@aglyn/plugins-ai 1.0.0-beta.292 → 1.0.0-beta.293

package/src/lib/site-import/site-import-html.js

@@ -50,7 +50,12 @@
     /** A mega-menu's entries and the footer's columns together. */ navLinks: 150,
     forms: 10,
     fieldsPerForm: 30,
-    icons: 12
+    icons: 12,
+    /** The structure's ceilings (AGL-4106): elements, inline icons, links and the page's own style sheets. */ elements: 20000,
+    svgIcons: 200,
+    anchors: 800,
+    styleChars: 1500000,
+    stylesheets: 4
 };
 /** Elements removed with everything inside them. */ const REMOVED_WITH_CONTENT = [
     'script',
@@ -68,10 +73,14 @@
     'video',
     'audio'
 ];
-/** Form controls, removed with their content once the forms are summarized. */ const FORM_CONTROLS = [
+/**
+ * Form controls, removed with their content once the forms are summarized.
+ * A `<button>` outside a form is not one (AGL-4106): its words are a line
+ * ("Watch the 40-second film", a pricing tab's name) and a picture inside it
+ * is still a picture. A form's own buttons go with the form.
+ */ const FORM_CONTROLS = [
     'textarea',
-    'select',
-    'button'
+    'select'
 ];
 /** Void elements removed wherever they appear. */ const REMOVED_VOID = /<(?:embed|param|source|track|link|base)\b(?:[^>"']|"[^"]*"|'[^']*')*>/gi;
 const PASSWORD_INPUT = /<input\b(?:[^>"']|"[^"]*"|'[^']*')*\btype\s*=\s*["']?password\b/i;
@@ -91,6 +100,8 @@ const BLOCK_TAGS = {
     figcaption: 'paragraph',
     caption: 'paragraph',
     label: 'paragraph',
+    /** A disclosure's question, a button's words (AGL-4106). */ summary: 'paragraph',
+    button: 'paragraph',
     blockquote: 'quote'
 };
 /** Tags that end a run of loose text without being a block themselves. */ const BREAK_TAGS = new Set([
@@ -109,6 +120,8 @@ const BLOCK_TAGS = {
     'tr',
     'tbody',
     'thead',
+    'tfoot',
+    'details',
     'figure',
     'form',
     'hr',
@@ -274,6 +287,8 @@ function removeElements(html, tags) {
             kept.push({
                 fields
             });
+            // Where the form sat, for the layout (AGL-4106): a marker, never words.
+            return ` <aglyn-form data-form="${kept.length - 1}"> `;
         }
         return ' ';
     });
@@ -430,12 +445,325 @@ const ORGANIZATION_TYPE = /(?:^|\b)(?:Organization|Corporation|LocalBusiness|NGO
         sameAs: [].concat((_node_sameAs = node['sameAs']) != null ? _node_sameAs : []).map((entry)=>urlOf(entry)).filter(Boolean).slice(0, 20)
     };
 }
+/*------------------------------------------
+ * How the page is built (AGL-4106)
+ *-----------------------------------------*/ const SVG_ELEMENT = /<svg\b((?:[^>"']|"[^"]*"|'[^']*')*)>([\s\S]*?)<\/svg\s*>/gi;
+const SVG_SHAPE = /<(path|circle|rect|ellipse|line|polyline|polygon)\b((?:[^>"']|"[^"]*"|'[^']*')*)>/gi;
+/** What a drawing keeps of a shape: its geometry and paint, never a link, a script or a reference. */ const SVG_SHAPE_ATTRIBUTES = new Set([
+    'd',
+    'cx',
+    'cy',
+    'r',
+    'rx',
+    'ry',
+    'x',
+    'y',
+    'x1',
+    'y1',
+    'x2',
+    'y2',
+    'width',
+    'height',
+    'points',
+    'fill',
+    'fill-rule',
+    'clip-rule',
+    'fill-opacity',
+    'stroke',
+    'stroke-width',
+    'stroke-linecap',
+    'stroke-linejoin',
+    'opacity',
+    'transform'
+]);
+const SVG_SAFE_VALUE = /^[\w\s.,#%()+-]*$/;
+const SVG_PATH_DATA = /^[MmZzLlHhVvCcSsQqTtAa0-9eE.,\s+-]+$/;
+const VIEW_BOX = /^-?[\d.]+(?:[\s,]+-?[\d.]+){3}$/;
+/** An `<svg>`'s drawing area: its `viewBox`, else its size, else the 24-unit square of an icon set. */ function viewBoxOf(attributes) {
+    var _attributes_viewbox, _attributes_width, _attributes_height;
+    const written = ((_attributes_viewbox = attributes['viewbox']) != null ? _attributes_viewbox : '').trim().replace(/[\s,]+/g, ' ');
+    if (VIEW_BOX.test(written)) return written;
+    const width = parseFloat((_attributes_width = attributes['width']) != null ? _attributes_width : '');
+    const height = parseFloat((_attributes_height = attributes['height']) != null ? _attributes_height : '');
+    return width > 0 && height > 0 ? `0 0 ${width} ${height}` : '0 0 24 24';
+}
+/**
+ * Every inline `<svg>` turned into a marker the walk reads as an icon at its
+ * place, its drawing kept in `icons` once however often it repeats. A drawing
+ * that embeds anything but shapes (a picture, text, a reference) is no icon
+ * and goes, as every `<svg>` did before.
