@j0hanz/superfetch 2.4.3 → 2.4.4

package/dist/dom-noise-removal.js

@@ -5,43 +5,112 @@
 import { parseHTML } from 'linkedom';
 import { config } from './config.js';
 import { isObject } from './type-guards.js';
-// ─────────────────────────────────────────────────────────────────────────────
-// DOM Type Guards and Accessors
-// ─────────────────────────────────────────────────────────────────────────────
+/* -------------------------------------------------------------------------------------------------
+ * DOM guards & small helpers
+ * ------------------------------------------------------------------------------------------------- */
 function isElement(node) {
     return (isObject(node) &&
         'getAttribute' in node &&
         typeof node.getAttribute === 'function');
 }
-function getBodyInnerHtml(document) {
-    if (!isObject(document))
-        return undefined;
-    const { body } = document;
-    if (isObject(body) && typeof body.innerHTML === 'string') {
-        return body.innerHTML;
-    }
-    return undefined;
+function isNodeListLike(value) {
+    return (isObject(value) &&
+        typeof value.length === 'number');
 }
-function getDocumentToString(document) {
-    if (!isObject(document))
-        return undefined;
-    if (typeof document.toString === 'function') {
-        return document.toString.bind(document);
+function getNodeListItem(nodes, index) {
+    if ('item' in nodes && typeof nodes.item === 'function') {
+        return nodes.item(index);
     }
-    return undefined;
+    return nodes[index] ?? null;
 }
-function getDocumentElementOuterHtml(document) {
-    if (!isObject(document))
-        return undefined;
-    const docEl = document.documentElement;
-    if (isObject(docEl) && typeof docEl.outerHTML === 'string') {
-        return docEl.outerHTML;
+/**
+ * Remove nodes from a list/iterable.
+ * - For NodeList-like collections we iterate backwards to be safe with live collections.
+ * - For iterables we snapshot into an array first.
+ */
+function removeNodes(nodes, shouldRemove) {
+    if (isNodeListLike(nodes)) {
+        for (let i = nodes.length - 1; i >= 0; i -= 1) {
+            const node = getNodeListItem(nodes, i);
+            if (node && shouldRemove(node))
+                node.remove();
+        }
+        return;
+    }
+    for (const node of nodes) {
+        if (shouldRemove(node))
+            node.remove();
     }
-    return undefined;
 }
-// ─────────────────────────────────────────────────────────────────────────────
-// Noise Detection Constants
-// ─────────────────────────────────────────────────────────────────────────────
+/* -------------------------------------------------------------------------------------------------
+ * Fast-path parsing heuristics
+ * ------------------------------------------------------------------------------------------------- */
+const HTML_DOCUMENT_MARKERS = /<\s*(?:!doctype|html|head|body)\b/i;
+function isFullDocumentHtml(html) {
+    return HTML_DOCUMENT_MARKERS.test(html);
+}
+const NOISE_SCAN_LIMIT = 50_000;
+const NOISE_MARKERS = [
+    '<script',
+    '<style',
+    '<noscript',
+    '<iframe',
+    '<nav',
+    '<footer',
+    '<header',
+    '<form',
+    '<button',
+    '<input',
+    '<select',
+    '<textarea',
+    '<svg',
+    '<canvas',
+    ' aria-hidden="true"',
+    " aria-hidden='true'",
+    ' hidden',
+    ' role="navigation"',
+    " role='navigation'",
+    ' role="banner"',
+    " role='banner'",
+    ' role="complementary"',
+    " role='complementary'",
+    ' role="contentinfo"',
+    " role='contentinfo'",
+    ' role="tree"',
+    " role='tree'",
+    ' role="menubar"',
+    " role='menubar'",
+    ' role="menu"',
+    " role='menu'",
+    ' banner',
+    ' promo',
+    ' announcement',
+    ' cta',
+    ' advert',
+    ' newsletter',
+    ' subscribe',
+    ' cookie',
+    ' consent',
+    ' popup',
+    ' modal',
+    ' overlay',
+    ' toast',
