like hold neither text nor an image of their own, so they leave\n// no chunk behind and need no sweep of their own.\n\nconst USER_AGENT =\n \"Mozilla/5.0 (compatible; RSSGallery/1.0; +https://val.town)\";\n\n/** The page a site served for a URL, as the feed read it off the wire */\nexport type FetchedPage =\n | {\n kind: \"page\";\n /** The URL the response came from, after redirects */\n url: string;\n status: number;\n statusText: string;\n html: string;\n }\n | {\n kind: \"error\";\n /** The URL the feed asked for /\n url: string;\n error: string;\n };\n\nexport async function fetchPage(url: string): Promise {\n try {\n const response = await fetch(url, {\n headers: { \"User-Agent\": USER_AGENT, \"Accept\": \"text/html\" },\n });\n return {\n kind: \"page\",\n // A URL that redirects answers with the URL it settled on, which is the\n // canonical one the feed links back to\n url: response.url || url,\n status: response.status,\n statusText: response.statusText,\n html: await response.text(),\n };\n } catch (error) {\n // A host that is gone, a firewall that drops the request: there is no page\n // to read, and the feed says so rather than failing as a whole\n return { kind: \"error\", url, error: errorMessage(error) };\n }\n}\n\nexport function errorMessage(error: unknown): string {\n return error instanceof Error ? error.message : String(error);\n}\n\n// --- serializing what an item leaves behind --------------------------------\n\n// A feed item's content is made of blocks (paragraphs, lists, headings, images)\n// and the inline runs between them, which are only wrapped in a paragraph once\n// they turn out to be content of their own.\nexport interface Chunk {\n html: string;\n block: boolean;\n}\n\n// Wrapper markup (tables, divs, spans…) carries no meaning of its own, so it is\n// unwrapped and its content is read where it lies\nconst BLOCK_TAGS = new Set([\n \"p\",\n \"ul\",\n \"ol\",\n \"li\",\n \"h1\",\n \"h2\",\n \"h3\",\n \"h4\",\n \"h5\",\n \"h6\",\n \"blockquote\",\n \"pre\",\n \"figure\",\n \"figcaption\",\n \"dl\",\n \"dt\",\n \"dd\",\n]);\n\nconst INLINE_TAGS = new Set([\n \"a\",\n \"strong\",\n \"b\",\n \"em\",\n \"i\",\n \"u\",\n \"s\",\n \"strike\",\n \"del\",\n \"ins\",\n \"code\",\n \"kbd\",\n \"samp\",\n \"var\",\n \"mark\",\n \"small\",\n \"sub\",\n \"sup\",\n \"cite\",\n \"q\",\n \"abbr\",\n \"time\",\n \"br\",\n \"wbr\",\n]);\n\n// Markup a feed reader has no use for: scripts, embeds and form controls\nconst DROPPED_TAGS = new Set([\n \"script\",\n \"style\",\n \"noscript\",\n \"template\",\n \"iframe\",\n \"svg\",\n \"canvas\",\n \"video\",\n \"audio\",\n \"object\",\n \"embed\",\n \"form\",\n \"input\",\n \"select\",\n \"textarea\",\n \"button\",\n \"meta\",\n \"link\",\n \"base\",\n \"title\",\n \"head\",\n]);\n\nexport function chunksOf(\n $: CheerioAPI,\n node: AnyNode,\n baseUrl: string,\n): Chunk[] {\n if (isText(node)) {\n // A text run holds what it holds, whitespace included: the run that stands\n // between two links is the space that keeps their words apart, and a run\n // that turns out to hold nothing but whitespace is trimmed away where the\n // chunks around it are joined\n return node.data.length > 0\n ? [{ html: escapeHtml(node.data), block: false }]\n : [];\n }\n if (!isTag(node)) return [];\n\n const tag = node.tagName.toLowerCase();\n if (DROPPED_TAGS.has(tag)) return [];\n\n if (tag === \"img\") {\n const src = $(node).attr(\"src\");\n if (!src) return [];\n return [{\n html: imageHtml(src, $(node).attr(\"alt\") ?? \"\", baseUrl),\n block: true,\n }];\n }\n if (tag === \"br\") return [{ html: \"
\", block: false }];\n\n const inner = joinChunks(childrenChunks($, node, baseUrl), false);\n if (BLOCK_TAGS.has(tag)) {\n return [{ html: `<${tag}>${inner}`, block: true }];\n }\n if (tag === \"a\") {\n const href = $(node).attr(\"href\");\n if (!href) return [{ html: inner, block: false }];\n const url = escapeHtml(new URL(href, baseUrl).toString());\n return [{ html: `${inner}`, block: false }];\n }\n if (INLINE_TAGS.has(tag)) {\n return [{ html: `<${tag}>${inner}`, block: false }];\n }\n\n return childrenChunks($, node, baseUrl);\n}\n\nfunction childrenChunks(\n $: CheerioAPI,\n el: Element,\n baseUrl: string,\n): Chunk[] {\n return $(el).contents().toArray().flatMap((child) =>\n chunksOf($, child, baseUrl)\n );\n}\n\n// The content an element holds, serialized as feed HTML, without the element\n// around it: what an item writes inside a tag it writes for itself\nexport function innerHtml(\n $: CheerioAPI,\n el: Element,\n baseUrl: string,\n): string {\n return joinChunks(childrenChunks($, el, baseUrl), false);\n}\n\nexport function joinChunks(chunks: Chunk[], wrapRuns: boolean): string {\n const parts: string[] = [];\n let run: string[] = [];\n\n const flush = () => {\n const html = run.join(\"\").trim();\n run = [];\n if (hasContent(html)) parts.push(wrapRuns ? `
${html}
` : html);\n };\n\n for (const chunk of chunks) {\n if (!chunk.block) {\n run.push(chunk.html);\n continue;\n }\n flush();\n if (hasContent(chunk.html)) parts.push(chunk.html);\n }\n flush();\n\n return parts.join(\"\\n\");\n}\n\n// Text or an image makes a chunk content; whitespace, entities that only decode\n// to whitespace and empty markup do not\nexport function hasContent(html: string): boolean {\n return / 0;\n}\n\n// --- reading a page's links, images and words ------------------------------\n\nexport function absolutize($: CheerioAPI, el: Element, baseUrl: string): void {\n $(el).find(\"[href]\").each((_, a) => {\n const href = $(a).attr(\"href\");\n if (href) $(a).attr(\"href\", new URL(href, baseUrl).toString());\n });\n $(el).find(\"[src]\").each((_, img) => {\n const src = $(img).attr(\"src\");\n if (src) $(img).attr(\"src\", new URL(src, baseUrl).toString());\n });\n}\n\n// An image as a feed item carries it: absolute, with its alt text, and free of\n// the layout styles of the page it came from\nexport function imageHtml(src: string, alt: string, baseUrl: string): string {\n const url = escapeHtml(new URL(src, baseUrl).toString());\n return ` $\"${\n$ `;\n}\n\n// --- headlines ------------------------------------------------------------\n//\n// An item a page gives no name of its own is named by the line it opens with:\n// its first sentence, cut where it runs past what a headline carries.\n\n// How long the headline of an item may run before it is cut at a word boundary\nconst TITLE_LENGTH = 90;\n\n// Where one sentence of a note ends: at terminal punctuation followed by the\n// capital, digit, quote or emoji that opens the next one\nexport const SENTENCE_END =\n /(?<=[.!?…])\\s+(?=[\\p{Lu}\\p{N}\\p{Extended_Pictographic}\"'“”‘’(\\[])/u;\n\n// A line that runs past the cap is cut at the last word boundary within it, with\n// an ellipsis marking what was left out\nexport function cutToWords(text: string): string {\n if (text.length <= TITLE_LENGTH) return text;\n\n const cut = text.slice(0, TITLE_LENGTH);\n const boundary = cut.search(/\\s+\\S$/);\n const words = boundary > 0 ? cut.slice(0, boundary) : cut;\n return `${words.replace(/[\\s,;:.!?…]+$/, \"\")}…`;\n}\n\n// --- word links ------------------------------------------------------------\n//\n// An entry that links to a name of its own (\"STRICH\", \"Vite\") wears that word\n// as a tag too, so a recurring ad or a project can be filtered out by the name\n// it promotes without the split knowing anything about it. Only a link of one\n// word qualifies: a link that reads as a phrase (\"Source code\") is prose, and a\n// phrase is no name to filter on.\n\n// A word opens with a letter or a digit and holds only the marks names carry\n// inside them (Next.js, fast-check, C++). A link that reads as a path or a call\n// (rsc.io/markdown, browser()), or as a run of punctuation or an emoji, is not a\n// word and gets no tag.\nconst WORD = /^[\\p{L}\\p{N}][\\p{L}\\p{N}._+'-]$/u;\n\nexport function singleWordLinkTags(contentHtml: string): string[] {\n const $ = load(contentHtml);\n const words = new Set();\n $(\"a[href]\").each((_, a) => {\n const text = normalizeWhitespace($(a).text());\n if (WORD.test(text)) words.add(text);\n });\n return [...words];\n}\n\n// --- helpers --------------------------------------------------------------\n\nexport function htmlToText(html: string): string {\n return normalizeWhitespace(load(html).text());\n}\n\nexport function escapeHtml(value: string): string {\n return value\n .replace(/&/g, \"&\")\n .replace(//g, \">\")\n .replace(/\"/g, \""\");\n}\n\nexport function normalizeWhitespace(value: string): string {\n return value.replace(/\\s+/g, \" \").trim();\n}\n",[],[253,312],"49b48542-cb56-4615-b10f-a49a6fff3fd9","openrouter.ts","1166edd4-586b-11f1-8c74-ee650bb23af1",55,["D",1788677504293],"import { feedEndpoint, fetchJson } from \"../feed.ts\";\n\ninterface OpenRouterModel {\n slug: string;\n name: string;\n description?: string;\n created_at: string;\n author?: string;\n author_display_name?: string;\n group?: string;\n context_length?: number;\n input_modalities?: string[];\n output_modalities?: string[];\n supports_reasoning?: boolean;\n reasoning_config?: {\n is_mandatory_reasoning?: boolean;\n supported_reasoning_efforts?: string[];\n default_reasoning_effort?: string;\n };\n knowledge_cutoff?: string | null;\n endpoint?: {\n provider_name?: string;\n is_free?: boolean;\n pricing?: {\n prompt?: string;\n completion?: string;\n input_cache_read?: string;\n display_pricing?: Array<{\n sku_label?: string;\n price?: string;\n displayMultiplier?: number;\n unitLabel?: string;\n }>;\n };\n };\n}\n\ninterface OpenRouterFeedItem {\n id: string;\n url: string;\n title: string;\n content_text: string;\n content_html: string;\n date_published: string;\n author: { name: string };\n tags: string[];\n}\n\n// Endpoint the /models page uses; `order=newest` sorts by release date and\n// `max_age_days` keeps the response to recent releases instead of all models.\nconst OPENROUTER_MODELS_URL =\n \"https://openrouter.ai/api/frontend/v1/models/find?order=newest&active=true&fmt=cards\";\n\nconst openRouterExpression = `{\n \"version\": $jsonFeedVersion,\n \"title\": \"OpenRouter AI Models\",\n \"home_page_url\": \"https://openrouter.ai/models\",\n \"feed_url\": $feedUrl,\n \"description\": \"Latest AI models from OpenRouter\",\n \"items\": items\n}`;\n\nexport const openRouterHandler = feedEndpoint({\n expression: openRouterExpression,\n loadSource: async (c) => {\n const days = parseDays(c.req.query(\"days\"));\n const raw = await fetchJson(\n `${OPENROUTER_MODELS_URL}&max_age_days=${days}`,\n ) as { data: { models: OpenRouterModel[] } };\n // Item bodies are composed in TypeScript since the metadata (pricing,\n // reasoning options, modalities) is easier to format and escape there.\n return {\n data: { items: dedupeVariants(raw.data.models).map(toFeedItem) },\n };\n },\n});\n\n// The catalog lists variants (batch/free) under the same slug, so different\n// variants would otherwise collide as feed item ids. Keep one item per model.\nfunction dedupeVariants(models: OpenRouterModel[]): OpenRouterModel[] {\n const bySlug = new Map();\n for (const model of models) {\n const group = bySlug.get(model.slug) ?? [];\n group.push(model);\n bySlug.set(model.slug, group);\n }\n return [...bySlug.values()].map((group) =>\n group.sort((a, b) => a.name.localeCompare(b.name))[0]\n );\n}\n\nfunction toFeedItem(model: OpenRouterModel): OpenRouterFeedItem {\n return {\n id: model.slug,\n url: `https://openrouter.ai/${model.slug}`,\n title: model.name,\n content_text: buildBodyText(model),\n content_html: buildBodyHtml(model),\n date_published: model.created_at,\n author: { name: displayName(model) },\n tags: buildTags(model),\n };\n}\n\nfunction buildBodyHtml(model: OpenRouterModel): string {\n const parts: string[] = [];\n if (model.description) {\n for (\n const paragraph of renderDescription(model.description).split(/\\n{2,}/)\n ) {\n if (paragraph.trim()) parts.push(`
