feat(Experimental): add support for metaformats

.github/workflows/build_and_test.yaml

@@ -17,7 +17,7 @@ jobs:
       - uses: actions/checkout@v2
       - uses: actions/setup-node@v1
         with:
-          node-version: 16
+          node-version: 18
       - name: Install dependencies
         run: yarn
       - name: Lint code

.nvmrc

@@ -1 +1 @@
-16
+18

README.md

@@ -109,6 +109,10 @@ These are sourced from the element themselves, a parent microformat, the HTML do
 
 When parsing microformats for text content, all the consecutive whitespace is collapsed into a single space. `<br/>` and `<p>` tags are treated as line breaks.
 
+#### `metaformats`
+
+Enables fallback to [metaformats](https://microformats.org/wiki/metaformats) parsing which looks at `<meta>` tags to infer content.
+
 ## Contributing
 
 See our [contributing guidelines](./CONTRIBUTING.md) for more information.

demo/demo.js

@@ -32,6 +32,10 @@ window.parseHtml = () => {
   const baseUrl = document.getElementById("base-url").value;
   const lang = document.getElementById("lang").checked;
   const textContent = document.getElementById("textContent").checked;
+  const metaformats = document.getElementById("metaformats").checked;
 
-  return parse(html, { baseUrl, experimental: { lang, textContent } });
+  return parse(html, {
+    baseUrl,
+    experimental: { lang, textContent, metaformats },
+  });
 };

demo/index.tpl.html

@@ -72,6 +72,16 @@ <h3>Experimental options</h3>
             />
             <span>Better text content</span>
           </label>
+          <label>
+            <input
+              type="checkbox"
+              name="metaformats"
+              id="metaformats"
+              value="true"
+              checked
+            />
+            <span>Metaformats parsing</span>
+          </label>
         </p>
 
