@superbee/core 0.2.0-pre.4 → 0.2.0-pre.41

package/dist/search/index.d.ts.map

@@ -0,0 +1 @@
+{"version":3,"file":"index.d.ts","sourceRoot":"","sources":["../../src/search/index.ts"],"names":[],"mappings":"AAAA;;;;;;GAMG;AACH,OAAO,EACL,mBAAmB,EACnB,gBAAgB,EAChB,aAAa,EACb,eAAe,EACf,aAAa,EACb,cAAc,EACd,mBAAmB,EACnB,uBAAuB,EACvB,mBAAmB,EACnB,oBAAoB,EACpB,KAAK,gBAAgB,EACrB,KAAK,aAAa,EAClB,KAAK,kBAAkB,EACvB,KAAK,eAAe,EACpB,KAAK,WAAW,EAChB,KAAK,kBAAkB,EACvB,KAAK,iBAAiB,EACtB,KAAK,aAAa,EAClB,KAAK,eAAe,EACpB,KAAK,cAAc,EACnB,KAAK,YAAY,GAClB,MAAM,eAAe,CAAC;AACvB,OAAO,EACL,gBAAgB,EAChB,qBAAqB,EACrB,gBAAgB,EAChB,gBAAgB,EAChB,KAAK,iBAAiB,GACvB,MAAM,aAAa,CAAC;AACrB,OAAO,EACL,SAAS,EACT,cAAc,EACd,MAAM,EACN,UAAU,EACV,QAAQ,EACR,KAAK,iBAAiB,EACtB,KAAK,cAAc,GACpB,MAAM,WAAW,CAAC;AACnB,OAAO,EACL,mBAAmB,EACnB,qBAAqB,EACrB,aAAa,EACb,WAAW,EACX,KAAK,eAAe,GACrB,MAAM,eAAe,CAAC;AACvB,YAAY,EACV,kBAAkB,EAClB,qBAAqB,EACrB,sBAAsB,EACtB,gBAAgB,EAChB,YAAY,EACZ,YAAY,GACb,MAAM,aAAa,CAAC;AACrB,OAAO,EAAE,kBAAkB,EAAE,KAAK,2BAA2B,EAAE,KAAK,oBAAoB,EAAE,MAAM,oBAAoB,CAAC;AACrH,OAAO,EAAE,eAAe,EAAE,qBAAqB,EAAE,KAAK,eAAe,EAAE,KAAK,aAAa,EAAE,MAAM,aAAa,CAAC;AAE/G,OAAO,EAAE,uBAAuB,EAAE,MAAM,qBAAqB,CAAC"}

package/dist/search/index.js

@@ -0,0 +1,15 @@
+/**
+ * `@superbee/core/search`: the one `documents.search.v1` engine. The contract, query rules,
+ * ranking, the any-term fallback, facets, cursors and page budgets live here; a backend supplies
+ * what it searches through the `SearchSource` seam, and `memorySearchSource` indexes a complete
+ * corpus in memory. Pure ECMAScript with no dependency, so a Worker, a browser page and the CLI run
+ * the same code. Ranking is portal-webmcp 0.3.1's (`portal-webmcp-search/v1`).
+ */
+export { BODY_OCCURRENCE_CAP, SEARCH_ALGORITHM, SEARCH_FACETS, SEARCH_GUIDANCE, SEARCH_LIMITS, SEARCH_MATCHES, SEARCH_MATCH_FIELDS, SEARCH_MAX_OUTPUT_BYTES, SEARCH_OPERATION_ID, SEARCH_TEXT_REVISION, } from "./contract.js";
+export { SearchAbortError, SearchBroadQueryError, SearchInputError, SearchLimitError, } from "./errors.js";
+export { normalize, normalizeQuery, phrase, relaxQuery, tokenize, } from "./text.js";
+export { documentSourceBytes, normalizerFingerprint, searchRowText, textVersion, } from "./row-text.js";
+export { memorySearchSource } from "./memory-source.js";
+export { searchDocuments, validateSearchRequest } from "./search.js";
+export { searchSourceFromBackend } from "./backend-source.js";
+//# sourceMappingURL=index.js.map