${paragraph.trim()}
`);\n }\n }\n const rows = metadataRows(model);\n if (rows.length) {\n parts.push(\"
\");\n for (const [term, detail] of rows) {\n parts.push(`
${escapeHtml(term)}
${detail}
`);\n }\n parts.push(\"
\");\n }\n parts.push(\n `
View on OpenRouter
`,\n );\n return parts.join(\"\\n\");\n}\n\nfunction buildBodyText(model: OpenRouterModel): string {\n const lines: string[] = [];\n const description = toPlainText(model.description ?? \"\")\n .replace(/\\n{2,}/g, \"\\n\")\n .trim();\n if (description) lines.push(description);\n for (const [term, detail] of metadataRows(model)) {\n lines.push(`${term}: ${stripHtml(detail)}`);\n }\n lines.push(`Model page: https://openrouter.ai/${model.slug}`);\n return lines.join(\"\\n\");\n}\n\nfunction metadataRows(model: OpenRouterModel): Array<[string, string]> {\n const rows: Array<[string, string]> = [];\n const author = displayName(model);\n if (author !== \"Unknown author\") rows.push([\"Author\", escapeHtml(author)]);\n if (model.group) rows.push([\"Group\", escapeHtml(model.group)]);\n const context = formatContextLength(model.context_length);\n if (context) rows.push([\"Context\", escapeHtml(context)]);\n const input = formatModalities(model.input_modalities);\n const output = formatModalities(model.output_modalities);\n if (input) rows.push([\"Input\", escapeHtml(input)]);\n if (output) rows.push([\"Output\", escapeHtml(output)]);\n const reasoning = formatReasoning(model);\n if (reasoning) rows.push([\"Reasoning\", escapeHtml(reasoning)]);\n if (model.knowledge_cutoff) {\n rows.push([\"Knowledge cutoff\", escapeHtml(model.knowledge_cutoff)]);\n }\n const pricing = pricingDetail(model);\n if (pricing) rows.push([\"Pricing\", pricing]);\n return rows;\n}\n\nfunction displayName(model: OpenRouterModel): string {\n return model.author_display_name ?? model.author ?? \"Unknown author\";\n}\n\nfunction formatContextLength(tokens?: number): string | undefined {\n if (!tokens || tokens <= 0) return undefined;\n return `${tokens.toLocaleString(\"en-US\")} tokens`;\n}\n\nfunction formatModalities(modalities?: string[]): string | undefined {\n const list = (modalities ?? []).filter(Boolean).map((modality) =>\n modality.charAt(0).toUpperCase() + modality.slice(1)\n );\n return list.length ? list.join(\", \") : undefined;\n}\n\nfunction formatReasoning(model: OpenRouterModel): string | undefined {\n if (!model.supports_reasoning) return undefined;\n const config = model.reasoning_config;\n const parts = [\"Supported\"];\n if (config?.is_mandatory_reasoning) parts.push(\"mandatory\");\n if (config?.default_reasoning_effort) {\n parts.push(`default effort ${config.default_reasoning_effort}`);\n }\n if (config?.supported_reasoning_efforts?.length) {\n parts.push(`options: ${config.supported_reasoning_efforts.join(\", \")}`);\n }\n return parts.join(\" \\u00b7 \");\n}\n\nfunction pricingDetail(model: OpenRouterModel): string | undefined {\n const endpoint = model.endpoint;\n if (!endpoint) return undefined;\n const parts: string[] = [];\n if (endpoint.provider_name) parts.push(escapeHtml(endpoint.provider_name));\n if (endpoint.is_free) parts.push(\"Free\");\n const prices = pricingParts(model);\n if (prices.length) parts.push(escapeHtml(prices.join(\" \\u00b7 \")));\n return parts.length ? parts.join(\" \\u2014 \") : undefined;\n}\n\nfunction pricingParts(model: OpenRouterModel): string[] {\n const pricing = model.endpoint?.pricing;\n if (!pricing) return [];\n const display = pricing.display_pricing ?? [];\n if (display.length) {\n const seen = new Set();\n const out: string[] = [];\n for (const item of display) {\n const price = formatPrice(\n Number(item.price ?? \"\") (item.displayMultiplier ?? 1),\n );\n if (!price) continue;\n const label = PRICE_LABEL_OVERRIDES[item.sku_label ?? \"\"] ??\n (item.sku_label ?? \"\").trim();\n const unit = (item.unitLabel ?? \"\").trim();\n const text = `${label} ${price}${unit}`.trim();\n if (text && !seen.has(text)) {\n seen.add(text);\n out.push(text);\n }\n }\n return out;\n }\n const out: string[] = [];\n const prompt = tokenRate(pricing.prompt);\n const cached = tokenRate(pricing.input_cache_read);\n const completion = tokenRate(pricing.completion);\n if (prompt) out.push(`Input ${prompt}/M`);\n if (cached && cached !== prompt) out.push(`Cached input ${cached}/M`);\n if (completion) out.push(`Output ${completion}/M`);\n return out;\n}\n\n// Shorter labels for the rows OpenRouter's UI shows on the model cards.\nconst PRICE_LABEL_OVERRIDES: Record = {\n \"Input Price\": \"Input\",\n \"Output Price\": \"Output\",\n \"Cache Read\": \"Cached input\",\n \"Cache Read (5min)\": \"Cached input (5min)\",\n \"Cache Creation (5min)\": \"Cache write (5min)\",\n \"Cache Write\": \"Cache write\",\n \"Cache Write (1h)\": \"Cache write (1h)\",\n};\n\nfunction tokenRate(raw?: string): string | undefined {\n if (raw === undefined || raw === \"\") return undefined;\n return formatPrice(Number(raw) * 1_000_000);\n}\n\nfunction formatPrice(value: number): string | undefined {\n if (!Number.isFinite(value) || value <= 0) return undefined;\n return `$${Number(value.toPrecision(3))}`;\n}\n\nfunction buildTags(model: OpenRouterModel): string[] {\n const tags: string[] = [];\n if (model.group) tags.push(model.group);\n const modalities = joinModalities(model.input_modalities);\n if (modalities) tags.push(modalities);\n if (model.endpoint?.is_free) tags.push(\"free\");\n if (model.supports_reasoning) tags.push(\"reasoning\");\n return tags;\n}\n\nfunction joinModalities(modalities?: string[]): string | undefined {\n return modalities?.length ? modalities.join(\"+\") : undefined;\n}\n\n// Turn a small subset of markdown (links, code) into HTML and linkify bare URLs.\nfunction renderDescription(value: string): string {\n return escapeHtml(value)\n .replace(\n /(\\[([^\\]]+)\\]\$(https?:\\/\\/[^)\\s]+)\$|(https?:\\/\\/[^\\s<]+))/g,\n (_match, _markdown, label, markdownUrl, bareUrl) =>\n `${label ?? bareUrl}`,\n )\n .replace(/`([^`]+)`/g, \"`$1`\");\n}\n\nfunction toPlainText(value: string): string {\n return value\n .replace(/\\[([^\\]]+)\\]\$[^)]+\$/g, \"$1\")\n .replace(/`([^`]+)`/g, \"$1\");\n}\n\n// For the plain-text body: drop tags and decode entities from HTML rows.\nfunction stripHtml(value: string): string {\n return value.replace(/<[^>]+>/g, \"\")\n .replace(/&/g, \"&\")\n .replace(/</g, \"<\")\n .replace(/>/g, \">\")\n .replace(/"/g, '\"');\n}\n\nfunction escapeHtml(value: string): string {\n return value\n .replace(/&/g, \"&\")\n .replace(//g, \">\")\n .replace(/\"/g, \""\");\n}\n\nfunction parseDays(raw: string | undefined): number {\n const days = raw === undefined ? NaN : Number.parseInt(raw, 10);\n return Number.isFinite(days) && days >= 1 ? Math.min(days, 365) : 30;\n}\n",[],[253,320],"68e831d4-586c-11f1-b008-ee650bb23af1","luma.ts","11603516-586b-11f1-ab2b-ee650bb23af1",["D",1779735806469],"import ICAL from \"npm:ical.js@2.2.1\";\nimport { feedEndpoint, fetchJson, HttpError } from \"../feed.ts\";\n\nconst LUMA_DISCOVER_BOOTSTRAP_URL =\n \"https://api2.luma.com/discover/bootstrap-page\";\n\nexport interface LumaPlace {\n apiId: string;\n name: string;\n slug: string;\n timezone: string;\n eventCount: number;\n}\n\ninterface LumaBootstrapPlace {\n place?: {\n api_id?: string;\n name?: string;\n slug?: string;\n timezone?: string;\n };\n event_count?: number;\n num_events?: number;\n}\n\ninterface LumaBootstrapResponse {\n places?: LumaBootstrapPlace[];\n}\n\nconst lumaExpression = `{\n \"version\": $jsonFeedVersion,\n \"title\": calendarName ? calendarName : \"Luma Calendar Feed\",\n \"home_page_url\": calendarUrl ? calendarUrl : \"https://lu.ma\",\n \"feed_url\": $feedUrl,\n \"description\": calendarDescription,\n \"items\": events.{\n \"id\": uid,\n \"url\": lumaUrl ? lumaUrl : url,\n \"title\": shortDateTime & \" \" & summary,\n \"content_text\": description,\n \"summary\": description ? $substring(description, 0, 100) & \"...\" : null,\n \"date_published\": start,\n \"date_modified\": modified,\n \"author\": organizer ? { \"name\": organizer } : null,\n \"tags\": categories,\n \"attachments\": attachment ? [{ \"url\": attachment }] : null\n }\n}`;\n\nexport const lumaHandler = feedEndpoint({\n expression: lumaExpression,\n loadSource: async (c) => {\n const placeSlug = c.req.param(\"place\")?.replace(/\\.(rss|json)$/, \"\");\n if (!placeSlug) throw new HttpError(400, \"Missing Luma place\");\n\n const place = (await getLumaPlaces()).find((place) =>\n place.slug === placeSlug\n );\n if (!place) throw new HttpError(404, \"Luma place not found\");\n\n const timezone = c.req.query(\"timezone\") ?? place.timezone;\n\n const icalResponse = await fetch(\n `https://api.lu.ma/ics/get?entity=discover&id=${\n encodeURIComponent(place.apiId)\n }`,\n );\n if (!icalResponse.ok) {\n throw new Error(\n `Failed to fetch Luma calendar: ${icalResponse.status} ${icalResponse.statusText}`,\n );\n }\n\n return { data: parseLumaCalendar(await icalResponse.text(), timezone) };\n },\n});\n\nexport async function getLumaPlaces(): Promise {\n const data = await fetchJson(\n LUMA_DISCOVER_BOOTSTRAP_URL,\n ) as LumaBootstrapResponse;\n return (data.places ?? [])\n .map(({ place, event_count, num_events }) => ({\n apiId: place?.api_id ?? \"\",\n name: place?.name ?? \"\",\n slug: place?.slug ?? \"\",\n timezone: place?.timezone ?? \"UTC\",\n eventCount: event_count ?? num_events ?? 