+    ' fixed',
+    ' sticky',
+    ' z-50',
+    ' z-4',
+    ' isolate',
+    ' breadcrumb',
+    ' pagination',
+];
+function mayContainNoise(html) {
+    const sample = html.length > NOISE_SCAN_LIMIT ? html.slice(0, NOISE_SCAN_LIMIT) : html;
+    const haystack = sample.toLowerCase();
+    return NOISE_MARKERS.some((marker) => haystack.includes(marker));
+}
+/* -------------------------------------------------------------------------------------------------
+ * Noise selectors & classification
+ * ------------------------------------------------------------------------------------------------- */
 const STRUCTURAL_TAGS = new Set([
     'script',
     'style',
@@ -82,10 +151,10 @@ const CANDIDATE_NOISE_SELECTOR = [
     '[style]',
 ].join(',');
 function buildNoiseSelector(extraSelectors) {
-    const extra = extraSelectors.filter((selector) => selector.trim().length > 0);
-    if (extra.length === 0)
-        return BASE_NOISE_SELECTOR;
-    return `${BASE_NOISE_SELECTOR},${extra.join(',')}`;
+    const extra = extraSelectors.filter((s) => s.trim().length > 0);
+    return extra.length === 0
+        ? BASE_NOISE_SELECTOR
+        : `${BASE_NOISE_SELECTOR},${extra.join(',')}`;
 }
 const NAVIGATION_ROLES = new Set([
     'navigation',
@@ -140,214 +209,151 @@ const BASE_PROMO_TOKENS = [
     'pager',
     'taglist',
 ];
-/**
- * Get promo tokens merged with any user-configured extra tokens.
- * Memoized because it is used in hot paths when scanning many nodes.
- */
-let promoTokensCache = null;
-function getPromoTokens() {
-    if (promoTokensCache)
-        return promoTokensCache;
-    const tokens = new Set(BASE_PROMO_TOKENS);
-    for (const token of config.noiseRemoval.extraTokens) {
-        const normalized = token.toLowerCase().trim();
-        if (normalized)
-            tokens.add(normalized);
-    }
-    promoTokensCache = tokens;
-    return tokens;
-}
-let promoRegexCache = null;
-function getPromoRegex() {
-    if (promoRegexCache)
-        return promoRegexCache;
-    const tokens = Array.from(getPromoTokens());
-    const escaped = tokens.map((t) => t.replace(/[.*+?^${}()|[\]\\]/g, '\\$&'));
-    const pattern = `(?:^|[^a-z0-9])(?:${escaped.join('|')})(?:$|[^a-z0-9])`;
-    promoRegexCache = new RegExp(pattern, 'i');
-    return promoRegexCache;
-}
 const HEADER_NOISE_PATTERN = /\b(site-header|masthead|topbar|navbar|nav(?:bar)?|menu|header-nav)\b/i;
 const FIXED_PATTERN = /\b(fixed|sticky)\b/;
 const HIGH_Z_PATTERN = /\bz-(?:4\d|50)\b/;
 const ISOLATE_PATTERN = /\bisolate\b/;
-const HTML_DOCUMENT_MARKERS = /<\s*(?:!doctype|html|head|body)\b/i;
-const NOISE_MARKERS = [
-    '<script',
-    '<style',
-    '<noscript',
-    '<iframe',
-    '<nav',
-    '<footer',
-    '<header',
-    '<form',
-    '<button',
-    '<input',
-    '<select',
-    '<textarea',
-    '<svg',
-    '<canvas',
-    ' aria-hidden="true"',
-    " aria-hidden='true'",
-    ' hidden',
-    ' role="navigation"',
-    " role='navigation'",
-    ' role="banner"',
-    " role='banner'",
-    ' role="complementary"',
-    " role='complementary'",
-    ' role="contentinfo"',
-    " role='contentinfo'",
-    ' role="tree"',
-    " role='tree'",
-    ' role="menubar"',
-    " role='menubar'",
-    ' role="menu"',
-    " role='menu'",
-    ' banner',
-    ' promo',
-    ' announcement',
-    ' cta',
-    ' advert',
-    ' newsletter',
-    ' subscribe',
-    ' cookie',
-    ' consent',
-    ' popup',
-    ' modal',
-    ' overlay',
-    ' toast',
-    ' fixed',
-    ' sticky',
-    ' z-50',
-    ' z-4',
-    ' isolate',
-    ' breadcrumb',
-    ' pagination',
-];
-// ─────────────────────────────────────────────────────────────────────────────
-// Noise Detection Functions
-// ─────────────────────────────────────────────────────────────────────────────
-const NOISE_SCAN_LIMIT = 50_000;
-function mayContainNoise(html) {
-    // Fast path: only scan a bounded prefix; parsing is the expensive step anyway.