package/dist/search/index.js.map

@@ -0,0 +1 @@
+{"version":3,"file":"index.js","sourceRoot":"","sources":["../../src/search/index.ts"],"names":[],"mappings":"AAAA;;;;;;GAMG;AACH,OAAO,EACL,mBAAmB,EACnB,gBAAgB,EAChB,aAAa,EACb,eAAe,EACf,aAAa,EACb,cAAc,EACd,mBAAmB,EACnB,uBAAuB,EACvB,mBAAmB,EACnB,oBAAoB,GAYrB,MAAM,eAAe,CAAC;AACvB,OAAO,EACL,gBAAgB,EAChB,qBAAqB,EACrB,gBAAgB,EAChB,gBAAgB,GAEjB,MAAM,aAAa,CAAC;AACrB,OAAO,EACL,SAAS,EACT,cAAc,EACd,MAAM,EACN,UAAU,EACV,QAAQ,GAGT,MAAM,WAAW,CAAC;AACnB,OAAO,EACL,mBAAmB,EACnB,qBAAqB,EACrB,aAAa,EACb,WAAW,GAEZ,MAAM,eAAe,CAAC;AASvB,OAAO,EAAE,kBAAkB,EAA+D,MAAM,oBAAoB,CAAC;AACrH,OAAO,EAAE,eAAe,EAAE,qBAAqB,EAA4C,MAAM,aAAa,CAAC;AAE/G,OAAO,EAAE,uBAAuB,EAAE,MAAM,qBAAqB,CAAC"}

package/dist/search/memory-source.d.ts

@@ -0,0 +1,36 @@
+import { type SearchDocumentV1, type SearchStopV1 } from "./contract.js";
+import type { SearchSource } from "./source.js";
+export interface MemorySearchSourceOptionsV1 {
+    /** Identifies this exact corpus (a snapshot digest, a heads digest); cursors are bound to it. */
+    readonly key: string;
+    /** A host's own bound on the corpus's source bytes, at most SEARCH_LIMITS.maxIndexSourceBytes. */
+    readonly maxSourceBytes?: number;
+    /** A host's own bound on the corpus's documents, at most SEARCH_LIMITS.maxIndexDocuments. */
+    readonly maxDocuments?: number;
+    /** Stops the build. */
+    readonly stop?: SearchStopV1;
+}
+export interface MemorySearchSourceV1 extends SearchSource {
+    readonly key: string;
+    /** Identity of the built index; two identical corpora index identically. Computed on first read. */
+    readonly indexDigest: `sha256:${string}`;
+}
+/** Shared bounds for direct and backend-fed memory corpora. */
+export declare function memorySearchBounds(options: Omit<MemorySearchSourceOptionsV1, "key">): {
+    maxDocuments: {
+        name: string;
+        max: number;
+    };
+    maxSourceBytes: {
+        name: string;
+        max: number;
+    };
+};
+/**
+ * Builds the in-memory source over a complete corpus. The corpus limits (SEARCH_LIMITS: documents,
+ * source bytes) and the host's bounds throw `SearchLimitError`; a document over the per-document
+ * term cap is skipped and counted; two documents with one id throw `TypeError`; a stop throws as
+ * its check does.
+ */
+export declare function memorySearchSource(documents: readonly SearchDocumentV1[], options: MemorySearchSourceOptionsV1): MemorySearchSourceV1;
+//# sourceMappingURL=memory-source.d.ts.map