0,\n }))\n .filter((place) => place.apiId && place.name && place.slug)\n .sort((a, b) => a.name.localeCompare(b.name));\n}\n\nfunction parseLumaCalendar(icalData: string, timezone: string): unknown {\n const jcalData = ICAL.parse(icalData);\n const comp = new ICAL.Component(jcalData);\n const events = comp.getAllSubcomponents(\"vevent\").map((event) => {\n const description = stringValue(event.getFirstPropertyValue(\"description\"));\n const start = event.getFirstPropertyValue(\"dtstart\");\n const end = event.getFirstPropertyValue(\"dtend\");\n const modified = event.getFirstPropertyValue(\"last-modified\");\n\n return {\n uid: stringValue(event.getFirstPropertyValue(\"uid\")),\n summary: stringValue(event.getFirstPropertyValue(\"summary\")),\n description,\n start: toIsoDate(start),\n end: toIsoDate(end),\n modified: toIsoDate(modified),\n shortDateTime: toDate(start)\n ? formatShortDateTime(toDate(start)!, timezone)\n : \"\",\n location: stringValue(event.getFirstPropertyValue(\"location\")),\n url: stringValue(event.getFirstPropertyValue(\"url\")),\n lumaUrl: description ? extractLumaLink(description) : undefined,\n organizer: stringValue(event.getFirstPropertyValue(\"organizer\")),\n categories: arrayValue(event.getFirstPropertyValue(\"categories\")),\n attachment: stringValue(event.getFirstPropertyValue(\"attach\")),\n };\n });\n\n return {\n calendarName: stringValue(comp.getFirstPropertyValue(\"x-wr-calname\")),\n calendarUrl: stringValue(comp.getFirstPropertyValue(\"url\")),\n calendarDescription: stringValue(\n comp.getFirstPropertyValue(\"x-wr-caldesc\"),\n ),\n events,\n };\n}\n\nfunction extractLumaLink(text: string): string | undefined {\n return text.match(/https?:\\/\\/(?:www\\.)?luma\\.com\\/[^\\s<>()\"']+/)?.[0];\n}\n\nfunction formatShortDateTime(date: Date, timezone: string): string {\n return date.toLocaleString(\"en-US\", {\n timeZone: timezone,\n month: \"short\",\n day: \"numeric\",\n hour: \"numeric\",\n minute: \"2-digit\",\n weekday: \"short\",\n hour12: true,\n });\n}\n\nfunction toDate(value: unknown): Date | undefined {\n if (\n value && typeof value === \"object\" && \"toJSDate\" in value &&\n typeof value.toJSDate === \"function\"\n ) {\n return value.toJSDate();\n }\n return undefined;\n}\n\nfunction toIsoDate(value: unknown): string | undefined {\n return toDate(value)?.toISOString();\n}\n\nfunction stringValue(value: unknown): string | undefined {\n return typeof value === \"string\" && value.length > 0 ? value : undefined;\n}\n\nfunction arrayValue(value: unknown): string[] | undefined {\n if (Array.isArray(value)) return value.map(String);\n if (typeof value === \"string\" && value.length > 0) return [value];\n return undefined;\n}\n",[],[253,328],"20477bd4-5e24-11f1-8db3-1607ee4eb77e","lineageos.ts","99e18668-5872-11f1-9b36-ee650bb23af1",32,["D",1780364467840],"import { feedEndpoint, fetchJson } from \"../feed.ts\";\n\nconst LINEAGEOS_DEVICES_URL = \"https://wiki.lineageos.org/devices/\";\nconst LINEAGEOS_SEARCH_URL = \"https://wiki.lineageos.org/search.json\";\n\nconst lineageOsExpression = `{\n \"version\": $jsonFeedVersion,\n \"title\": \"LineageOS Devices\",\n \"home_page_url\": \"${LINEAGEOS_DEVICES_URL}\",\n \"feed_url\": $feedUrl,\n \"description\": \"List of supported devices for LineageOS\",\n \"items\": $map(\n $filter(entries, function($v) {\n $contains($v.url, \"/devices/\") and $not($contains($v.title, \"Info about\"))\n }),\n function($v) {\n {\n \"id\": $v.url,\n \"url\": \"https://wiki.lineageos.org\" & $v.url,\n \"title\": $v.title,\n \"date_published\": \"2024-01-01T00:00:00.000Z\",\n \"content_text\": $v.summary ? $v.summary : $v.title,\n \"content_html\": \"\" & ($v.summary ? $v.summary : $v.title)\n }\n }\n )\n}`;\n\nexport const lineageOsHandler = feedEndpoint({\n expression: lineageOsExpression,\n loadSource: async () => ({\n data: { entries: await fetchJson(LINEAGEOS_SEARCH_URL) },\n }),\n});\n",[],[253,335],"6972489c-586c-11f1-b008-ee650bb23af1","huggingFace.ts","114c7328-586b-11f1-8152-ee650bb23af1",19,["D",1779735807379],"import { load } from \"npm:cheerio@1.2.0\";\nimport { feedEndpoint, HttpError } from \"../feed.ts\";\n\nconst hfActivityExpression = `{\n \"version\": $jsonFeedVersion,\n \"title\": username & \" - Hugging Face Activity\",\n \"home_page_url\": profileUrl & \"/activity/all\",\n \"feed_url\": $feedUrl,\n \"description\": \"Recent activity from \" & username & \" on Hugging Face\",\n \"items\": activities.{\n \"id\": id,\n \"url\": items[0].url ? items[0].url : $$.profileUrl & \"/activity/all\",\n \"title\": action,\n \"content_html\": $join(items.(\"\" & name & \"\"), \"
\\n\"),\n \"date_published\": timestamp\n }\n}`;\n\nexport const huggingFaceHandler = feedEndpoint({\n expression: hfActivityExpression,\n loadSource: async (c) => {\n const username = c.req.param(\"usernameFile\")?.replace(/\\.(rss|json)$/, \"\");\n if (!username) throw new HttpError(400, \"Missing Hugging Face username\");\n return { data: await scrapeHuggingFaceActivity(username) };\n },\n});\n\nasync function scrapeHuggingFaceActivity(username: string): Promise {\n const profileUrl = `https://huggingface.co/${username}`;\n const response = await fetch(`${profileUrl}/activity/all`, {\n headers: {\n \"User-Agent\":\n \"Mozilla/5.0 (compatible; RSSGallery/1.0; +https://val.town)\",\n },\n });\n\n if (response.status === 404) {\n throw new HttpError(404, \"Hugging Face user not found\");\n }\n if (!response.ok) {\n throw new Error(\n `Failed to fetch Hugging Face activity: ${response.status} ${response.statusText}`,\n );\n }\n\n const $ = load(await response.text());\n const activities: Array<\n {\n action: string;\n id: string;\n items: Array<{ name: string; url: string }>;\n timestamp: string;\n }\n > = [];\n\n $(\".user-profile-content div.mb-7\").each((_i, group) => {\n const $group = $(group);\n const headerText = $group.children(\"div\").first().text().replace(\n /\\s+/g,\n \" \",\n ).trim();\n const { action, timeAgo } = parseActivityHeader(headerText);\n if (!action) return;\n\n const timestampValue = $group.find(\"time[datetime]\").first().attr(\n \"datetime\",\n );\n const timestamp = timestampValue\n ? new Date(timestampValue)\n : parseTimeAgo(timeAgo);\n const items = collectActivityItems($, $group, action);\n const timestampIso = timestamp.toISOString();\n const id = buildActivityId(\n username,\n action,\n items,\n timestampValue ? timestampIso : undefined,\n );\n activities.push({ action, id, items, timestamp: timestampIso });\n });\n\n return { username, profileUrl, activities };\n}\n\nfunction collectActivityItems(\n $: ReturnType,\n $group: ReturnType>,\n action: string,\n): Array<{ name: string; url: string }> {\n const items: Array<{ name: string; url: string }> = [];\n const seen = new Set();\n\n $group.find(\"article\").each((_i, article) => {\n const $article = $(article);\n const name = $article.find(\"h4.text-md\").first().text().trim();\n if (!name || seen.has(name)) return;\n\n const href = $article.find(\"a[href]\").toArray()\n .map((link) => $(link).attr(\"href\") ?? \"\")\n .find((href) => /^\\/[^/]+\\/[^/]+$/.test(href));\n\n seen.add(name);\n items.push({ name, url: toHuggingFaceUrl(href ?? `/${name}`) });\n });\n\n const repoMatch = action.match(/^New activity in (\\S+)/);\n if (repoMatch && !seen.has(repoMatch[1])) {\n seen.add(repoMatch[1]);\n items.unshift({\n name: repoMatch[1],\n url: toHuggingFaceUrl(`/${repoMatch[1]}`),\n });\n }\n\n $group.find(\"h4.overflow-hidden\").each((_i, titleElement) => {\n const name = $(titleElement).text().trim();\n if (!name || seen.has(name)) return;\n\n const href = $(titleElement).closest(\"a\").attr(\"href\");\n seen.add(name);\n items.push({\n name,\n url: toHuggingFaceUrl(href ?? `/${repoMatch?.[1] ?? \"\"}`),\n });\n });\n\n return items;\n}\n\nfunction buildActivityId(\n username: string,\n action: string,\n items: Array<{ name: string; url: string }>,\n exactTimestampIdPart?: string,\n): string {\n const itemDiscriminator = items\n .map((item) => item.url || item.name)\n .join(\"|\");\n const discriminator = exactTimestampIdPart ??\n (itemDiscriminator || \"activity\");\n\n return [\"hf-activity\", username, action, discriminator]\n .map(normalizeIdPart)\n .join(\":\");\n}\n\nfunction normalizeIdPart(value: string): string {\n return value.replace(/\\s+/g, \" \").trim();\n}\n\nfunction parseActivityHeader(\n text: string,\n): { action: string; timeAgo: string } {\n const match = text.match(\n /^(.+?)\\s+((?:about\\s+)?\\d+\\s+(?:second|minute|hour|day|week|month|year)s?\\s+ago|just now)$/i,\n );\n return match\n ? { action: match[1].trim(), timeAgo: match[2].trim() }\n : { action: \"\", timeAgo: \"\" };\n}\n\nfunction parseTimeAgo(timeAgo: string): Date {\n const cleaned = timeAgo.replace(/^about\\s+/, \"\");\n if (cleaned === \"just now\") return new Date();\n\n const match = cleaned.match(\n /(\\d+)\\s+(second|minute|hour|day|week|month|year)s?/,\n );\n if (!match) return new Date();\n\n const durationMs: Record = {\n second: 1_000,\n minute: 60_000,\n hour: 3_600_000,\n day: 86_400_000,\n week: 604_800_000,\n month: 2_592_000_000,\n year: 31_536_000_000,\n };\n\n return new Date(Date.now() - Number(match[1]) * durationMs[match[2]]);\n}\n\nfunction toHuggingFaceUrl(href: string): string {\n return href.startsWith(\"http\") ? href : `https://huggingface.co${href}`;\n}\n",[],[253,343],"6e982654-9a22-4db9-b983-4a61e1e8f654","hnDaily.ts","00cf79ac-b18d-11f1-8da4-1607ee4eb77e",64,["D",1789535458723],"import { load } from \"npm:cheerio@1.2.0\";\nimport { feedEndpoint, HttpError } from \"../feed.ts\";\n\n// Hacker News Daily (daemonology.net/hn-daily) posts one digest a day: the ten\n// highest-rated articles on Hacker News that day which have not appeared on any\n// previous Hacker News Daily. Its archive page carries the most recent digests\n// on one page, each a heading, a list of ten entries and a link to the page the\n// digest keeps. The page format has been stable for years, so the entries are\n// read straight off it and handed out one item per article, in the body shape\n// of hnrss.org items, each carrying the day it belongs to and the rank it held\n// that day.\nconst HN_DAILY_URL = \"https://www.daemonology.net/hn-daily/\";\n// The only icon Daemonology offers. It rides the feed metadata as the JSON Feed\n// icon, which jsonfeed-to-rss turns into the RSS channel image.\nconst HN_DAILY_ICON = \"https://www.daemonology.net/favicon.ico\";\n\nconst USER_AGENT =\n \"Mozilla/5.0 (compatible; RSSGallery/1.0; +https://val.town)\";\n\nconst hnDailyExpression = `{\n \"version\": $jsonFeedVersion,\n \"title\": \"HN Daily (decomposed)\",\n \"home_page_url\": \"${HN_DAILY_URL}\",\n \"feed_url\": $feedUrl,\n \"icon\": \"${HN_DAILY_ICON}\",\n \"description\": \"Every article of the recent Hacker News Daily digests as an individual item, carrying the day it belongs to and the rank it held that day, and, like hnrss.org feeds, linking to the article and to its comments. A digest holds the ten highest-rated articles on Hacker News of one day which have not appeared on any previous Hacker News Daily, and the day of each item links back to the digest it was split from.