-    // Most noise markers appear near the top of the document (nav, scripts, meta, etc.).
-    const sample = html.length > NOISE_SCAN_LIMIT ? html.slice(0, NOISE_SCAN_LIMIT) : html;
-    const haystack = sample.toLowerCase();
-    return NOISE_MARKERS.some((marker) => haystack.includes(marker));
-}
-function isFullDocumentHtml(html) {
-    return HTML_DOCUMENT_MARKERS.test(html);
-}
-function isStructuralNoiseTag(tagName) {
-    return STRUCTURAL_TAGS.has(tagName);
-}
-function isInteractiveComponent(element) {
-    const role = element.getAttribute('role');
-    if (role && INTERACTIVE_CONTENT_ROLES.has(role))
-        return true;
-    // Check for common UI framework data attributes that indicate managed visibility
-    const dataState = element.getAttribute('data-state');
-    if (dataState === 'inactive' || dataState === 'closed')
-        return true;
-    const dataOrientation = element.getAttribute('data-orientation');
-    if (dataOrientation === 'horizontal' || dataOrientation === 'vertical') {
-        return true;
+class PromoDetector {
+    tokenCache = null;
+    regexCache = null;
+    matches(className, id) {
+        const regex = this.getRegex();
+        return regex.test(className) || regex.test(id);
     }
-    // Check for accordion/collapse patterns
-    if (element.getAttribute('data-accordion-item') !== null)
-        return true;
-    if (element.getAttribute('data-radix-collection-item') !== null)
-        return true;
-    return false;
-}
-function isElementHidden(element) {
-    const style = element.getAttribute('style') ?? '';
-    return (element.getAttribute('hidden') !== null ||
-        element.getAttribute('aria-hidden') === 'true' ||
-        /\bdisplay\s*:\s*none\b/i.test(style) ||
-        /\bvisibility\s*:\s*hidden\b/i.test(style));
-}
-function hasNoiseRole(role) {
-    return role !== null && NAVIGATION_ROLES.has(role);
-}
-function matchesPromoIdOrClass(className, id) {
-    const regex = getPromoRegex();
-    return regex.test(className) || regex.test(id);
-}
-function matchesFixedOrHighZIsolate(className) {
-    return (FIXED_PATTERN.test(className) ||
-        (HIGH_Z_PATTERN.test(className) && ISOLATE_PATTERN.test(className)));
-}
-function readElementMetadata(element) {
-    return {
-        tagName: element.tagName.toLowerCase(),
-        className: element.getAttribute('class') ?? '',
-        id: element.getAttribute('id') ?? '',
-        role: element.getAttribute('role'),
-        isHidden: isElementHidden(element),
-    };
-}
-function isBoilerplateHeader({ className, id, role, }) {
-    if (hasNoiseRole(role))
-        return true;
-    const combined = `${className} ${id}`.toLowerCase();
-    return HEADER_NOISE_PATTERN.test(combined);
-}
-function isNoiseElement(node) {
-    const metadata = readElementMetadata(node);
-    const isComplementaryAside = metadata.tagName === 'aside' && metadata.role === 'complementary';
-    const shouldCheckHidden = metadata.isHidden && !isInteractiveComponent(node);
-    const isInteractiveStructural = isStructuralNoiseTag(metadata.tagName) && isInteractiveComponent(node);
-    return ((isStructuralNoiseTag(metadata.tagName) && !isInteractiveStructural) ||
-        ALWAYS_NOISE_TAGS.has(metadata.tagName) ||
-        (metadata.tagName === 'header' && isBoilerplateHeader(metadata)) ||