package/dist/search/memory-source.d.ts.map

@@ -0,0 +1 @@
+{"version":3,"file":"memory-source.d.ts","sourceRoot":"","sources":["../../src/search/memory-source.ts"],"names":[],"mappings":"AAUA,OAAO,EASL,KAAK,gBAAgB,EAGrB,KAAK,YAAY,EAClB,MAAM,eAAe,CAAC;AAIvB,OAAO,KAAK,EAIV,YAAY,EAEb,MAAM,aAAa,CAAC;AAiCrB,MAAM,WAAW,2BAA2B;IAC1C,iGAAiG;IACjG,QAAQ,CAAC,GAAG,EAAE,MAAM,CAAC;IACrB,kGAAkG;IAClG,QAAQ,CAAC,cAAc,CAAC,EAAE,MAAM,CAAC;IACjC,6FAA6F;IAC7F,QAAQ,CAAC,YAAY,CAAC,EAAE,MAAM,CAAC;IAC/B,uBAAuB;IACvB,QAAQ,CAAC,IAAI,CAAC,EAAE,YAAY,CAAC;CAC9B;AAED,MAAM,WAAW,oBAAqB,SAAQ,YAAY;IACxD,QAAQ,CAAC,GAAG,EAAE,MAAM,CAAC;IACrB,oGAAoG;IACpG,QAAQ,CAAC,WAAW,EAAE,UAAU,MAAM,EAAE,CAAC;CAC1C;AAaD,+DAA+D;AAC/D,wBAAgB,kBAAkB,CAAC,OAAO,EAAE,IAAI,CAAC,2BAA2B,EAAE,KAAK,CAAC;;;;;;;;;EASnF;AAqBD;;;;;GAKG;AACH,wBAAgB,kBAAkB,CAChC,SAAS,EAAE,SAAS,gBAAgB,EAAE,EACtC,OAAO,EAAE,2BAA2B,GACnC,oBAAoB,CAwGtB"}