\",\n \"items\": items.{\n \"id\": id,\n \"url\": url,\n \"title\": title,\n \"content_html\": content_html,\n \"content_text\": content_text,\n \"date_published\": date_published\n }\n}`;\n\nexport const hnDailyHandler = feedEndpoint({\n expression: hnDailyExpression,\n loadSource: async () => {\n const items = decomposeArchive(await fetchArchive());\n if (items.length === 0) {\n throw new HttpError(502, `No articles found on ${HN_DAILY_URL}`);\n }\n return { data: { items } };\n },\n});\n\n/** One article of a digest, as the digest lists it */\nexport interface DailyEntry {\n title: string;\n url: string;\n /** The Hacker News discussion of the article, when the digest links one */\n commentsUrl?: string;\n /** The place the article held among the ten of its day */\n rank: number;\n}\n\n/** One day's digest, as the archive page spells it out */\nexport interface DailyDigest {\n /** The day the articles are drawn from, e.g. 2026-09-15 */\n date: string;\n /** The page the digest keeps for itself */\n url: string;\n entries: DailyEntry[];\n}\n\nexport interface FeedItem {\n id: string;\n url: string;\n title: string;\n content_html: string;\n content_text: string;\n date_published: string;\n}\n\n// The archive page holds the digests newest first, and the items follow that\n// order, all ten of a day together\nexport function decomposeArchive(html: string): FeedItem[] {\n return parseDigests(html).flatMap((digest) =>\n digest.entries.map((entry) => feedItem(digest, entry))\n );\n}\n\nconst DAY_HEADING = /^Daily Hacker News for (\\d{4}-\\d{2}-\\d{2})$/;\n\n// A digest is a heading (\"Daily Hacker News for 2026-09-15\") over the list of\n// its ten entries. The site's own headings and lists sit outside the content\n// column, and a list with no heading above it belongs to no digest.\nexport function parseDigests(\n html: string,\n pageUrl = HN_DAILY_URL,\n): DailyDigest[] {\n const $ = load(html);\n const digests: DailyDigest[] = [];\n\n $(\"div.content ul\").each((_, list) => {\n const heading = normalizeWhitespace($(list).prevAll(\"h2\").first().text());\n const date = DAY_HEADING.exec(heading)?.[1];\n if (!date) return;\n\n const entries: DailyEntry[] = [];\n $(list).children(\"li\").each((_, li) => {\n const $story = $(li).find(\".storylink a\").first();\n const title = normalizeWhitespace($story.text());\n const href = $story.attr(\"href\");\n if (!href || !title) return;\n\n // Comments are dressed as \".postlink\" since 2019 and as \".commentlink\"\n // before that\n const comments = $(li).find(\".postlink a, .commentlink a\").first().attr(\n \"href\",\n );\n\n entries.push({\n title,\n url: new URL(href, pageUrl).href,\n commentsUrl: comments ? new URL(comments, pageUrl).href : undefined,\n rank: entries.length + 1,\n });\n });\n\n if (entries.length > 0) {\n digests.push({\n date,\n url: new URL(`${date}.html`, pageUrl).href,\n entries,\n });\n }\n });\n\n return digests;\n}\n\n// An article of a digest as a feed item. No article is ever carried twice, by\n// its own day or by a later one, so the discussion is a name for it that holds\n// however far the archive page rolls on.\nfunction feedItem(digest: DailyDigest, entry: DailyEntry): FeedItem {\n const lines = bodyLines(digest, entry);\n return {\n id: entry.commentsUrl ?? entry.url,\n url: entry.url,\n title: entry.title,\n content_html: lines\n .map(({ label, value, url }) =>\n `
${label}: ${\n url\n ? `${escapeHtml(value)}`\n : escapeHtml(value)\n }
`\n )\n .join(\"\\n\"),\n content_text: lines\n .map(({ label, value }) => `${label}: ${value}`)\n .join(\"\\n\"),\n date_published: `${digest.date}T00:00:00.000Z`,\n };\n}\n\ninterface BodyLine {\n label: string;\n value: string;\n url?: string;\n}\n\n// The body of an hnrss.org item, with the day and the rank of the digest where\n// those feeds carry points and comment counts: neither is on the archive page\nfunction bodyLines(digest: DailyDigest, entry: DailyEntry): BodyLine[] {\n return [\n { label: \"Article URL\", value: entry.url, url: entry.url },\n entry.commentsUrl\n ? {\n label: \"Comments URL\",\n value: entry.commentsUrl,\n url: entry.commentsUrl,\n }\n : undefined,\n { label: \"Date\", value: digest.date, url: digest.url },\n { label: \"Rank\", value: String(entry.rank) },\n ].filter((line): line is BodyLine => Boolean(line));\n}\n\nasync function fetchArchive(): Promise {\n const response = await fetch(HN_DAILY_URL, {\n headers: { \"User-Agent\": USER_AGENT, \"Accept\": \"text/html\" },\n });\n if (!response.ok) {\n throw new Error(\n `Failed to fetch ${HN_DAILY_URL}: ${response.status} ${response.statusText}`,\n );\n }\n return await response.text();\n}\n\nfunction normalizeWhitespace(value: string): string {\n return value.replace(/\\s+/g, \" \").trim();\n}\n\nfunction escapeHtml(value: string): string {\n return value\n .replace(/&/g, \"&\")\n .replace(//g, \">\")\n .replace(/\"/g, \""\");\n}\n",[],[253,351],"5b784fbe-f00b-4605-9a92-34ab52375e1f","crowdSupply.ts","846aa1ca-a98a-11f1-90b0-1607ee4eb77e",43,["D",1788654781703],"import { load } from \"npm:cheerio@1.2.0\";\nimport { feedEndpoint, HttpError } from \"../feed.ts\";\n\nconst BASE_URL = \"https://www.crowdsupply.com\";\n\ninterface CrowdSupplyPage {\n title: string;\n description: string;\n path: string;\n}\n\nconst PAGES: Record = {\n prelaunch: {\n title: \"Crowd Supply Prelaunch\",\n description: \"Upcoming Crowd Supply campaigns that haven't launched yet\",\n path: \"/prelaunch\",\n },\n crowdfunding: {\n title: \"Crowd Supply Crowdfunding\",\n description: \"Active crowdfunding campaigns on Crowd Supply\",\n path: \"/crowdfunding\",\n },\n};\n\ninterface CrowdSupplyProject {\n id: string;\n url: string;\n title: string;\n description: string;\n imageUrl?: string;\n program?: string;\n status?: string;\n facts: Array<{ number: string; label: string }>;\n endsOn?: string;\n content_html: string;\n content_text: string;\n}\n\nconst crowdSupplyExpression = `{\n \"version\": $jsonFeedVersion,\n \"title\": pageTitle,\n \"home_page_url\": pageUrl,\n \"feed_url\": $feedUrl,\n \"description\": pageDescription,\n \"items\": items.{\n \"id\": id,\n \"url\": url,\n \"title\": title,\n \"content_html\": content_html,\n \"content_text\": content_text,\n \"date_published\": $now()\n }\n}`;\n\nexport const crowdSupplyHandler = feedEndpoint({\n expression: crowdSupplyExpression,\n loadSource: async (c) => {\n const pageKey = c.req.param(\"page\")?.replace(/\\.(rss|json)$/, \"\");\n if (!pageKey) throw new HttpError(400, \"Missing Crowd Supply page\");\n\n const page = PAGES[pageKey];\n if (!page) {\n throw new HttpError(\n 404,\n \"Crowd Supply page not found. Use: prelaunch or crowdfunding\",\n );\n }\n\n const items = await scrapeProjects(page);\n if (items.length === 0) {\n throw new HttpError(502, `No projects found on ${BASE_URL}${page.path}`);\n }\n\n return {\n data: {\n items,\n pageTitle: page.title,\n pageDescription: page.description,\n pageUrl: `${BASE_URL}${page.path}`,\n },\n };\n },\n});\n\nasync function scrapeProjects(page: CrowdSupplyPage): Promise<\n CrowdSupplyProject[]\n> {\n const url = `${BASE_URL}${page.path}`;\n const response = await fetch(url, {\n headers: {\n \"User-Agent\":\n \"Mozilla/5.0 (compatible; RSSGallery/1.0; +https://val.town)\",\n \"Accept\": \"text/html\",\n },\n });\n if (!response.ok) {\n throw new Error(\n `Failed to fetch ${url}: ${response.status} ${response.statusText}`,\n );\n }\n\n const $ = load(await response.text());\n const projects: CrowdSupplyProject[] = [];\n\n $(\"a.project-tile\").each((_, tile) => {\n const $tile = $(tile);\n const title = $tile.find(\"h3\").first().text().trim();\n const href = $tile.attr(\"href\") ?? \"\";\n if (!title || !href) return;\n\n const description = $tile.find(\".project-tile-overview p\").first().text()\n .trim();\n const imageEl = $tile.find(\"img\").first();\n const imageSrc = imageEl.attr(\"src\") ?? \"\";\n const imageAlt = imageEl.attr(\"alt\") ?? title;\n\n const status = [\n $tile.find(\".status-progress\").first().text(),\n $tile.find(\".status-bar\").first().children(\"div\").first().text(),\n ].map((text) => text.trim()).find((text) => text.length > 0);\n\n const program = normalizeWhitespace(\n $tile.find(\".project-tile-program\").text(),\n );\n\n const facts: Array<{ number: string; label: string }> = [];\n $tile.find(\".fact\").each((_, fact) => {\n const number = normalizeWhitespace($(fact).find(\".fact-number\").text());\n const label = normalizeWhitespace($(fact).find(\".fact-label\").text());\n if (number || label) facts.push({ number, label });\n });\n\n const endsOn = formatEndsOn($tile.find(\"[data-countdown-to]\").attr(\n \"data-countdown-to\",\n ));\n\n projects.push({\n id: $tile.attr(\"data-project-id\") ?? `${BASE_URL}${href}`,\n url: `${BASE_URL}${href}`,\n title,\n description,\n imageUrl: imageSrc ? `${BASE_URL}${imageSrc}` : undefined,\n program,\n status,\n facts,\n endsOn,\n content_html: buildContentHtml({\n title,\n description,\n imageUrl: imageSrc ? `${BASE_URL}${imageSrc}` : undefined,\n imageAlt,\n program,\n status,\n facts,\n endsOn,\n url: `${BASE_URL}${href}`,\n }),\n content_text: buildContentText({\n title,\n description,\n program,\n status,\n facts,\n endsOn,\n }),\n });\n });\n\n return projects;\n}\n\nfunction buildContentHtml(project: {\n title: string;\n description: string;\n imageUrl?: string;\n imageAlt: string;\n program?: string;\n status?: string;\n facts: Array<{ number: string; label: string }>;\n endsOn?: string;\n url: string;\n}): string {\n const parts: string[] = [];\n if (project.imageUrl) {\n parts.push(\n ` $\"${escapeHtml(project.imageAlt)}\"$ `,\n );\n }\n if (project.status) {\n parts.push(`
${escapeHtml(project.status)}
`);\n }\n if (project.description) {\n parts.push(`
${escapeHtml(project.description)}
`);\n }\n if (project.program) {\n parts.push(programToHtml(project.program));\n }\n const factText = project.facts\n .map((fact) =>\n escapeHtml([fact.number, fact.label].filter(Boolean).join(\" \"))\n )\n .filter(Boolean)\n .join(\" · \");\n const detailLines = [\n factText,\n project.endsOn ? `Ends ${escapeHtml(project.endsOn)}` : \"\",\n ].filter(Boolean);\n if (detailLines.length > 0) {\n parts.push(`
${detailLines.join(\"
\")}
`);\n }\n parts.push(\n `
View project