-        shouldCheckHidden ||
-        (!isComplementaryAside && hasNoiseRole(metadata.role)) ||
-        matchesFixedOrHighZIsolate(metadata.className) ||
-        matchesPromoIdOrClass(metadata.className, metadata.id));
-}
-function isNodeListLike(value) {
-    return isObject(value) && typeof value.length === 'number';
-}
-function tryGetNodeListItem(nodes, index) {
-    if ('item' in nodes && typeof nodes.item === 'function') {
-        return nodes.item(index);
-    }
-    return nodes[index] ?? null;
-}
-function removeNoiseFromNodeListLike(nodes, shouldCheckNoise) {
-    for (let index = nodes.length - 1; index >= 0; index -= 1) {
-        const node = tryGetNodeListItem(nodes, index);
-        if (!node)
-            continue;
-        if (isElement(node) && (!shouldCheckNoise || isNoiseElement(node))) {
-            node.remove();
+    getTokens() {
+        if (this.tokenCache)
+            return this.tokenCache;
+        const tokens = new Set(BASE_PROMO_TOKENS);
+        for (const token of config.noiseRemoval.extraTokens) {
+            const normalized = token.toLowerCase().trim();
+            if (normalized)
+                tokens.add(normalized);
         }
+        this.tokenCache = tokens;
+        return tokens;
+    }
+    getRegex() {
+        if (this.regexCache)
+            return this.regexCache;
+        const tokens = [...this.getTokens()];
+        const escaped = tokens.map((t) => t.replace(/[.*+?^${}()|[\]\\]/g, '\\$&'));
+        const pattern = `(?:^|[^a-z0-9])(?:${escaped.join('|')})(?:$|[^a-z0-9])`;
+        this.regexCache = new RegExp(pattern, 'i');
+        return this.regexCache;
     }
 }
-function removeNoiseNodes(nodes, shouldCheckNoise = true) {
-    if (isNodeListLike(nodes)) {
-        removeNoiseFromNodeListLike(nodes, shouldCheckNoise);
-        return;
+class NoiseClassifier {
+    promo;
+    constructor(promo) {
+        this.promo = promo;
     }
-    const nodeList = Array.from(nodes);
-    for (const node of nodeList) {
-        if (isElement(node) && (!shouldCheckNoise || isNoiseElement(node))) {
-            node.remove();
-        }
+    isNoise(element) {
+        const meta = this.readMetadata(element);
+        if (this.isStructuralNoise(meta, element))
+            return true;
+        if (ALWAYS_NOISE_TAGS.has(meta.tagName))
+            return true;
+        if (this.isHeaderBoilerplate(meta))
+            return true;
+        if (this.isHiddenNoise(meta, element))
+            return true;
+        if (this.isRoleNoise(meta))
+            return true;
+        if (this.matchesFixedOrHighZIsolate(meta.className))
+            return true;
+        if (this.promo.matches(meta.className, meta.id))
+            return true;
+        return false;
+    }
+    readMetadata(element) {
+        return {
+            tagName: element.tagName.toLowerCase(),
+            className: element.getAttribute('class') ?? '',
+            id: element.getAttribute('id') ?? '',
+            role: element.getAttribute('role'),
+            isHidden: this.isHidden(element),
+        };
+    }
+    isStructuralNoise(meta, element) {
+        if (!STRUCTURAL_TAGS.has(meta.tagName))
+            return false;
+        // Interactive structural components (dialogs, menus) are handled elsewhere.