package/dist/search/memory-source.js

@@ -0,0 +1,235 @@
+/**
+ * The in-memory source: a complete corpus indexed once, answering evidence by walking its rows.
+ *
+ * The build is portal-webmcp 0.3.1's (`buildSearchIndex`): field term sets and a positional body
+ * index holding every occurrence's original byte range, so the lazy `indexDigest` is Portal's
+ * index identity byte for byte. A document over SEARCH_LIMITS.maxTermsPerDocument body terms is
+ * left out and counted as skipped; the corpus bounds still refuse the whole corpus. A host builds
+ * one source per corpus key and keeps it; nothing here re-reads a document after the build.
+ */
+import { sha256HexOfUtf8 } from "../sha256.js";
+import { BODY_OCCURRENCE_CAP, SEARCH_ALGORITHM, SEARCH_LIMITS, SIGNAL_INTERVAL, documentDescription, documentTags, documentTitle, documentType, } from "./contract.js";
+import { SearchLimitError, stopIfAsked } from "./errors.js";
+import { canonicalJson, codeUnitCompare } from "./identity.js";
+import { documentSourceBytes, fieldTokens } from "./row-text.js";
+import { scanBody } from "./text.js";
+function limitExceeded(limit, max, actual) {
+    throw new SearchLimitError(limit, max, actual);
+}
+/** A host bound, held to the engine's own ceiling; the binding limit names itself in a refusal. */
+function bound(name, value, ceilingName, ceiling) {
+    if (value === undefined)
+        return { name: ceilingName, max: ceiling };
+    if (!Number.isSafeInteger(value) || value < 1)
+        throw new TypeError(`${name} must be a positive integer`);
+    return value < ceiling ? { name, max: value } : { name: ceilingName, max: ceiling };
+}
+/** Shared bounds for direct and backend-fed memory corpora. */
+export function memorySearchBounds(options) {
+    const maxDocuments = bound("maxDocuments", options.maxDocuments, "maxIndexDocuments", SEARCH_LIMITS.maxIndexDocuments);
+    const maxSourceBytes = bound("maxSourceBytes", options.maxSourceBytes, "maxIndexSourceBytes", SEARCH_LIMITS.maxIndexSourceBytes);
+    return { maxDocuments, maxSourceBytes };
+}
+/** Every body occurrence's byte range by term, or nothing when the body is over the term cap. */
+function indexBody(body, maxTermsPerDocument) {
+    const terms = new Map();
+    const state = { count: 0, over: false };
+    scanBody(body, (term, start, end) => {
+        state.count += 1;
+        // Stop before retaining an over-budget occurrence.
+        if (state.count > maxTermsPerDocument) {
+            state.over = true;
+            return true;
+        }
+        const existing = terms.get(term);
+        if (existing)
+            existing.push({ start, end });
+        else
+            terms.set(term, [{ start, end }]);
+        return false;
+    });
+    return state.over ? undefined : terms;
+}
+/**
+ * Builds the in-memory source over a complete corpus. The corpus limits (SEARCH_LIMITS: documents,
+ * source bytes) and the host's bounds throw `SearchLimitError`; a document over the per-document
+ * term cap is skipped and counted; two documents with one id throw `TypeError`; a stop throws as
+ * its check does.
+ */
+export function memorySearchSource(documents, options) {
+    const key = options?.key;
+    if (typeof key !== "string" || key === "")
+        throw new TypeError("memorySearchSource needs a non-empty key");
+    const { maxDocuments, maxSourceBytes } = memorySearchBounds(options);
+    const stop = options.stop;
+    stopIfAsked(stop);
+    if (documents.length > maxDocuments.max)
+        limitExceeded(maxDocuments.name, maxDocuments.max, documents.length);
+    const rows = [];
+    const ids = new Set();
+    let sourceBytes = 0;
+    for (const [position, document] of documents.entries()) {
+        if (position % SIGNAL_INTERVAL === 0)
+            stopIfAsked(stop);
+        // text() answers by id, so two documents under one id would share one excerpt.
+        if (ids.has(document.id))
+            throw new TypeError("memorySearchSource needs one document per id");
+        ids.add(document.id);
+        // Check the total source budget before allocating postings for this document.
+        const documentBytes = documentSourceBytes(document);
+        if (sourceBytes + documentBytes > maxSourceBytes.max) {
+            limitExceeded(maxSourceBytes.name, maxSourceBytes.max, sourceBytes + documentBytes);
+        }
+        sourceBytes += documentBytes;
+        const body = document.body;
+        const indexed = indexBody(body, SEARCH_LIMITS.maxTermsPerDocument);
+        if (indexed === undefined)
+            continue;
+        const bodyTerms = new Map();
+        for (const [term, tokens] of indexed)
+            bodyTerms.set(term, Object.freeze(tokens));
+        const title = documentTitle(document);
+        const type = documentType(document);
+        const tags = documentTags(document);
+        const fields = fieldTokens(document);
+        const fieldInstances = new Map();
+        for (const tokens of [fields.id, fields.title, fields.type, fields.tag, fields.description])
+            for (const term of tokens)
+                fieldInstances.set(term, (fieldInstances.get(term) ?? 0) + 1);
+        rows.push({
+            indexedId: document.id,
+            indexedVersion: document.version,
+            idPhrase: fields.id.join(" "),
+            idSegments: new Set(fields.id),
+            titlePhrase: fields.title.join(" "),
+            titleTerms: new Set(fields.title),
+            typeTerms: new Set(fields.type),
+            tagTerms: new Set(fields.tag),
+            descriptionTerms: new Set(fields.description),
+            bodyTerms,