`,\n );\n return parts.join(\"\");\n}\n\nfunction buildContentText(project: {\n title: string;\n description: string;\n program?: string;\n status?: string;\n facts: Array<{ number: string; label: string }>;\n endsOn?: string;\n}): string {\n const factText = project.facts\n .map((fact) => [fact.number, fact.label].filter(Boolean).join(\" \"))\n .join(\" | \");\n return [\n project.status,\n project.description,\n project.program,\n factText,\n project.endsOn ? `Ends ${project.endsOn}` : \"\",\n ].filter(Boolean).join(\" | \");\n}\n\n// \"Part of PCBWay Assembly Hub\" -> emphasizes the program name\nfunction programToHtml(program: string): string {\n const prefix = \"Part of\";\n if (program.startsWith(prefix)) {\n const rest = program.slice(prefix.length).trim();\n return `
${escapeHtml(prefix)} ${escapeHtml(rest)}
`;\n }\n return `
${escapeHtml(program)}
`;\n}\n\nfunction formatEndsOn(value: string | undefined): string | undefined {\n if (!value) return undefined;\n const date = new Date(value);\n if (Number.isNaN(date.getTime())) return undefined;\n return date.toISOString().slice(0, 10);\n}\n\nfunction normalizeWhitespace(value: string): string {\n return value.replace(/\\s+/g, \" \").trim();\n}\n\nfunction escapeHtml(value: string): string {\n return value\n .replace(/&/g, \"&\")\n .replace(//g, \">\")\n .replace(/\"/g, \""\");\n}",[],[253,359],"0e4ee765-388e-4e99-8d4a-f1fff577959d","cooperPress.ts","97a2e1f6-b0c6-11f1-b20f-1607ee4eb77e",72,["D",1789538342848],"import { type CheerioAPI, contains, load } from \"npm:cheerio@1.2.0\";\nimport { type Element, isTag } from \"npm:domhandler@5.0.3\";\nimport { feedEndpoint, HttpError } from \"../feed.ts\";\nimport {\n absolutize,\n chunksOf,\n escapeHtml,\n type FetchedPage,\n fetchPage,\n hasContent,\n htmlToText,\n imageHtml,\n joinChunks,\n normalizeWhitespace,\n singleWordLinkTags,\n} from \"./pageDrain.ts\";\n\n// The Cooper Press newsletters (JavaScript Weekly, Frontend Focus, Node Weekly,\n// React Status, Postgres Weekly, Go Weekly and Ruby Weekly) each publish one\n// page per issue that holds every link they gathered that week. These feeds\n// pull the latest issue apart into one item per entry, so an issue can be\n// skimmed section by section. Whatever the split does not recognize is offered\n// as one closing item, so that a change of editorial style on their side cannot\n// drop content from the feed. A page that is no issue of the publication at all\n// — a firewall's block screen, a parking page, an error document — is never\n// read as one: it reaches the feed as a single item noting the failed fetch.\nexport interface Publication {\n title: string;\n site: string;\n description: string;\n}\n\nexport const cooperPressPublications: Record = {\n javascriptweekly: {\n title: \"JavaScript Weekly\",\n site: \"javascriptweekly.com\",\n description: \"Every entry of the latest JavaScript Weekly issue\",\n },\n frontendfocus: {\n title: \"Frontend Focus\",\n site: \"frontendfoc.us\",\n description: \"Every entry of the latest Frontend Focus issue\",\n },\n nodeweekly: {\n title: \"Node Weekly\",\n site: \"nodeweekly.com\",\n description: \"Every entry of the latest Node Weekly issue\",\n },\n reactstatus: {\n title: \"React Status\",\n site: \"react.statuscode.com\",\n description: \"Every entry of the latest React Status issue\",\n },\n postgresweekly: {\n title: \"Postgres Weekly\",\n site: \"postgresweekly.com\",\n description: \"Every entry of the latest Postgres Weekly issue\",\n },\n golangweekly: {\n title: \"Go Weekly\",\n site: \"golangweekly.com\",\n description: \"Every entry of the latest Go Weekly issue\",\n },\n rubyweekly: {\n title: \"Ruby Weekly\",\n site: \"rubyweekly.com\",\n description: \"Every entry of the latest Ruby Weekly issue\",\n },\n};\n\nconst cooperPressExpression = `{\n \"version\": $jsonFeedVersion,\n \"title\": feedTitle,\n \"home_page_url\": homePageUrl,\n \"feed_url\": $feedUrl,\n \"description\": feedDescription,\n \"items\": items.{\n \"id\": id,\n \"url\": url,\n \"title\": title,\n \"content_html\": content_html,\n \"content_text\": content_text,\n \"date_published\": date_published,\n \"tags\": tags\n }\n}`;\n\nexport const cooperPressHandler = feedEndpoint({\n expression: cooperPressExpression,\n loadSource: async (c) => {\n const slug = c.req.param(\"publication\")?.replace(/\\.(rss|json)$/, \"\");\n if (!slug) throw new HttpError(400, \"Missing Cooper Press publication\");\n\n const publication = cooperPressPublications[slug];\n if (!publication) {\n throw new HttpError(\n 404,\n `Unknown publication. Use: ${\n Object.keys(cooperPressPublications).join(\", \")\n }`,\n );\n }\n\n const page = await fetchLatestPage(publication);\n\n return {\n data: {\n feedTitle: `${publication.title} (decomposed)`,\n // The feed holds the latest issue, but it is about the publication as a\n // whole, so it links to the publication's own home page rather than to\n // the issue it currently carries. A reader who wants that issue finds\n // it at the bottom of every item\n homePageUrl: `https://${publication.site}/`,\n feedDescription:\n `${publication.description}, split into one item per section entry, each tagged with the section it appears under, any sponsorship and every single word it links to, plus a closing item for any content the split did not recognize and a single item noting the failed fetch where the publication serves no issue at all`,\n items: latestIssueItems(page, publication),\n },\n };\n },\n});\n\ninterface CooperPressIssue {\n url: string;\n html: string;\n number?: number;\n date?: string;\n}\n\nexport interface FeedItem {\n id: string;\n url?: string;\n title: string;\n content_html: string;\n content_text: string;\n date_published?: string;\n tags?: string[];\n}\n\n// An entry as the split reads it off the page: the fields of the feed item it\n// becomes, plus the section it appears under, which labels and tags the item\ninterface Entry {\n title: string;\n url?: string;\n contentHtml: string;\n section?: string;\n /** Tags the entry wears besides the section it appears under */\n tags?: string[];\n}\n\n/** Why no issue could be read, and what answered in its place */\nexport type FetchFailure =\n | { kind: \"unreachable\"; url: string; error: string }\n | {\n kind: \"http-error\";\n url: string;\n status: number;\n statusText?: string;\n /** The title of the page that came with the error, when it has one */\n pageTitle?: string;\n }\n | { kind: \"not-an-issue\"; url: string; pageTitle?: string };\n\n/** The latest issue of a publication, or the failure that stands in for it /\nexport type LatestIssue =\n | ({ kind: \"issue\" } & CooperPressIssue)\n | { kind: \"failure\"; failure: FetchFailure };\n\n// /latest redirects to the issue of the moment, and the response URL is the\n// canonical issue URL the feed links back to\nfunction fetchLatestPage(pub: Publication): Promise {\n return fetchPage(`https://${pub.site}/latest`);\n}\n\n// --- reading what /latest served ------------------------------------------\n//\n// The split reads one issue of a publication. A change of host, or a firewall\n// put in front of the site, can leave /latest serving something else entirely:\n// a Cloudflare block or challenge screen, a parking page, a generic error\n// document. None of that is an issue, so none of it is decomposed. The entries\n// such a page holds are not entries of the publication, and reading them would\n// hand the feed a block screen's prose dressed up as an issue. The feed asks\n// the page to be an issue of the publication it was asked for first, and notes\n// the failure as one item where it is not.\n\n// What makes a page an issue of a publication is that it names the publication\n// in its document title (\"Issue #801: When 8 chunks weigh more than 355 —\n// JavaScript Weekly\"). Both of Cooper Press's templates do, and the name is the\n// page's own identity rather than a piece of its layout, so it outlives a\n// change of markup. What it does not survive is being served another page: a\n// block screen, a parking page or another publication's issue names that page\n// instead, and so is no issue of this one.\n//\n// The guard deliberately stops there, at the page's name: whether an issue can\n// be decomposed is another matter, and an issue the split cannot read is still\n// offered whole as its closing item. Only a page that is no issue at all is\n// reported as a failure.\nexport function readLatestIssue(\n fetched: FetchedPage,\n pub: Publication,\n): LatestIssue {\n if (fetched.kind === \"error\") {\n return {\n kind: \"failure\",\n failure: { kind: \"unreachable\", url: fetched.url, error: fetched.error },\n };\n }\n\n const pageTitle = normalizeWhitespace(\n load(fetched.html)(\"title\").first().text(),\n ) || undefined;\n\n if (fetched.status < 200 || fetched.status >= 300) {\n return {\n kind: \"failure\",\n failure: {\n kind: \"http-error\",\n url: fetched.url,\n status: fetched.status,\n statusText: fetched.statusText,\n pageTitle,\n },\n };\n }\n\n if (!pageTitle || !namesPublication(pageTitle, pub)) {\n return {\n kind: \"failure\",\n failure: { kind: \"not-an-issue\", url: fetched.url, pageTitle },\n };\n }\n\n return {\n kind: \"issue\",\n url: fetched.url,\n html: fetched.html,\n ...parseIssueMeta(fetched.html),\n };\n}\n\n// The publication's name as a name of its own, so that the \"Go Weekly\" inside a\n// longer publication's name answers for that publication and no other\nfunction namesPublication(title: string, pub: Publication): boolean {\n const name = pub.title.replace(/[.+?^${}()|[\\]\\\\]/g, \"\\\\$&\");\n return new RegExp(`\\\\b${name}\\\\b`, \"i\").test(title);\n}\n\n// The items the feed carries for the page /latest served: the entries of the\n// latest issue, or the one item that notes why there are none\nexport function latestIssueItems(\n fetched: FetchedPage,\n pub: Publication,\n): FeedItem[] {\n const latest = readLatestIssue(fetched, pub);\n return latest.kind === \"issue\"\n ? decomposeIssue(latest.html, latest)\n : [fetchFailureItem(pub, latest.failure)];\n}\n\n// The tag of the item that stands in for an issue the publication did not serve\nexport const UNAVAILABLE_TAG = \"Unavailable\";\n\nconst UNAVAILABLE_TITLE = \"Latest issue unavailable\";\n\n// The one item the feed carries where no issue could be read. It holds no\n// sign-off: there is no issue to read on. It names itself and what answered in\n// place of the issue, so that a reader who finds no entries learns why.\nfunction fetchFailureItem(pub: Publication, failure: FetchFailure): FeedItem {\n const url = failure.url;\n const lead = `Fetching the latest ${pub.title} issue failed.`;\n const link = `${escapeHtml(url)}`;\n const contentHtml = `
${escapeHtml(lead)} ${link} ${\n escapeHtml(failureAnswer(failure))\n }.