+ */ function readSvgIcons(html, icons) {
+    const seen = new Map();
+    return html.replace(SVG_ELEMENT, (whole, attributeSource, inner)=>{
+        var _icon_path;
+        if (/<(?:image|foreignobject|use|text|script)\b/i.test(inner)) return ' ';
+        const shapes = [
+            ...inner.matchAll(SVG_SHAPE)
+        ];
+        if (!shapes.length) return ' ';
+        const viewBox = viewBoxOf(parseAttributes(attributeSource));
+        let icon = null;
+        if (shapes.length === 1 && shapes[0][1].toLowerCase() === 'path') {
+            var _parseAttributes_d;
+            const d = ((_parseAttributes_d = parseAttributes(shapes[0][2])['d']) != null ? _parseAttributes_d : '').trim();
+            if (d && d.length <= 4096 && SVG_PATH_DATA.test(d)) icon = {
+                viewBox,
+                path: d
+            };
+        }
+        if (!icon && whole.length <= 16384 && shapes.length <= 64) {
+            const drawn = shapes.map(([, tag, source])=>{
+                const kept = Object.entries(parseAttributes(source)).filter(([name, value])=>SVG_SHAPE_ATTRIBUTES.has(name) && SVG_SAFE_VALUE.test(value) && !/url\s*\(/i.test(value)).map(([name, value])=>`${name}="${value}"`);
+                return `<${tag.toLowerCase()}${kept.length ? ` ${kept.join(' ')}` : ''}/>`;
+            });
+            icon = {
+                viewBox,
+                svg: `<svg xmlns="http://www.w3.org/2000/svg" viewBox="${viewBox}">${drawn.join('')}</svg>`
+            };
+        }
+        if (!icon) return ' ';
+        const key = `${icon.viewBox}|${(_icon_path = icon.path) != null ? _icon_path : icon.svg}`;
+        let index = seen.get(key);
+        if (index === undefined) {
+            if (icons.length >= SITE_IMPORT_PAGE_LIMITS.svgIcons) return ' ';
+            index = icons.length;
+            icons.push(icon);
+            seen.set(key, index);
+        }
+        return ` <aglyn-svg data-icon="${index}"> `;
+    });
+}
+/** The page's own `<style>` sheets and the stylesheets it links, print ones left out. */ function readStyleSheets(html, pageUrl) {
+    const styles = [];
+    let chars = 0;
+    for (const match of html.matchAll(/<style\b((?:[^>"']|"[^"]*"|'[^']*')*)>([\s\S]*?)<\/style\s*>/gi)){
+        var _parseAttributes_media, _match_;
+        const media = ((_parseAttributes_media = parseAttributes(match[1])['media']) != null ? _parseAttributes_media : '').toLowerCase();
+        if (media && !/\b(?:all|screen)\b/.test(media)) continue;
+        const css = (_match_ = match[2]) != null ? _match_ : '';
+        if (!css.trim() || chars + css.length > SITE_IMPORT_PAGE_LIMITS.styleChars) continue;
+        chars += css.length;
+        styles.push(css);
+    }
+    const stylesheets = [];
+    for (const match of html.matchAll(/<link\b((?:[^>"']|"[^"]*"|'[^']*')*)>/gi)){
+        var _attributes_rel, _attributes_media;
+        const attributes = parseAttributes(match[1]);
+        if (!((_attributes_rel = attributes['rel']) != null ? _attributes_rel : '').toLowerCase().split(/\s+/).includes('stylesheet')) continue;
+        const media = ((_attributes_media = attributes['media']) != null ? _attributes_media : '').toLowerCase();
+        if (media && !/\b(?:all|screen)\b/.test(media)) continue;
+        const url = resolveSiteUrl(attributes['href'], pageUrl);
+        if (!url || url.protocol !== 'https:' || stylesheets.includes(url.toString())) continue;
+        stylesheets.push(url.toString());
+        if (stylesheets.length >= SITE_IMPORT_PAGE_LIMITS.stylesheets) break;
+    }
+    return {
+        styles,
+        stylesheets
+    };
+}
+/** Elements with no content and no end tag. */ const VOID_TAGS = new Set([
+    'area',
+    'base',
+    'br',
+    'col',
+    'embed',
+    'hr',
+    'img',
+    'input',
+    'link',
+    'meta',
+    'param',
+    'source',
+    'track',
+    'wbr',
+    'aglyn-svg',
+    'aglyn-form'
+]);
+/** Start tags that end an open `<p>`, as a browser reads them. */ const CLOSES_P = new Set([
+    'address',
+    'article',
+    'aside',
+    'blockquote',
+    'details',