+        return !this.isInteractiveComponent(element);
+    }
+    isHeaderBoilerplate(meta) {
+        if (meta.tagName !== 'header')
+            return false;
+        if (this.hasNoiseRole(meta.role))
+            return true;
+        const combined = `${meta.className} ${meta.id}`.toLowerCase();
+        return HEADER_NOISE_PATTERN.test(combined);
+    }
+    isHiddenNoise(meta, element) {
+        if (!meta.isHidden)
+            return false;
+        // Don't remove hidden interactive components (they may be managed by UI framework state).
+        return !this.isInteractiveComponent(element);
+    }
+    isRoleNoise(meta) {
+        const isComplementaryAside = meta.tagName === 'aside' && meta.role === 'complementary';
+        if (isComplementaryAside)
+            return false;
+        return this.hasNoiseRole(meta.role);
+    }
+    hasNoiseRole(role) {
+        return role !== null && NAVIGATION_ROLES.has(role);
+    }
+    matchesFixedOrHighZIsolate(className) {
+        return (FIXED_PATTERN.test(className) ||
+            (HIGH_Z_PATTERN.test(className) && ISOLATE_PATTERN.test(className)));
+    }
+    isHidden(element) {
+        const style = element.getAttribute('style') ?? '';
+        return (element.getAttribute('hidden') !== null ||
+            element.getAttribute('aria-hidden') === 'true' ||
+            /\bdisplay\s*:\s*none\b/i.test(style) ||
+            /\bvisibility\s*:\s*hidden\b/i.test(style));
+    }
+    isInteractiveComponent(element) {
+        const role = element.getAttribute('role');
+        if (role && INTERACTIVE_CONTENT_ROLES.has(role))
+            return true;
+        const dataState = element.getAttribute('data-state');
+        if (dataState === 'inactive' || dataState === 'closed')
+            return true;
+        const dataOrientation = element.getAttribute('data-orientation');
+        if (dataOrientation === 'horizontal' || dataOrientation === 'vertical')
+            return true;
+        if (element.getAttribute('data-accordion-item') !== null)
+            return true;
+        if (element.getAttribute('data-radix-collection-item') !== null)
+            return true;
+        return false;
     }
 }
-function stripNoiseNodes(document) {
-    const targetSelectors = buildNoiseSelector(config.noiseRemoval.extraSelectors);
-    const potentialNoiseNodes = document.querySelectorAll(targetSelectors);
-    removeNoiseNodes(potentialNoiseNodes, false);
-    const allElements = document.querySelectorAll(CANDIDATE_NOISE_SELECTOR);
-    removeNoiseNodes(allElements, true);
+class NoiseStripper {
+    classifier;
+    constructor(classifier) {
+        this.classifier = classifier;
+    }
+    strip(document) {
+        this.removeBySelector(document, buildNoiseSelector(config.noiseRemoval.extraSelectors), 
+        /* checkNoise */ false);
+        this.removeBySelector(document, CANDIDATE_NOISE_SELECTOR, 
+        /* checkNoise */ true);
+    }
+    removeBySelector(document, selector, checkNoise) {
+        const nodes = document.querySelectorAll(selector);
+        removeNodes(nodes, (node) => {
+            if (!isElement(node))
+                return false;
+            return checkNoise ? this.classifier.isNoise(node) : true;
+        });
+    }
 }
-// ─────────────────────────────────────────────────────────────────────────────
-// URL Resolution
-// ─────────────────────────────────────────────────────────────────────────────
+/* -------------------------------------------------------------------------------------------------
+ * Relative URL resolution
+ * ------------------------------------------------------------------------------------------------- */
 const SKIP_URL_PREFIXES = [
     '#',
     'java' + 'script:',
@@ -360,9 +366,6 @@ function shouldSkipUrlResolution(url) {
     const normalized = url.trim().toLowerCase();
     return SKIP_URL_PREFIXES.some((prefix) => normalized.startsWith(prefix));
 }
-/**
- * Safely resolve a relative URL to absolute using base URL.
