@wgtechlabs/mdd-engine 0.1.0-pr.08f3b80 → 0.1.0-pr.0cb765a

package/dist/search.js

@@ -70,11 +70,76 @@ function normalize(value) {
         .replace(/[^\p{L}\p{N}\p{M}_]+/gu, " ")
         .trim();
 }
-function excerpt(value) {
-    const characters = Array.from(value.replace(/\s+/gu, " ").trim());
-    return characters.length > 160
-        ? `${characters.slice(0, 159).join("")}…`
-        : characters.join("");
+/** Map a normalized match back to the original text, including NFKC expansions. */
+function matchRange(value, terms) {
+    const normalized = value.normalize("NFKC").toLowerCase();
+    let start = -1;
+    let end = -1;
+    for (const term of terms) {
+        const found = normalized.indexOf(term);
+        if (found >= 0 &&
+            (start < 0 ||
+                found < start ||
+                (found === start && found + term.length > end))) {
+            start = found;
+            end = found + term.length;
+        }
+    }
+    if (start < 0)
+        return;
+    const groups = [];
+    for (const { segment, index } of new Intl.Segmenter(undefined, {
+        granularity: "grapheme",
+    }).segment(value)) {
+        let group = {
+            text: segment.normalize("NFKC"),
+            start: index,
+            end: index + segment.length,
+        };
+        // NFKC can combine adjacent graphemes, e.g. compatibility Hangul ㄱㅏ → 가.
+        while (groups.length) {
+            const previous = groups.at(-1);
+            if (!previous)
+                break;
+            const joined = previous.text + group.text;
+            const combined = joined.normalize("NFKC");
+            if (combined === joined)
+                break;
+            groups.pop();
+            group = { text: combined, start: previous.start, end: group.end };
+        }
+        groups.push(group);
+    }
+    // Match against whole-field lowercase for contextual letters such as sigma.
+    // Per-group lowercase is used only for lengths (including expansions like İ).
+    let offset = 0;
+    let originalStart = 0;
+    for (const group of groups) {
+        const next = offset + group.text.toLowerCase().length;
+        if (offset <= start && start < next)
+            originalStart = group.start;
+        if (offset < end && end <= next)
+            return [originalStart, group.end];
+        offset = next;
+    }
+}
+function excerpt(value, terms) {
+    const text = value.replace(/\s+/gu, " ").trim();
+    const characters = Array.from(text);
+    if (characters.length <= 160)
+        return text;
+    const match = matchRange(text, terms);
+    const matchStart = match ? Array.from(text.slice(0, match[0])).length : 0;
+    const matchLength = match
+        ? Array.from(text.slice(match[0], match[1])).length
+        : 0;
+    // Reserve room for both ellipses and the matched term before adding context.
+    const context = Math.min(40, Math.max(0, 158 - matchLength));
+    const start = Math.min(Math.max(0, matchStart - context), characters.length - 159);
+    let end = Math.min(characters.length, start + 160 - (start > 0 ? 1 : 0));
+    if (end < characters.length)
+        end--;
+    return `${start > 0 ? "…" : ""}${characters.slice(start, end).join("")}${end < characters.length ? "…" : ""}`;
 }
 /** Return at most one hit per page, without filesystem, network, or DOM access. */
 export function search(index, query, options = {}) {
@@ -132,14 +197,23 @@ export function search(index, query, options = {}) {
         matches.sort((a, b) => b.score - a.score);
         const best = matches[0]?.section;
         const destination = metadataMatch ? undefined : best;
+        const sources = [
+            best?.text ?? "",
+            page.description,
+            ...page.sections.map((section) => section.text),
+        ];
+        const source = sources.find((text) => {
+            const normalized = normalize(text);
+            return terms.some((term) => normalized.includes(term));
+        }) ??
+            sources.find((text) => text) ??
+            page.title;
         results.push({
             title: page.title,
             url: destination?.url ?? page.url,
             ...(destination?.title ? { section: destination.title } : {}),
-            excerpt: excerpt(best?.text ||
-                page.description ||
-                page.sections.find((section) => section.text)?.text ||
-                page.title),
+            // Keep source text until ranking so only returned hits need a snippet.
+            excerpt: source,
             score: weights.reduce((sum, weight) => sum + weight, 0) +
                 (title === phrase
                     ? 24
@@ -149,5 +223,8 @@ export function search(index, query, options = {}) {
         });
     }
     results.sort((a, b) => b.score - a.score || (a.url < b.url ? -1 : a.url > b.url ? 1 : 0));
-    return results.slice(0, limit);
+    return results.slice(0, limit).map((result) => ({
+        ...result,
+        excerpt: excerpt(result.excerpt, terms),
+    }));
 }

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@wgtechlabs/mdd-engine",
-  "version": "0.1.0-pr.08f3b80",
+  "version": "0.1.0-pr.0cb765a",
   "description": "Headless Markdown documentation compiler for mdd",
   "type": "module",
   "license": "MIT",