+    'div',
+    'dl',
+    'fieldset',
+    'figcaption',
+    'figure',
+    'footer',
+    'form',
+    'h1',
+    'h2',
+    'h3',
+    'h4',
+    'h5',
+    'h6',
+    'header',
+    'hr',
+    'main',
+    'menu',
+    'nav',
+    'ol',
+    'p',
+    'pre',
+    'section',
+    'table',
+    'ul'
+]);
+/** For each tag that ends an open one of its own kind: the kinds it ends, and the tags that wall the search off. */ const IMPLIED_END = {
+    li: {
+        ends: new Set([
+            'li'
+        ]),
+        walls: new Set([
+            'ul',
+            'ol',
+            'menu'
+        ])
+    },
+    dt: {
+        ends: new Set([
+            'dt',
+            'dd'
+        ]),
+        walls: new Set([
+            'dl'
+        ])
+    },
+    dd: {
+        ends: new Set([
+            'dt',
+            'dd'
+        ]),
+        walls: new Set([
+            'dl'
+        ])
+    },
+    tr: {
+        ends: new Set([
+            'tr',
+            'td',
+            'th'
+        ]),
+        walls: new Set([
+            'table',
+            'tbody',
+            'thead',
+            'tfoot'
+        ])
+    },
+    td: {
+        ends: new Set([
+            'td',
+            'th'
+        ]),
+        walls: new Set([
+            'tr',
+            'table'
+        ])
+    },
+    th: {
+        ends: new Set([
+            'td',
+            'th'
+        ]),
+        walls: new Set([
+            'tr',
+            'table'
+        ])
+    },
+    option: {
+        ends: new Set([
+            'option'
+        ]),
+        walls: new Set([
+            'select',
+            'datalist'
+        ])
+    },
+    tbody: {
+        ends: new Set([
+            'tbody',
+            'thead',
+            'tfoot',
+            'tr',
+            'td',
+            'th'
+        ]),
+        walls: new Set([
+            'table'
+        ])
+    },
+    thead: {
+        ends: new Set([
+            'tbody',
+            'thead',
+            'tfoot',
+            'tr',
+            'td',
+            'th'
+        ]),
+        walls: new Set([
+            'table'
+        ])
+    },
+    tfoot: {
+        ends: new Set([
+            'tbody',
+            'thead',
+            'tfoot',
+            'tr',
+            'td',
+            'th'
+        ]),
+        walls: new Set([
+            'table'
+        ])
+    }
+};
+const P_WALLS = new Set([
+    'button',
+    'table',
+    'td',
+    'th',
+    'caption',
+    'object',
+    'template',
+    'html',
+    'marquee',
+    'applet'
+]);
+/** The attributes the layout reader reads. */ function elementAttrs(attributes) {
+    var _attributes_datasectiontheme, _attributes_ariahidden;
+    var _text, _text1, _text2;
+    const attrs = {};
+    const text = (value, max)=>{
+        const clean = String(value != null ? value : '').replace(/\s+/g, ' ').trim();
+        return clean ? clean.slice(0, max) : undefined;
+    };
+    const put = (key, value)=>{
+        if (value !== undefined) attrs[key] = value;
+    };
+    put('class', text(attributes['class'], 600));
+    put('id', text(attributes['id'], 120));
+    put('style', text(attributes['style'], 1000));
+    put('role', (_text = text(attributes['role'], 40)) == null ? void 0 : _text.toLowerCase());
+    put('label', text(attributes['aria-label'], 200));
+    put('href', text(attributes['href'], 2048));
+    put('theme', (_text1 = text((_attributes_datasectiontheme = attributes['data-section-theme']) != null ? _attributes_datasectiontheme : attributes['data-theme'], 40)) == null ? void 0 : _text1.toLowerCase());
+    put('type', (_text2 = text(attributes['type'], 40)) == null ? void 0 : _text2.toLowerCase());
+    put('controls', text(attributes['aria-controls'], 120));
+    put('labelledby', text(attributes['aria-labelledby'], 120));
+    if ('hidden' in attributes || ((_attributes_ariahidden = attributes['aria-hidden']) != null ? _attributes_ariahidden : '').toLowerCase() === 'true') attrs.hidden = true;
+    return attrs;
+}
 /**
  * The words, structure, pictures and links of one page.