- */
 function tryResolveUrl(relativeUrl, baseUrl) {
     try {
         return new URL(relativeUrl, baseUrl).href;
@@ -371,87 +374,142 @@ function tryResolveUrl(relativeUrl, baseUrl) {
         return null;
     }
 }
-function resolveAnchorElement(element, base) {
-    const href = element.getAttribute('href');
-    if (href && !shouldSkipUrlResolution(href)) {
+class RelativeUrlResolver {
+    resolve(document, baseUrl) {
+        let base;
+        try {
+            base = new URL(baseUrl);
+        }
+        catch {
+            // invalid base URL - skip resolution
+            return;
+        }
+        for (const element of document.querySelectorAll('a[href], img[src], source[srcset]')) {
+            const tag = element.tagName.toLowerCase();
+            if (tag === 'a')
+                this.resolveAnchor(element, base);
+            else if (tag === 'img')
+                this.resolveImage(element, base);
+            else if (tag === 'source')
+                this.resolveSource(element, base);
+        }
+    }
+    resolveAnchor(element, base) {
+        const href = element.getAttribute('href');
+        if (!href || shouldSkipUrlResolution(href))
+            return;
         const resolved = tryResolveUrl(href, base);
         if (resolved)
             element.setAttribute('href', resolved);
     }
-}
-function resolveImageElement(element, base) {
-    const src = element.getAttribute('src');
-    if (src && !shouldSkipUrlResolution(src)) {
+    resolveImage(element, base) {
+        const src = element.getAttribute('src');
+        if (!src || shouldSkipUrlResolution(src))
+            return;
         const resolved = tryResolveUrl(src, base);
         if (resolved)
             element.setAttribute('src', resolved);
     }
-}
-function resolveSourceElement(element, base) {
-    const srcset = element.getAttribute('srcset');
-    if (!srcset)
-        return;
-    const resolved = srcset
-        .split(',')
-        .map((entry) => {
-        const parts = entry.trim().split(/\s+/);
-        const url = parts[0];
-        if (url) {
-            const resolvedUrl = tryResolveUrl(url, base);
-            if (resolvedUrl)
-                parts[0] = resolvedUrl;
-        }
-        return parts.join(' ');
-    })
-        .join(', ');
-    element.setAttribute('srcset', resolved);
-}
-function resolveRelativeUrls(document, baseUrl) {
-    try {
-        const base = new URL(baseUrl);
-        for (const element of document.querySelectorAll('a[href], img[src], source[srcset]')) {
-            const tag = element.tagName.toLowerCase();
-            if (tag === 'a') {
-                resolveAnchorElement(element, base);
-            }
-            else if (tag === 'img') {
-                resolveImageElement(element, base);
-            }
-            else if (tag === 'source') {
-                resolveSourceElement(element, base);
+    /**
+     * Keep original behavior: srcset entries are always attempted to be resolved (no prefix skipping).
+     */
+    resolveSource(element, base) {
+        const srcset = element.getAttribute('srcset');
+        if (!srcset)
+            return;
+        const resolved = srcset
+            .split(',')
+            .map((entry) => {
+            const parts = entry.trim().split(/\s+/);
+            const url = parts[0];
+            if (url) {
+                const resolvedUrl = tryResolveUrl(url, base);
+                if (resolvedUrl)
+                    parts[0] = resolvedUrl;
             }
+            return parts.join(' ');
+        })
+            .join(', ');
+        element.setAttribute('srcset', resolved);
+    }
+}
+/* -------------------------------------------------------------------------------------------------
+ * Serialization
+ * ------------------------------------------------------------------------------------------------- */
+class DocumentSerializer {
+    /**
+     * Preserve existing behavior:
+     * - Prefer body.innerHTML only if it has "substantial" content (> 100 chars).