`;\n\n return {\n // One note per URL, however often the failure is reported\n id: `${url}#unavailable`,\n url,\n title: UNAVAILABLE_TITLE,\n content_html: contentHtml,\n content_text: htmlToText(contentHtml),\n tags: [UNAVAILABLE_TAG],\n };\n}\n\n// What answered in place of the issue, e.g. \"answered with HTTP 403 Forbidden\"\nfunction failureAnswer(failure: FetchFailure): string {\n switch (failure.kind) {\n case \"unreachable\":\n return `could not be reached (${failure.error})`;\n case \"http-error\":\n return `answered with HTTP ${failure.status}${\n failure.statusText ? ` ${failure.statusText}` : \"\"\n }${pageTitleNote(failure.pageTitle)}`;\n case \"not-an-issue\":\n return failure.pageTitle\n ? `answered with the page \"${failure.pageTitle}\" instead of an issue of the publication`\n : \"answered with something other than an issue of the publication\";\n }\n}\n\nfunction pageTitleNote(pageTitle?: string): string {\n return pageTitle ? ` (\"${pageTitle}\")` : \"\";\n}\n\n// Every issue names itself (\"#793 — July 7, 2026\") somewhere in the content, but\n// the surrounding markup differs between publications\nexport function parseIssueMeta(html: string): {\n number?: number;\n date?: string;\n} {\n const $ = load(html);\n const body = normalizeWhitespace($(\"#content\").text() || $(\"body\").text());\n const match = /#\\u200b?(\\d+)\\s[—-]\\s([A-Z][a-z]+ \\d{1,2}, \\d{4})/.exec(\n body,\n );\n if (!match) return {};\n\n const date = new Date(`${match[2]} UTC`);\n return {\n number: Number(match[1]),\n date: Number.isNaN(date.getTime()) ? undefined : date.toISOString(),\n };\n}\n\n// --- issue decomposition --------------------------------------------------\n//\n// The split drains the page of one issue: one pass over its elements that reads\n// each entry, hands it to the feed as an item and takes the elements the item\n// holds off the page. Whatever the pass leaves behind is the content of the\n// closing item.\n\n// Parts of the page that never hold editorial content: the site's own furniture\nconst FURNITURE_REGIONS = [\n \".el-splitbar\",\n \".norss\",\n \".el-masthead\",\n \".masthead\",\n \".footer\",\n \".preheader\",\n \"style\",\n \"script\",\n].join(\",\");\n\nconst ITEM_CONTAINERS = \"table.el-item, table.miniitem\";\n\n// Label of the item that carries whatever the split did not recognize\nexport const NON_DECOMPOSED_TAG = \"Non-decomposed\";\n\n// Tag of an entry Cooper Press was paid to place. A sponsor says so in the\n// byline of its entry, and a sponsored classified ad says so in the element the\n// issue wraps it in.\nexport const SPONSORED_TAG = \"Sponsored\";\n\nconst SPONSOR_BADGE = \".tag-sponsor\";\n// The two signals live in different tiers and never overlap: the badge dresses\n// the featured sponsor boxes, the element wraps an ad-server sponsored\n// classified. A classified without that element is a plain classified ad, i.e.\n// holding the classifieds slot is not by itself a sponsorship.\nconst SPONSOR_MARKERS = `${SPONSOR_BADGE}, classified[tags~=\"sponsor\"]`;\n\nconst EMOJI_START =\n /^[\\p{Extended_Pictographic}\\p{Emoji_Presentation}\\u2600-\\u27BF\\u2B00-\\u2BFF\\uFE0F]/u;\nconst EMOJI_ONLY =\n /^[\\p{Extended_Pictographic}\\p{Emoji_Presentation}\\u2600-\\u27BF\\u2B00-\\u2BFF\\uFE0F\\s]+$/u;\n\nexport interface IssueRef {\n url: string;\n date?: string;\n number?: number;\n}\n\nexport function decomposeIssue(html: string, issue: IssueRef): FeedItem[] {\n const $ = load(html);\n const content = $(\"#content\").first();\n const scope = (content.length ? content : $(\"body\")).first().get(0)!;\n\n const items = [...drainIssue($, scope, issue)];\n\n // By now an item holds all the content it will carry (its own, the notes the\n // drain attached to it and the issue's banner), so the words it links to can\n // join its tags last, and the issue the entry was split out of can sign off\n // below everything else\n const signOff = issueSignOff(issue);\n for (const item of items) {\n const tags = [...(item.tags ?? [])];\n for (const word of singleWordLinkTags(item.content_html)) {\n if (!tags.includes(word)) tags.push(word);\n }\n item.tags = tags.length > 0 ? tags : undefined;\n item.content_html += `\\n${signOff.html}`;\n item.content_text = item.content_text\n ? `${item.content_text}\\n${signOff.text}`\n : signOff.text;\n }\n\n return items;\n}\n\n// The line every item of an issue signs off with: the entry belongs to an issue\n// as a whole, and the issue carries content the split files elsewhere, so a\n// reader is handed the one link that holds all of it\nfunction issueSignOff(issue: IssueRef): { html: string; text: string } {\n const label = issue.number\n ? `Read the full issue #${issue.number}`\n : \"Read the full issue\";\n return {\n html: `
${label}
`,\n text: label,\n };\n}\n\n// An item of an issue, numbered by its place among the issue's items\nfunction feedItem(\n issue: IssueRef,\n index: number,\n fields: { title: string; url?: string; contentHtml: string; tags?: string[] },\n): FeedItem {\n return {\n id: `${issue.url}#${index}`,\n url: fields.url,\n title: fields.title,\n content_html: fields.contentHtml,\n content_text: htmlToText(fields.contentHtml),\n date_published: issue.date,\n tags: fields.tags,\n };\n}\n\n// The drain loop over the page of one issue: it hands out one item per entry it\n// reads, in page order, and closes with the item that holds what it did not\n// read. An item takes the elements it holds off the page as it is handed out, so\n// the closing item is made of what no item before it claimed.\nfunction* drainIssue(\n $: CheerioAPI,\n scope: Element,\n issue: IssueRef,\n): Generator {\n const items: FeedItem[] = [];\n // The section currently in effect. Compact list markers (\"IN BRIEF:\") and the\n // classifieds box only label their own block, so they carry a scope.\n let section:\n | { name: string; block?: Element; heading: Element; story?: boolean }\n | undefined;\n const classifiedsSeen = new Set();\n // A run of prose between entries is one item; it has no headline of its own,\n // so it borrows the first link in the run\n let pendingProse:\n | {\n section?: string;\n link?: { title: string; url?: string };\n elements: Element[];\n parts: string[];\n }\n | undefined;\n // The image at the head of the issue heads the issue as a whole, so it is held\n // back to sign off every item. Every other image introduces the entry below\n // it, so it is held until that entry is read.\n let issueImage: { image: string; el: Element } | undefined;\n let pendingImages: Array<{ image: string; el: Element }> = [];\n\n // Taking an element off the page is what claiming it means: what an item\n // holds, the closing item does not offer a second time\n const take = (elements: Element[]) => {\n for (const el of elements) $(el).remove();\n };\n\n // The images held for an entry lead its content\n const takeImages = (): string[] => {\n const images = pendingImages.map((pending) => pending.image);\n take(pendingImages.map((pending) => pending.el));\n pendingImages = [];\n return images;\n };\n\n // A note of content that belongs to the entry above it joins that entry\n const appendNote = (into: FeedItem, note: string) => {\n into.content_html += `\\n${note}`;\n into.content_text += `\\n${htmlToText(note)}`;\n };\n\n // Folds an element that belongs to the entry above it into that entry\n const foldInto = (into: FeedItem, el: Element) => {\n absolutize($, el, issue.url);\n const note = $.html($(el)) ?? \"\";\n take([el]);\n appendNote(into, note);\n };\n\n const sectionFor = (el: Element): string | undefined => {\n if (!section) return undefined;\n if (section.block && !$(el).parents().toArray().includes(section.block)) {\n return undefined;\n }\n return section.name;\n };\n\n // Hands one entry to the feed. The heading of a section is provided by the\n // entries that wear its name, so a heading no entry ever wears is not taken\n // and reaches the closing item instead.\n function* pushItem(entry: Entry): Generator {\n const label = entry.section && entry.section !== entry.title\n ? entry.section\n : undefined;\n if (entry.section && section) take([section.heading]);\n // The images held for this entry lead its body\n const contentHtml = [...takeImages(), entry.contentHtml]\n .filter(Boolean)\n .join(\"\\n\");\n // The section names the entry first; the tags it wears follow\n const tags = [label, ...