  *
  * `isSameSite(url)` says whether an address is on the site being imported;
  * only those become `links`, and images are marked by it.
- */ export function extractSiteImportPage(html, pageUrl, isSameSite) {
+ */ export function extractSiteImportPage(html, pageUrl, isSameSite, options = {}) {
     var _ref, _ref1;
     var _resolveSiteUrl, _parseAttributes_lang;
     const empty = {
@@ -465,6 +793,14 @@ const ORGANIZATION_TYPE = /(?:^|\b)(?:Organization|Corporation|LocalBusiness|NGO
     // Read before scripts go: the structured data is a script's body.
     const organization = readOrganization(source);
     const icons = readIcons(source, pageUrl);
+    const keep = options.structure === true;
+    const sheets = keep ? readStyleSheets(source, pageUrl) : {
+        styles: [],
+        stylesheets: []
+    };
+    // An inline icon is a marker at its place (AGL-4106); its drawing is never words.
+    const svgIcons = [];
+    source = readSvgIcons(source, svgIcons);
     source = removeElements(readPictures(source), REMOVED_WITH_CONTENT).replace(REMOVED_VOID, ' ');
     // Forms next, so a field list is read before the controls go.
     const forms = takeForms(source);
@@ -513,24 +849,112 @@ const ORGANIZATION_TYPE = /(?:^|\b)(?:Organization|Corporation|LocalBusiness|NGO
     let buffer = '';
     /** The link being read: its menu address (http(s) only), where it goes, its words and its first picture. */ let anchor = null;
     /** The links read into the block being built, written after it. */ let pending = [];
+    // How the page is built (AGL-4106): every element, in a flat table, and
+    // the element each block's words sit in. Kept only when asked for.
+    const elements = [];
+    const blockElement = [];
+    const anchors = [];
+    const texts = [];
+    /** Where in `buffer` the words no link has taken begin: a menu's own words are read from here. */ let ownFrom = 0;
+    /** A menu's or a header's own words, outside its links (AGL-4106), from `ownFrom` on. */ const takeOwnWords = (element)=>{
+        if (!keep || element < 0 || !(navDepth > 0 || headerDepth > 0) || texts.length >= SITE_IMPORT_PAGE_LIMITS.anchors) return;
+        const words = clip(collapse(decodeHtmlEntities(buffer.slice(ownFrom))), 300);
+        if (words) texts.push({
+            element,
+            text: words
+        });
+    };
+    /** The elements open where the walk is, innermost last. */ const open = [];
+    let mainElement = -1;
+    let roleMain = -1;
+    let bodyElement = -1;
+    /** The element the loose words being read began in. */ let looseElement = -1;
+    const top = ()=>open.length ? open[open.length - 1] : -1;
     const region = ()=>footerDepth > 0 ? 'footer' : headerDepth > 0 ? 'header' : 'main';
     // A site without a <nav> keeps its menu in the header; its links are the nav.