+     * - Otherwise fall back to document.toString(), then documentElement.outerHTML, then original HTML.
+     */
+    serialize(document, fallbackHtml) {
+        const bodyInner = this.getBodyInnerHtml(document);
+        if (bodyInner && bodyInner.trim().length > 100)
+            return bodyInner;
+        const toStringFn = this.getDocumentToString(document);
+        if (toStringFn)
+            return toStringFn();
+        const outer = this.getDocumentElementOuterHtml(document);
+        if (outer)
+            return outer;
+        return fallbackHtml;
+    }
+    getBodyInnerHtml(document) {
+        if (!isObject(document))
+            return undefined;
+        const { body } = document;
+        if (isObject(body) &&
+            typeof body.innerHTML === 'string') {
+            return body.innerHTML;
         }
+        return undefined;
     }
-    catch {
-        /* invalid base URL - skip resolution */
+    getDocumentToString(document) {
+        if (!isObject(document))
+            return undefined;
+        const fn = document.toString;
+        if (typeof fn === 'function')
+            return fn.bind(document);
+        return undefined;
+    }
+    getDocumentElementOuterHtml(document) {
+        if (!isObject(document))
+            return undefined;
+        const docEl = document.documentElement;
+        if (isObject(docEl) &&
+            typeof docEl.outerHTML === 'string') {
+            return docEl.outerHTML;
+        }
+        return undefined;
     }
 }
-export function removeNoiseFromHtml(html, document, baseUrl) {
-    const shouldParse = isFullDocumentHtml(html) || mayContainNoise(html);
-    if (!shouldParse)
-        return html;
-    try {
-        const resolvedDocument = document ?? parseHTML(html).document;
-        stripNoiseNodes(resolvedDocument);
-        // Resolve relative URLs before converting to markdown
-        if (baseUrl) {
-            resolveRelativeUrls(resolvedDocument, baseUrl);
+/* -------------------------------------------------------------------------------------------------
+ * Public pipeline
+ * ------------------------------------------------------------------------------------------------- */
+class HtmlNoiseRemovalPipeline {
+    promo = new PromoDetector();
+    classifier = new NoiseClassifier(this.promo);
+    stripper = new NoiseStripper(this.classifier);
+    urlResolver = new RelativeUrlResolver();
+    serializer = new DocumentSerializer();
+    removeNoise(html, document, baseUrl) {
+        const shouldParse = isFullDocumentHtml(html) || mayContainNoise(html);
+        if (!shouldParse)
+            return html;
+        try {
+            const resolvedDocument = document ?? parseHTML(html).document;
+            this.stripper.strip(resolvedDocument);
+            if (baseUrl) {
+                this.urlResolver.resolve(resolvedDocument, baseUrl);
+            }
+            return this.serializer.serialize(resolvedDocument, html);
         }
-        const bodyInnerHtml = getBodyInnerHtml(resolvedDocument);
-        // Only use body innerHTML if it has substantial content
-        // On some sites (e.g., Framer), noise removal empties body but leaves content in documentElement
-        if (bodyInnerHtml && bodyInnerHtml.trim().length > 100) {
-            return bodyInnerHtml;
+        catch {
+            return html;
         }
-        const docToString = getDocumentToString(resolvedDocument);
-        if (docToString)
-            return docToString();
-        const documentElementOuterHtml = getDocumentElementOuterHtml(resolvedDocument);
-        if (documentElementOuterHtml)
-            return documentElementOuterHtml;
-        return html;
-    }
-    catch {
-        return html;
     }
 }
+const pipeline = new HtmlNoiseRemovalPipeline();
+export function removeNoiseFromHtml(html, document, baseUrl) {
+    return pipeline.removeNoise(html, document, baseUrl);
+}