(entry.tags ?? [])]\n .filter((tag): tag is string => Boolean(tag));\n const item = feedItem(issue, items.length, {\n ...entry,\n title: label ? `${label}: ${entry.title}` : entry.title,\n contentHtml,\n tags: tags.length > 0 ? tags : undefined,\n });\n items.push(item);\n yield item;\n }\n\n // Reads an element as the entry it presents: the element as a whole, or the\n // parts of it an item holds when the box around them holds content of its own\n function* emit(\n el: Element,\n entry: Omit,\n /** Parts of the entry the item holds; defaults to the entry as a whole. /\n held?: Element[],\n ): Generator {\n // The section is read off the page before the entry leaves it\n const worn = sectionFor(el);\n take(held ?? [el]);\n yield pushItem({ ...entry, section: worn });\n }\n\n // Hands the run of prose read since the last entry to the feed\n function* flushProse(): Generator {\n if (!pendingProse) return;\n const { section: proseSection, link, elements, parts } = pendingProse;\n pendingProse = undefined;\n\n // A run with no link of its own has no headline to offer, so its text\n // stays with the entry above it (footnotes, closing notes, and the like)\n if (!link) {\n const previous = items[items.length - 1];\n if (previous && sectionTag(previous) === proseSection) {\n take(elements);\n appendNote(previous, parts.join(\"\\n\"));\n }\n return;\n }\n\n const tags = sponsorTags($, ...elements);\n take(elements);\n yield* pushItem({\n title: link.title,\n url: link.url,\n contentHtml: parts.join(\"\\n\"),\n section: proseSection,\n tags,\n });\n }\n\n // Adds the parts read since the last entry to the run of prose open on the\n // page, opening a run for them where the run they belong to has ended: a list\n // of release highlights closes the story above it, and no entry ever broke\n // the two apart\n const prose = (\n fields: { section?: string; link?: { title: string; url?: string } },\n elements: Element[],\n ) => {\n if (!pendingProse) {\n pendingProse = {\n section: fields.section,\n link: fields.link,\n elements: [],\n parts: [],\n };\n }\n pendingProse.elements.push(...elements);\n pendingProse.parts.push(...elements.map((el) => $.html($(el)) ?? \"\"));\n };\n\n for (const el of $(scope).find(\"\").toArray()) {\n // Whatever an item above took off the page is not read a second time\n if (!contains(scope, el)) continue;\n\n const $el = $(el);\n if ($el.parents(FURNITURE_REGIONS).length > 0) continue;\n if ($el.parents(ITEM_CONTAINERS).length > 0) continue;\n\n // Cooper Press puts the image an entry leads with above it, in a block of\n // its own rather than inside a paragraph\n if (el.tagName === \"img\" && isHeroImage($, el, scope)) {\n const image = heroImageHtml($, el, issue.url);\n if (image) {\n if (items.length === 0 && !issueImage) issueImage = { image, el };\n else pendingImages.push({ image, el });\n }\n continue;\n }\n\n const heading = detectHeading($, el);\n if (heading) {\n yield flushProse();\n const classifieds = $el.closest(\"table.classifieds\").get(0) as\n | Element\n | undefined;\n // Only the first heading of the classifieds box names a section; the\n // rest are titles of single ads\n if (classifieds && classifiedsSeen.has(classifieds)) continue;\n if (classifieds) classifiedsSeen.add(classifieds);\n const name = cleanHeading(heading.text);\n if (name) {\n section = {\n name,\n block: classifieds ?? heading.block,\n heading: el,\n story: heading.story,\n };\n }\n continue;\n }\n\n // Callout boxes (\"💡 ...\") expand on the entry above them\n if (\n (el.tagName === \"p\" || el.tagName === \"li\") &&\n $el.closest(\"table.el-content.blue\").length > 0\n ) {\n yield* flushProse();\n const previous = items[items.length - 1];\n if (previous) {\n foldInto(previous, el);\n continue;\n }\n }\n\n // A list no item of which links anywhere has no headline to offer either, so\n // it reads as prose of the entry above it rather than as entries of its own\n if (\n (el.tagName === \"ul\" || el.tagName === \"ol\") &&\n $el.find(\"a[href]\").length === 0\n ) {\n prose({ section: sectionFor(el), link: findEntryLink($, el) }, [el]);\n continue;\n }\n\n if ($el.is(ITEM_CONTAINERS)) {\n const link = findEntryLink($, el);\n if (!link) continue;\n yield* flushProse();\n // A story is told with prose, and the entry box that follows brings the\n // next one, so a centered title names no further than the story it opened\n if (section?.story) section = undefined;\n absolutize($, el, issue.url);\n // The badge a paid entry wears is the ad's own furniture: its meaning\n // travels as a tag, so it does not stay in the entry's content\n const tags = sponsorTags($, el);\n if (tags.length > 0) $el.find(SPONSOR_BADGE).remove();\n const held = [\"p.desc\", \"p.name\"]\n .map((part) => $el.find(part).first())\n .filter((part) => part.length > 0)\n .map((part) => part.get(0)!);\n\n if (held.length === 0) {\n yield* emit(el, { ...link, contentHtml: $.html($el), tags });\n continue;\n }\n\n // The logo an entry box decorates itself with is the ad's own furniture,\n // not editorial content, so it is accounted for alongside the ad\n yield* emit(\n el,\n {\n ...link,\n contentHtml: held.map((part) => $.html(part)).join(\"\\n\"),\n tags,\n },\n [...held, ...$el.find(\"img\").toArray()],\n );\n continue;\n }\n\n // Classifieds ads written as plain paragraphs are ruled off from each other\n if (el.tagName === \"hr\") {\n yield* flushProse();\n continue;\n }\n\n if (el.tagName === \"p\" || el.tagName === \"li\") {\n // List items are emitted whole, so skip the paragraphs inside them\n if (el.tagName === \"p\" && $el.parents(\"li\").length > 0) continue;\n\n if (el.tagName === \"p\" && !isEntryParagraph($, el)) {\n prose({ section: sectionFor(el), link: findEntryLink($, el) }, [el]);\n continue;\n }\n\n const link = findEntryLink($, el);\n if (!link) continue;\n yield* flushProse();\n absolutize($, el, issue.url);\n yield* emit(el, {\n ...link,\n contentHtml: $.html($el),\n tags: sponsorTags($, el),\n });\n }\n }\n\n yield* flushProse();\n\n // An image left over has no entry below it to lead, so it stays with the\n // entry above instead of being dropped\n const last = items[items.length - 1];\n if (last && pendingImages.length > 0) {\n last.content_html += `\\n${takeImages().join(\"\\n\")}`;\n }\n\n // The headline image is the issue's own banner, so it signs off every item.\n // An issue whose entries the split could not read at all keeps its images for\n // the closing item instead.\n const banner = items.length > 0 ? issueImage : undefined;\n if (banner) take([banner.el]);\n\n // What the drain did not take off the page closes the issue as one item\n const closing = leftoverItem($, scope, issue, items.length);\n if (closing) items.push(closing);\n if (banner) {\n for (const item of items) item.content_html += `\\n${banner.image}`;\n }\n if (closing) yield closing;\n}\n\n// --- closing item ---------------------------------------------------------\n//\n// The drain above only reads the entry styles Cooper Press use today, so a style\n// they add tomorrow would drop its content from the feed. Whatever no item of an\n// issue took off the page is serialized as plain feed HTML into one closing item\n// — or comes back undefined when nothing readable is left, which is the usual\n// outcome for a page whose entries were all read. See pageDrain.ts for what\n// that serialization keeps and what it drops.\n\nfunction leftoverItem(\n $: CheerioAPI,\n scope: Element,\n issue: IssueRef,\n /** The place the closing item takes among the items of its issue */\n index: number,\n): FeedItem | undefined {\n $(scope).find(FURNITURE_REGIONS).remove();\n\n const html = joinChunks(chunksOf($, scope, issue.url), true);\n if (!hasContent(html)) return undefined;\n\n return feedItem(issue, index, {\n title: issue.number\n ? `Non-decomposed issue #${issue.number} content`\n : \"Non-decomposed content\",\n url: issue.url,\n contentHtml: html,\n tags: [NON_DECOMPOSED_TAG],\n });\n}\n\n// A paragraph that introduces an entry starts with the entry's link, e.g.\n// \"📄 Title – description\" or \"Title Author\". Prose that\n// merely links mid-sentence (\"RubyKaigi, the conference…\") does not.\nfunction isEntryParagraph($: CheerioAPI, el: Element): boolean {\n const nodes = $(el).contents().toArray();\n let i = 0;\n while (i < nodes.length) {\n const node = nodes[i];\n if (\n node.type === \"text\" && (!node.data.trim() || EMOJI_ONLY.test(node.data))\n ) {\n i++;\n continue;\n }\n break;\n }\n\n const first = nodes[i];\n if (!first || first.type !== \"tag\" || !$(first).is(\"a[href]\")) return false;\n\n const after = nodes[i + 1];\n if (!after || after.type === \"tag\") return true;\n if (after.type !== \"text\") return true;\n // A trailing comma reads as \", the conference…\", i.e. mid-sentence\n return !after.data.trimStart().startsWith(\",\");\n}\n\n// --- headings -------------------------------------------------------------\n\ninterface DetectedHeading {\n text: string;\n block?: Element;\n /**\n * A centered title (\"What's New in React 19.3\") names the story it opens,\n * written below it, rather than the run of entries a section holds\n /\n story?: boolean;\n}\n\nfunction detectHeading(\n $: CheerioAPI,\n el: Element,\n): DetectedHeading | undefined {\n const $el = $(el);\n\n if (el.tagName === \"table\") {\n return $el.hasClass(\"el-heading\") ? { text: $el.text() } : undefined;\n }\n if (el.tagName !== \"p\") return undefined;\n if ($el.find(\"a[href]\").length > 0) return undefined;\n\n const text = normalizeWhitespace($el.text());\n if (!text || text.length > 80) return undefined;\n\n const style = $el.attr(\"style\") ?? \"\";\n // A centered paragraph names a section like any other mark of emphasis: the\n // \"What's New in React 19.3\" that opens a big release heads the story below\n // it, exactly as a section heading heads the entries below it\n if (/text-align:\\scenter/.test(style)) return { text, story: true };\n\n const upperSpan = $el.find(\"span\").toArray().some((span) =>\n /text-transform:\\suppercase/.test($(span).attr(\"style\") ?? \"\")\n );\n if (upperSpan) return { text };\n if (/font-size:\\s1\\.[4-9]em/.test(style)) return { text };\n if (/text-transform:\\suppercase/.test(style)) return { text };\n if (EMOJI_START.test(text)) return { text };\n\n // Compact list markers (\"IN BRIEF:\", \"RELEASES AND CODE:\"). They label the\n // list they open, so they stay in effect for later blocks only when their\n // own block holds nothing to label.\n const letters = text.replace(/[^A-Za-z]/g, \"\");\n if (\n letters.length >= 3 && letters === letters.toUpperCase() &&\n /^[A-Z0-9][A-Z0-9 ,'&/()\\-–—.:!?]:?