     const inNav = ()=>navDepth > 0 || !hasNav && headerDepth > 0;
     const room = ()=>blocks.length < SITE_IMPORT_PAGE_LIMITS.blocks;
+    /** A block, and where its words sit. */ const push = (block, element)=>{
+        blocks.push(block);
+        if (!keep) return;
+        blockElement.push(element);
+        const owner = elements[element];
+        if (!owner) return;
+        owner.blocks.push(blocks.length - 1);
+        if (owner.order === undefined) owner.order = blocks.length - 1;
+    };
+    /** Pops the open elements down to the innermost `tags` one, unless a wall stands first. */ const endOpen = (tags, walls)=>{
+        for(let at = open.length - 1; at >= 0; at -= 1){
+            var _ref;
+            var _elements_open_at;
+            const tag = (_ref = (_elements_open_at = elements[open[at]]) == null ? void 0 : _elements_open_at.tag) != null ? _ref : '';
+            if (tags.has(tag)) {
+                open.length = at;
+                return;
+            }
+            if (walls.has(tag)) return;
+        }
+    };
+    const P_ONLY = new Set([
+        'p'
+    ]);
+    /** An element opens: what it implies is closed first, then it is the innermost. */ const openElement = (tag, attributes, selfClosing)=>{
+        var _attributes_role;
+        var _elements_parent;
+        if (!keep) return -1;
+        if (CLOSES_P.has(tag)) endOpen(P_ONLY, P_WALLS);
+        const implied = IMPLIED_END[tag];
+        if (implied) endOpen(implied.ends, implied.walls);
+        const isVoid = VOID_TAGS.has(tag);
+        if (isVoid && tag !== 'img' && tag !== 'aglyn-svg' && tag !== 'aglyn-form') return -1;
+        if (elements.length >= SITE_IMPORT_PAGE_LIMITS.elements) return -1;
+        const id = elements.length;
+        const parent = top();
+        elements.push({
+            id,
+            tag,
+            parent,
+            children: [],
+            region: region(),
+            nav: inNav(),
+            attrs: elementAttrs(attributes),
+            blocks: []
+        });
+        (_elements_parent = elements[parent]) == null ? void 0 : _elements_parent.children.push(id);
+        if (!isVoid && !selfClosing) open.push(id);
+        if (tag === 'body' && bodyElement < 0) bodyElement = id;
+        if (tag === 'main' && mainElement < 0) mainElement = id;
+        if (roleMain < 0 && ((_attributes_role = attributes['role']) != null ? _attributes_role : '').toLowerCase() === 'main') roleMain = id;
+        return id;
+    };
+    /** An element closes: it and everything still open inside it. */ const closeElement = (tag)=>{
+        if (!keep) return;
+        for(let at = open.length - 1; at >= 0; at -= 1){
+            var _elements_open_at;
+            if (((_elements_open_at = elements[open[at]]) == null ? void 0 : _elements_open_at.tag) === tag) {
+                open.length = at;
+                return;
+            }
+        }
+    };
     const flush = ()=>{
-        var _ref;
+        var _ref, _ref1;
         const raw = buffer;
         const written = pending;
+        const loose = current === null;
+        const element = (_ref = current == null ? void 0 : current.element) != null ? _ref : looseElement;
+        if (!anchor) takeOwnWords(element);
+        ownFrom = 0;
         buffer = '';
         pending = [];
+        looseElement = -1;
         if (anchor && collapse(decodeHtmlEntities(raw.slice(anchor.start)))) anchor.crossed = true;
         if (anchor) anchor.start = 0;
-        const kind = (_ref = current == null ? void 0 : current.kind) != null ? _ref : 'paragraph';
+        const kind = (_ref1 = current == null ? void 0 : current.kind) != null ? _ref1 : 'paragraph';
         const level = current == null ? void 0 : current.level;
-        const loose = current === null;
         current = null;
-        if (navDepth > 0) return;
         const text = clip(collapse(decodeHtmlEntities(raw)), SITE_IMPORT_PAGE_LIMITS.blockChars);
+        if (navDepth > 0) return;
         // Loose words that are only links (a row of buttons, an address in a
         // card) are those links, each a block of its own; words around a link
         // keep their block, and the link follows it.