         <div class="submit">

src/helpers/metaformats.ts

@@ -0,0 +1,246 @@
+import { Document, Element } from "parse5";
+
+import { MicroformatRoot, ParsingOptions } from "../types";
+import {
+  getAttributeIfTag,
+  getAttributeValue,
+  hasRelIntersect,
+} from "./attributes";
+import { isEnabled } from "./experimental";
+import { isElement, isTag } from "./nodeMatchers";
+
+/** Special key for title tag in meta collection */
+const TITLE_TAG_KEY = "<title>";
+const CANONICAL_URL_KEY = "<canonical>";
+const MEDIA_TYPES = ["image", "video", "audio"];
+
+interface ComplexMediaMeta {
+  value: string;
+  alt: string;
+}
+type MetaTagContent = string | ComplexMediaMeta;
+
+/**
+ * Creates a normalized store for meta tags
+ */
+const initializeMetaContentCollection = (): MetaContentCollection => {
+  /**
+   * Collection of all relevant meta tag content
+   * Since tag order isn't guaranteed, need to collect all value before applying defaults
+   */
+  const metaContent: Record<string, MetaTagContent[]> = {};
+
+  /**
+   * Gets the values of the first property found
+   * @param properties Array of properties to look for, preferred item first
+   */
+  const get = (properties: string[]) => {
+    for (const key of properties) {
+      if (metaContent[key]) {
+        return metaContent[key];
+      }
+    }
+    return;
+  };
+
+  /**
+   * Stores meta tag values.
+   *
+   * Includes following normalization rules:
+   * - Duplicates are removed from repeated (array) tags
+   * - src, url, and secure_url media tags are treated same as base (e.g. og:image:url -> og:image)
+   * - Alt text is added as property on last image url
+   */
+  const set = (key: string, value: string) => {
+    // Split tag name to normalize values like "og:video:url"
+    const [domain, type, subtype] = key.split(":");
+
+    // Media tags specific parsing
+    if (
+      (domain === "og" || domain === "twitter") &&
+      MEDIA_TYPES.includes(type)
+    ) {
+      if (subtype === "alt") {
+        const existingMedia = metaContent[`${domain}:${type}`];
+
+        if (existingMedia?.length) {
+          const last = existingMedia.pop();
+
+          if (typeof last === "string") {
+            existingMedia.push({ value: last, alt: value });
+          } else if (last) {
+            // Found duplicate alt text tag so re-inserting existing
+            // last should always be object. if condition added for types
+            existingMedia.push(last);
+          }
+        }
+
+        return; // Stop as alt text is already added
+      } else if (["url", "secure_url"].includes(subtype)) {
+        // Mutate key to normalize different url values
+        // Duplicates will be cleaned up on insertion
+        key = `${domain}:${type}`;
+      }
+    }
+    const existing = metaContent[key];
+
+    if (existing) {
+      const isDuplicate = existing
+        .map((existingValue) =>
+          typeof existingValue === "string"
+            ? existingValue
+            : existingValue.value
+        )
+        .some((existingValue) => value === existingValue);
+
+      if (!isDuplicate) {
+        metaContent[key].push(value);
+      } // Else ignore duplicates
+    } else {
+      metaContent[key] = [value];
+    }
+  };
+
+  return {
+    metaContent,
+    set,
+    get,
+  };
+};
+
+interface MetaContentCollection {
+  metaContent: Record<string, MetaTagContent[]>;
+  set: (key: string, value: string) => void;
+  get: (properties: string[]) => MetaTagContent[] | undefined;
+}
+
+const collectMetaTags = (head: Element): MetaContentCollection => {
+  const metaTags = initializeMetaContentCollection();
+
+  for (const i in head.childNodes) {
+    const child = head.childNodes[i];
+
+    if (!isElement(child)) {
+      continue;
+    }
+
+    const content = getAttributeIfTag(child, ["meta"], "content");
+    if (content) {
+      // Tags keys usually use the "name" attribute but open graph uses "property"
+      // Consider them separately in case a meta tag uses both
+      // e.g. <meta property="og:title" name="author" content="Johnny Complex" >
+      const property = getAttributeValue(child, "property");
+      if (property) {
+        metaTags.set(property, content);
+      }
+
+      const name = getAttributeValue(child, "name");
+      if (name && name !== property) {
+        metaTags.set(name, content);
+      }
+    } else if (child.tagName === "title" && "value" in child.childNodes[0]) {
+      metaTags.set(TITLE_TAG_KEY, child.childNodes[0].value);
+    } else if (
+      child.tagName === "link" &&
+      hasRelIntersect(child, ["canonical"])
+    ) {
+      const canonicalUrl = getAttributeValue(child, "href");
+      if (canonicalUrl) {
+        metaTags.set(CANONICAL_URL_KEY, canonicalUrl);
+      }
+    }
+  }
+  return metaTags;
+};
+
+/**
+ * Collect meta content into a microformat object
+ * @param metaTags Previously parsed meta tag collection
+ * @param options Library parsing options
+ */
+const combineRoot = (
+  metaTags: MetaContentCollection,
+  options: ParsingOptions
+): MicroformatRoot[] => {
+  const item: MicroformatRoot = { properties: {} };
+
+  if (isEnabled(options, "lang") && options.inherited.lang) {
+    item.lang = options.inherited.lang;
+  }
+
+  /**
+   * Define property on microformat root if values are found
+   * @param property Key of microformats property
+   * @param value Array of values for the property. Empty and undefined values are not added.
+   */
+  const setMicroformatProp = (
+    property: string,
+    value: MetaTagContent[] = []
+  ) => {
+    const filteredValue = value.filter(Boolean);
+    if (filteredValue.length) {
+      item.properties[property] = filteredValue;
+    }
+  };
+
+  let impliedRootClass = "h-entry";
+  const [ogType] = metaTags.get(["og:type"]) ?? [];
+  if (ogType && typeof ogType === "string") {
+    if (ogType === "profile") {
+      impliedRootClass = "h-card";
+    } else if (["music", "video"].some((type) => ogType.includes(type))) {
+      impliedRootClass = "h-cite";
+    } // else h-entry
+  }
+  item.type = [impliedRootClass];
+
+  setMicroformatProp(
+    "name",
+    metaTags.get(["og:title", "twitter:title", TITLE_TAG_KEY])
+  );
+  setMicroformatProp(
+    "summary",
+    metaTags.get(["og:description", "twitter:description", "description"])
+  );
+  setMicroformatProp("featured", metaTags.get(["og:image", "twitter:image"]));
+  setMicroformatProp("video", metaTags.get(["og:video", "twitter:video"]));
+  setMicroformatProp("audio", metaTags.get(["og:audio", "twitter:audio"]));
+  setMicroformatProp(
+    "published",
+    metaTags.get(["article:published_time", "date"])
+  );
+  setMicroformatProp("updated", metaTags.get(["article:modified_time"]));
+  setMicroformatProp("author", metaTags.get(["article:author", "author"]));
+  setMicroformatProp("url", metaTags.get(["og:url", CANONICAL_URL_KEY]));
+
+  // Publication properties useful for h-cite
+  setMicroformatProp(
+    "publication",
+    metaTags.get(["og:site_name", "publisher"])
+  );
+
+  if (impliedRootClass === "h-card") {
+    setMicroformatProp("given-name", metaTags.get(["profile:first_name"]));
+    setMicroformatProp("family-name", metaTags.get(["profile:last_name"]));
+  }
+
+  if (Object.keys(item.properties).length === 0) {
+    return [];
+  }
+
+  return [item];
+};
+
+export const parseMetaformats = (
+  doc: Document,
+  options: ParsingOptions
+): MicroformatRoot[] => {
+  // Per validation, html element will always be found
+  const html = doc.childNodes.find(isTag("html"));
+  const head = html?.childNodes.find(isTag("head"));
+
+  // Per manual testing, head will always be defined
+  // eslint-disable-next-line @typescript-eslint/no-non-null-assertion
+  const metaContent = collectMetaTags(head!);
+  return combineRoot(metaContent, options);
+};