$/.test(text)\n ) {\n const block = ($el.closest(\"table\").get(0) as Element | undefined) ?? el;\n const scoped = $(block).find(\"a[href]\").length > 0;\n return { text, block: scoped ? block : undefined };\n }\n\n return undefined;\n}\n\nfunction cleanHeading(value: string): string {\n return normalizeWhitespace(value)\n .replace(\n /^[\\p{Extended_Pictographic}\\p{Emoji_Presentation}\\u2600-\\u27BF\\u2B00-\\u2BFF\\uFE0F]+\\s/u,\n \"\",\n )\n .replace(/[:：]\\s$/, \"\")\n .trim();\n}\n\n// --- hero images ----------------------------------------------------------\n\n// An image introduces the entry below it when it stands alone in its block, i.e.\n// nothing but the image sits between it and the top of the issue. An image with\n// company (a caption, a story around it) belongs to that block instead.\nfunction isHeroImage($: CheerioAPI, el: Element, scope: Element): boolean {\n for (let node = el.parent; node && node !== scope; node = node.parent) {\n if (!isTag(node)) continue;\n if (normalizeWhitespace($(node).text()).length > 0) return false;\n if ($(node).find(\"img\").length > 1) return false;\n }\n return true;\n}\n\nfunction heroImageHtml(\n $: CheerioAPI,\n el: Element,\n baseUrl: string,\n): string | undefined {\n const src = $(el).attr(\"src\");\n if (!src) return undefined;\n return imageHtml(src, $(el).attr(\"alt\") ?? \"\", baseUrl);\n}\n\n// --- helpers --------------------------------------------------------------\n\n// The tags an entry wears besides the section the page places it under. Cooper\n// Press badges a paid placement in the entry's own byline, and wraps a\n// sponsored classified ad in an element that calls itself a sponsor.\nfunction sponsorTags($: CheerioAPI, ...elements: Element[]): string[] {\n const marked = elements.some((el) =>\n $(el).is(SPONSOR_MARKERS) || $(el).find(SPONSOR_MARKERS).length > 0\n );\n return marked ? [SPONSORED_TAG] : [];\n}\n\n// The section an item appears under, which is the tag that is not a sponsored\n// tag: an entry that sits under no section wears the sponsored tag alone.\nfunction sectionTag(item: FeedItem): string | undefined {\n return item.tags?.find((tag) => tag !== SPONSORED_TAG);\n}\n\n// The headline of an entry: the promoted \"mainlink\" when there is one,\n// otherwise the first link that has link text of its own\nfunction findEntryLink(\n $: CheerioAPI,\n el: Element,\n): { title: string; url?: string } | undefined {\n const main = $(el).find(\".mainlink a[href]\").first();\n if (main.length && normalizeWhitespace(main.text())) {\n return { title: normalizeWhitespace(main.text()), url: main.attr(\"href\") };\n }\n for (const a of $(el).find(\"a[href]\").toArray()) {\n const title = normalizeWhitespace($(a).text());\n if (title) return { title, url: $(a).attr(\"href\") };\n }\n return undefined;\n}\n",[],[253,367],"4e9e12b0-5e22-11f1-a88a-1607ee4eb77e","artificialAnalysis.ts","11573af6-586b-11f1-bbf2-ee650bb23af1",31,["D",1780363686579],"import * as cheerio from \"npm:cheerio@1.2.0\";\nimport { feedEndpoint } from \"../feed.ts\";\n\nconst CHANGELOG_URL = \"https://artificialanalysis.ai/changelog\";\nconst BASE_URL = \"https://artificialanalysis.ai\";\n\ninterface ArtificialAnalysisChangelogItem {\n id: string;\n url: string;\n title: string;\n content_html: string;\n summary?: string;\n date_published: string;\n}\n\nconst artificialAnalysisExpression = `{\n \"version\": $jsonFeedVersion,\n \"title\": \"Artificial Analysis Changelog\",\n \"home_page_url\": \"${CHANGELOG_URL}\",\n \"feed_url\": $feedUrl,\n \"description\": \"Latest Artificial Analysis changelog entries\",\n \"items\": items\n}`;\n\nexport const artificialAnalysisHandler = feedEndpoint({\n expression: artificialAnalysisExpression,\n loadSource: async () => ({\n data: { items: await fetchChangelogItems() },\n }),\n});\n\nasync function fetchChangelogItems(): Promise<\n ArtificialAnalysisChangelogItem[]\n> {\n const response = await fetch(CHANGELOG_URL, {\n headers: { \"Accept\": \"text/html\" },\n });\n if (!response.ok) {\n throw new Error(\n `Failed to fetch ${CHANGELOG_URL}: ${response.status} ${response.statusText}`,\n );\n }\n\n const $ = cheerio.load(await response.text());\n const items: ArtificialAnalysisChangelogItem[] = [];\n\n $(\"main h4\").each((_, heading) => {\n const dateText = $(heading).text().trim();\n const datePublished = parseChangelogDate(dateText);\n if (!datePublished) return;\n\n $(heading).nextUntil(\"h4\").find(\"div.flex.gap-8\").each((_, wrapper) => {\n const $wrapper = $(wrapper);\n\n // Type 1: full entries with title, kind, and summary inside an anchor\n const fullEntry = $wrapper.find(\"> a.flex.flex-col.gap-1\");\n if (fullEntry.length) {\n const link = fullEntry.first();\n const href = link.attr(\"href\");\n if (!href) return;\n\n absolutizeUrls($, link);\n const title = normalizeWhitespace(link.find(\"h3\").first().text()) ||\n normalizeWhitespace(link.find(\"img\").first().attr(\"alt\") || \"\");\n if (!title) return;\n\n const kind = normalizeWhitespace(\n link.find(\"span.text-xs\").first().text(),\n );\n const summary = normalizeWhitespace(link.find(\".prose p\").text());\n const url = new URL(href, BASE_URL).toString();\n\n items.push({\n id: `${dateText}-${items.length}-${url}`,\n url,\n title: kind ? `${title} - ${kind}` : title,\n content_html: link.html() ?? link.text(),\n summary: summary || undefined,\n date_published: datePublished,\n });\n return;\n }\n\n // Type 2: provider performance results\n const providerBlock = $wrapper.find(\"div.w-full.max-w-xl.py-2.px-2\");\n if (!providerBlock.length) return;\n\n const providerKind = normalizeWhitespace(\n providerBlock.find(\"span.text-sm\").text(),\n );\n if (!providerKind) return;\n\n providerBlock.find(\"a[href]\").each((_, modelLink) => {\n const $modelLink = $(modelLink);\n const href = $modelLink.attr(\"href\");\n if (!href) return;\n\n const title = normalizeWhitespace($modelLink.text());\n if (!title) return;\n\n const url = new URL(href, BASE_URL).toString();\n absolutizeUrls($, $modelLink);\n\n items.push({\n id: `${dateText}-${items.length}-${url}`,\n url,\n title: `${title} - ${providerKind}`,\n content_html: $modelLink.html() ?? $modelLink.text(),\n date_published: datePublished,\n });\n });\n });\n });\n\n const cutoff = Date.now() - 14 * 24 * 60 * 60 * 1000;\n return items.filter((item) => new Date(item.date_published).getTime() > cutoff);\n}\n\nfunction absolutizeUrls($: cheerio.CheerioAPI, root): void {\n root.find(\"[href]\").each((_, element) => {\n const href = $(element).attr(\"href\");\n if (href) $(element).attr(\"href\", new URL(href, BASE_URL).toString());\n });\n\n root.find(\"[src]\").each((_, element) => {\n const src = $(element).attr(\"src\");\n if (src) $(element).attr(\"src\", new URL(src, BASE_URL).toString());\n });\n}\n\nfunction parseChangelogDate(dateText: string): string | undefined {\n const match = /^(\\d{1,2})\\s+([A-Za-z]{3})\\s+(\\d{4})$/.exec(dateText);\n if (!match) return undefined;\n\n const months: Record = {\n Jan: 0,\n Feb: 1,\n Mar: 2,\n Apr: 3,\n May: 4,\n Jun: 5,\n Jul: 6,\n Aug: 7,\n Sep: 8,\n Oct: 9,\n Nov: 10,\n Dec: 11,\n };\n const month = months[match[2]];\n if (month === undefined) return undefined;\n\n return new Date(Date.UTC(Number(match[3]), month, Number(match[1])))\n .toISOString();\n}\n\nfunction normalizeWhitespace(value: string): string {\n return value.replace(/\\s+/g, \" \").trim();\n}\n",[],[253,375],"11350c2e-586b-11f1-ab2b-ee650bb23af1",["D",1779735229854],"directory",[165,166,167,168,169,170,171,172,173,174,175,176,177,178,179,180,181],[253],"b91a25d5-c80f-4625-99b8-b33457944a03",".nobs","01a07423-664e-75fc-bbb1-864df62bfd03",48,["D",1788654872144],[],[388],["D",1777602595753],["D",1789500119686],{"_11":12,"_122":68},{"_11":12,"_118":-5,"_67":68,"_30":66,"_119":120,"_121":24},{"_38":109,"_13":16,"_32":35,"_110":70,"_111":-5,"_112":-5,"_113":-5,"_114":-5,"_115":-5,"_116":14,"_117":12},"ctx",{"_51":-5,"_141":432,"_433":434,"_435":436,"_437":24,"_438":24,"_439":24,"_440":83,"_441":-5,"_442":-7,"_443":-7,"_444":26,"_445":24,"_446":24,"_447":-5,"_448":24,"_449":24,"_450":24,"_451":-5},"ENV",{"_403":404,"_405":406,"_407":16,"_408":409,"_410":411,"_412":413,"_414":415,"_416":417,"_418":419,"_420":421,"_422":423,"_424":425,"_426":427,"_428":429,"_430":431},"REACT_APP_SENTRY_DSN","https://586020d86f05489db63a554690acf1e9@o1398359.ingest.sentry.io/6724499","RENDER_GIT_COMMIT","e87eb04e707cf64e7e532c8e8ec395f209e17f15","RENDER_GIT_BRANCH","POSTHOG_API_TOKEN","phc_nmoyuAkStEZuUp69rYlg1DZgrIt19Gx0Y4Hwf8tGgJP","POSTHOG_API_HOST","https://wart.val.town","WEB_API_TEMPLATE","https://{val}.web.val.run","APP_URL","https://www.val.town","MODULE_URL","https://esm.town","VANITY_WEB_API_TEMPLATE","https://{label}.val.run/","DENOLS_URL","https://lsp.val.town","NODE_ENV","production","CLOUDFLARE_TURNSTILE_SITE_KEY","0x4AAAAAACtaIZJH7s9rBMZ6","GOOGLE_CLIENT_ID","875989696899-jtsqkhlbdv5jrlsk8631p7ipjg3c6o62.apps.googleusercontent.com","STRIPE_PUBLISHABLE_KEY","pk_live_51N2zvWEpZTTLipIkVEwXrUG9AT0yeOOEYxszarZT1jpsZV58Ly6wlg3SltZpHN26CUalOBmUSY7adQRQiSkCZz2500qSXNkf8Y","API_URL","https://api.val.town",[],"preferencesKeybindings","default","preferencesTheme","light","preferencesFixers","preferencesCopilot","preferencesSelectionAi","preferencesCloseBrackets","preferencesChatPanelState","preferencesDprint","preferencesDenofmt","notificationCount","flagDenols","isBusinessOrgMember","businessOrgHandle","isBusinessOrgAdmin","isHigherTownieTier","navCollapsed","currentOrgId"]

History · serve.ts