@@ -543,18 +967,18 @@ const ORGANIZATION_TYPE = /(?:^|\b)(?:Organization|Corporation|LocalBusiness|NGO
             const last = blocks[blocks.length - 1];
             if (!(last && last.kind !== 'image' && last.kind !== 'link' && last.text === text)) {
                 if (kind === 'heading') {
-                    blocks.push({
+                    push({
                         kind,
                         level: level != null ? level : 2,
                         text,
                         region: region()
-                    });
+                    }, element);
                 } else {
-                    blocks.push({
+                    push({
                         kind,
                         text,
                         region: region()
-                    });
+                    }, element);
                 }
             }
         }
@@ -563,14 +987,14 @@ const ORGANIZATION_TYPE = /(?:^|\b)(?:Organization|Corporation|LocalBusiness|NGO
             const last = blocks[blocks.length - 1];
             if ((last == null ? void 0 : last.kind) === 'link' && last.href === link.href && last.text === link.text) continue;
             const inline = !onlyLinks || link.crossed;
-            blocks.push(_extends({
+            push(_extends({
                 kind: 'link',
                 text: link.text,
                 href: link.href,
                 region: link.region
             }, inline ? {
                 inline: true
-            } : {}));
+            } : {}), link.element);
         }
     };
     /** A boundary between two runs of words: in the block, and in the link being read. */ const boundary = (block)=>{
@@ -579,22 +1003,9 @@ const ORGANIZATION_TYPE = /(?:^|\b)(?:Organization|Corporation|LocalBusiness|NGO
         anchor.text += RUN_BREAK;
         if (block && anchor.lines[anchor.lines.length - 1] !== '') anchor.lines.push('');
     };
-    for (const match of body.matchAll(new RegExp(`${TAG.source}|([^<]+)|<`, 'g'))){
-        var _match_;
-        const text = match[4];
-        if (text !== undefined || match[2] === undefined) {
-            const run = text != null ? text : '<';
-            buffer += run;
-            if (anchor) {
-                anchor.text += run;
-                anchor.lines[anchor.lines.length - 1] += run;
-            }
-            continue;
-        }
-        const closing = match[1] === '/';
-        const tag = match[2].toLowerCase();
-        const attributes = closing ? {} : parseAttributes((_match_ = match[3]) != null ? _match_ : '');
+    /** One tag of the page, read where the walk is. */ const read = (closing, tag, attributes, element)=>{
         if (tag === 'a') {
+            var _ref;
             if (closing) {
                 if (anchor) {
                     var _anchor_lines_find;
@@ -620,14 +1031,33 @@ const ORGANIZATION_TYPE = /(?:^|\b)(?:Organization|Corporation|LocalBusiness|NGO
                             start: anchor.start,
                             end: buffer.length,
                             crossed: anchor.crossed,
-                            region: anchor.region
+                            region: anchor.region,
+                            element: anchor.element
                         });
                     }
+                    if (keep && anchor.element >= 0 && anchors.length < SITE_IMPORT_PAGE_LIMITS.anchors) {
+                        const lines = anchor.lines.map((line)=>clip(collapse(decodeHtmlEntities(line)), 300)).filter(Boolean);
+                        anchors.push(_extends({
+                            element: anchor.element,
+                            href: anchor.href,
+                            label: firstLine,
+                            text: clip(lines.slice(1).join(' '), 300),
+                            region: anchor.region,
+                            nav: inNav()
+                        }, anchor.icon !== undefined ? {
+                            icon: anchor.icon
+                        } : {}, anchor.image ? {
+                            image: anchor.image.url
+                        } : {}));
+                    }
                     anchor = null;
                 }
                 boundary(false);
-                continue;
+                ownFrom = buffer.length;
+                return;
             }
+            // Words before a link in a menu are its own: a column's heading.
+            if (!anchor) takeOwnWords((_ref = current == null ? void 0 : current.element) != null ? _ref : looseElement >= 0 ? looseElement : top());
             boundary(false);
             const url = resolveSiteUrl(attributes['href'], pageUrl);
             anchor = {
@@ -640,7 +1070,8 @@ const ORGANIZATION_TYPE = /(?:^|\b)(?:Organization|Corporation|LocalBusiness|NGO
                 image: null,
                 start: buffer.length,
                 crossed: false,
-                region: region()
+                region: region(),
+                element
             };
             if (url && isSameSite(url) && links.length < SITE_IMPORT_PAGE_LIMITS.links) {
                 const key = url.toString();
@@ -649,7 +1080,27 @@ const ORGANIZATION_TYPE = /(?:^|\b)(?:Organization|Corporation|LocalBusiness|NGO
                     links.push(key);
                 }
             }
-            continue;
+            return;
+        }
+        if (tag === 'aglyn-svg') {
+            // An icon at its place: never words (AGL-4106).