src/helpers/nodeMatchers.ts

@@ -20,6 +20,11 @@ const propClassRegex = classRegex("(p|e|u|dt)");
 export const isElement = (node: Node): node is Element =>
   "tagName" in node && "childNodes" in node;
 
+export const isTag =
+  (tagName: string) =>
+  (node: Node): node is Element =>
+    isElement(node) && node.tagName === tagName;
+
 export const isTextNode = (node: Node): node is TextNode => "value" in node;
 
 export const isMicroformatV2Root = (node: Element): boolean =>

src/index.ts

@@ -7,6 +7,7 @@ export interface Options {
   experimental?: {
     lang?: boolean;
     textContent?: boolean;
+    metaformats?: boolean;
   };
 }
 

src/microformats/property.ts

@@ -34,6 +34,7 @@ export const parseP = (node: Element, options: ParsingOptions): string =>
   getAttributeIfTag(node, ["abbr", "link"], "title") ??
   getAttributeIfTag(node, ["data"], "value") ??
   getAttributeIfTag(node, ["img", "area"], "alt") ??
+  getAttributeIfTag(node, ["meta"], "content") ??
   textContent(node, options);
 
 export const parseU = (
@@ -49,6 +50,7 @@ export const parseU = (
     valueClassPattern(node, options) ??
     getAttributeIfTag(node, ["abbr"], "title") ??
     getAttributeIfTag(node, ["data", "input"], "value") ??
+    getAttributeIfTag(node, ["meta"], "content") ??
     textContent(node, options);
 
   if (typeof url === "string" && isLocalLink(url)) {
@@ -63,6 +65,7 @@ const parseDt = (node: Element, options: ParsingOptions): string =>
   getAttributeIfTag(node, ["time", "ins", "del"], "datetime") ??
   getAttributeIfTag(node, ["abbr"], "title") ??
   getAttributeIfTag(node, ["data", "input"], "value") ??
+  getAttributeIfTag(node, ["meta"], "content") ??
   textContent(node, options);
 
 export const parseE = (node: Element, options: ParsingOptions): Html => {

src/parser.ts

@@ -6,6 +6,8 @@ import { isMicroformatRoot } from "./helpers/nodeMatchers";
 import { ParsedDocument, ParserOptions, ParsingOptions } from "./types";
 import { validateParsedHtml } from "./validator";
 import { documentSetup } from "./helpers/documentSetup";
+import { parseMetaformats } from "./helpers/metaformats";
+import { isEnabled } from "./helpers/experimental";
 
 export const parser = (
   html: string,
@@ -22,12 +24,17 @@ export const parser = (
     idRefs,
     inherited: { roots: [], lang },
   };
+  let items = findChildren(doc, isMicroformatRoot).map((mf) =>
+    parseMicroformat(mf, parsingOptions)
+  );
+
+  if (items.length === 0 && isEnabled(parsingOptions, "metaformats")) {
+    items = parseMetaformats(doc, parsingOptions);
+  }
 
   return {
     rels,
     "rel-urls": relUrls,
-    items: findChildren(doc, isMicroformatRoot).map((mf) =>
-      parseMicroformat(mf, parsingOptions)
-    ),
+    items,
   };
 };

src/rels/rels.ts

@@ -43,6 +43,7 @@ export const parseRel = (
       relUrls[href] = { rels: [rel], text };
     } else if (!relUrls[href].rels.includes(rel)) {
       relUrls[href].rels.push(rel);
+      relUrls[href].rels.sort();
     }
 
     if (text && !relUrls[href].text) {