+            const index = Number(attributes['data-icon']);
+            if (keep && element >= 0 && Number.isInteger(index) && svgIcons[index]) {
+                const own = elements[element];
+                own.icon = index;
+                own.order = blocks.length - 0.5;
+                if (anchor && anchor.icon === undefined) anchor.icon = index;
+            }
+            return;
+        }
+        if (tag === 'aglyn-form') {
+            const index = Number(attributes['data-form']);
+            if (keep && element >= 0 && Number.isInteger(index)) {
+                const own = elements[element];
+                own.form = index;
+                own.order = blocks.length - 0.5;
+            }
+            return;
         }
         if (tag === 'img' && !closing) {
             const url = resolveSiteUrl(attributes['src'] || attributes['data-src'] || firstSrcset(attributes['srcset'] || attributes['data-srcset']), pageUrl);
@@ -672,16 +1123,16 @@ const ORGANIZATION_TYPE = /(?:^|\b)(?:Organization|Corporation|LocalBusiness|NGO
                     });
                     if (navDepth === 0) {
                         flush();
-                        if (room()) blocks.push({
+                        if (room()) push({
                             kind: 'image',
                             url: key,
                             alt,
                             region: region()
-                        });
+                        }, element);
                     }
                 }
             }
-            continue;
+            return;
         }
         if (tag === 'nav' || tag === 'header' || tag === 'footer') {
             flush();
@@ -690,41 +1141,69 @@ const ORGANIZATION_TYPE = /(?:^|\b)(?:Organization|Corporation|LocalBusiness|NGO
             if (tag === 'nav') navDepth = Math.max(0, navDepth + step);
             if (tag === 'header') headerDepth = Math.max(0, headerDepth + step);
             if (tag === 'footer') footerDepth = Math.max(0, footerDepth + step);
-            continue;
+            // The landmark is in its own region.
+            const own = elements[element];
+            if (own) {
+                own.region = region();
+                own.nav = inNav();
+            }
+            return;
         }
         const heading = /^h([1-6])$/.exec(tag);
         if (heading || BLOCK_TAGS[tag]) {
             if (closing) {
                 if ((current == null ? void 0 : current.tag) === tag) flush();
                 boundary(true);
-                continue;
+                return;
             }
             flush();
             boundary(true);
             current = heading ? {
                 kind: 'heading',
                 level: Number(heading[1]),
-                tag
+                tag,
+                element
             } : {
                 kind: BLOCK_TAGS[tag],
-                tag
+                tag,
+                element
             };
-            continue;
+            return;
         }
         if (BREAK_TAGS.has(tag)) {
             if (tag === 'br' && current) {
                 buffer += ' ';
                 if (anchor) anchor.text += ' ';
-                continue;
+                return;
             }
             flush();
             boundary(true);
-            continue;
+            return;
         }
         if (RUN_BOUNDARY.has(tag)) boundary(false);
+    };
+    for (const match of body.matchAll(new RegExp(`${TAG.source}|([^<]+)|<`, 'g'))){
+        var _match_, _match_1;
+        const text = match[4];
+        if (text !== undefined || match[2] === undefined) {
+            const run = text != null ? text : '<';
+            buffer += run;
+            if (keep && looseElement < 0 && current === null && run.trim()) looseElement = top();
+            if (anchor) {
+                anchor.text += run;
+                anchor.lines[anchor.lines.length - 1] += run;
+            }
+            continue;
+        }
+        const closing = match[1] === '/';
+        const tag = match[2].toLowerCase();
+        const attributes = closing ? {} : parseAttributes((_match_ = match[3]) != null ? _match_ : '');
+        const element = closing ? -1 : openElement(tag, attributes, /\/\s*$/.test((_match_1 = match[3]) != null ? _match_1 : ''));
+        read(closing, tag, attributes, element);
+        if (closing) closeElement(tag);
     }
     flush();
-    return {
+    return _extends({
         title,
         description,
         lang,
@@ -743,7 +1222,20 @@ const ORGANIZATION_TYPE = /(?:^|\b)(?:Organization|Corporation|LocalBusiness|NGO
             removed: forms.removed
         },
         loginForm: false
-    };
+    }, keep ? {
+        structure: {
+            elements,
+            blockElement,
+            icons: svgIcons,
+            anchors,
+            texts,
+            styles: sheets.styles,
+            stylesheets: sheets.stylesheets,
+            main: mainElement >= 0 ? mainElement : roleMain,
+            body: bodyElement,
+            root: htmlTag ? elementAttrs(parseAttributes(htmlTag[1])) : {}
+        }
+    } : {});
 }
 
 //# sourceMappingURL=site-import-html.js.map