+            fieldInstances,
+            title,
+            exactType: type,
+            exactTags: Object.freeze([...tags]),
+            description: documentDescription(document) ?? "",
+            body,
+        });
+    }
+    Object.freeze(rows);
+    const byId = new Map(rows.map((row) => [row.indexedId, row]));
+    let digest;
+    const size = Object.freeze({
+        documents: documents.length,
+        indexed: rows.length,
+        skipped: documents.length - rows.length,
+        reindexPending: 0,
+    });
+    return Object.freeze({
+        key,
+        // Lazy: hashed per document only when somebody asks for the identity.
+        get indexDigest() {
+            digest ??= computeIndexDigest(rows);
+            return digest;
+        },
+        async evidence(terms, filters, bounds, evidenceStop) {
+            stopIfAsked(evidenceStop);
+            let volume = 0;
+            for (const [position, row] of rows.entries()) {
+                if (position % SIGNAL_INTERVAL === 0)
+                    stopIfAsked(evidenceStop);
+                for (const term of terms)
+                    volume += instancesOf(row, term);
+            }
+            const held = [];
+            // Over the bound nothing is walked, as a native index refuses before it reads a posting.
+            if (volume > bounds.maxInstances)
+                return { key, size, volume, documents: held };
+            for (const [position, row] of rows.entries()) {
+                if (position % SIGNAL_INTERVAL === 0)
+                    stopIfAsked(evidenceStop);
+                if (filters.type !== undefined && row.exactType !== filters.type)
+                    continue;
+                if (filters.tag !== undefined && !row.exactTags.includes(filters.tag))
+                    continue;
+                if (filters.prefix !== undefined && !row.indexedId.startsWith(filters.prefix))
+                    continue;
+                const evidence = evidenceOf(row, terms);
+                if (evidence !== undefined)
+                    held.push(evidence);
+            }
+            return { key, size, volume, documents: held };
+        },
+        async text(hits) {
+            const texts = new Map();
+            for (const hit of hits) {
+                const row = byId.get(hit.id);
+                if (row !== undefined && row.indexedVersion === hit.version)
+                    texts.set(hit.id, { description: row.description, body: row.body });
+            }
+            return texts;
+        },
+    });
+}
+/** The term's capped instances in the row: searchRowText's `instances` for the same document. */
+function instancesOf(row, term) {
+    return (row.fieldInstances.get(term) ?? 0) + Math.min(row.bodyTerms.get(term)?.length ?? 0, BODY_OCCURRENCE_CAP);
+}
+function fieldsForTerm(row, term) {
+    const fields = [];
+    if (row.titleTerms.has(term))
+        fields.push("title");
+    if (row.typeTerms.has(term))
+        fields.push("type");
+    if (row.tagTerms.has(term))
+        fields.push("tag");
+    if (row.descriptionTerms.has(term))
+        fields.push("description");
+    if (row.idSegments.has(term))
+        fields.push("id");
+    if (row.bodyTerms.has(term))
+        fields.push("body");
+    return fields;
+}
+/** The row's evidence for the terms it holds, or nothing when it holds none of them. */
+function evidenceOf(row, terms) {
+    let held;
+    for (const term of terms) {
+        const fields = fieldsForTerm(row, term);
+        if (fields.length === 0)
+            continue;
+        held ??= new Map();
+        held.set(term, {
+            fields: Object.freeze(fields),
+            bodyOccurrences: Math.min(row.bodyTerms.get(term)?.length ?? 0, BODY_OCCURRENCE_CAP),
+        });
+    }
+    if (held === undefined)
+        return undefined;
+    return {
+        id: row.indexedId,
+        version: row.indexedVersion,
+        ...(row.title === undefined ? {} : { title: row.title }),
+        ...(row.exactType === undefined ? {} : { type: row.exactType }),
+        tags: row.exactTags,
+        titlePhrase: row.titlePhrase,
+        idPhrase: row.idPhrase,
+        terms: held,
+    };
+}
+/** Portal's index identity, byte for byte: one hash per document, then one over the list. */
+function computeIndexDigest(rows) {
+    const fingerprints = rows.map((row) => sha256HexOfUtf8(canonicalJson({
+        id: row.indexedId,
+        version: row.indexedVersion,
+        idPhrase: row.idPhrase,
+        titlePhrase: row.titlePhrase,
+        type: row.exactType ?? null,
+        tags: [...row.exactTags],
+        description: [...row.descriptionTerms].sort(codeUnitCompare),
+        body: [...row.bodyTerms.keys()]
+            .sort(codeUnitCompare)
+            .map((term) => [term, row.bodyTerms.get(term).map((token) => [token.start, token.end])]),
+    })));
+    return `sha256:${sha256HexOfUtf8(canonicalJson({ algorithm: SEARCH_ALGORITHM, documents: fingerprints }))}`;
+}
+//# sourceMappingURL=memory-source.js.map

package/dist/search/memory-source.js.map

@@ -0,0 +1 @@
+{"version":3,"file":"memory-source.js","sourceRoot":"","sources":["../../src/search/memory-source.ts"],"names":[],"mappings":"AAAA;;;;;;;;GAQG;AACH,OAAO,EAAE,eAAe,EAAE,MAAM,cAAc,CAAC;AAC/C,OAAO,EACL,mBAAmB,EACnB,gBAAgB,EAChB,aAAa,EACb,eAAe,EACf,mBAAmB,EACnB,YAAY,EACZ,aAAa,EACb,YAAY,GAKb,MAAM,eAAe,CAAC;AACvB,OAAO,EAAE,gBAAgB,EAAE,WAAW,EAAE,MAAM,aAAa,CAAC;AAC5D,OAAO,EAAE,aAAa,EAAE,eAAe,EAAE,MAAM,eAAe,CAAC;AAC/D,OAAO,EAAE,mBAAmB,EAAE,WAAW,EAAE,MAAM,eAAe,CAAC;AAQjE,OAAO,EAAE,QAAQ,EAAE,MAAM,WAAW,CAAC;AAiDrC,SAAS,aAAa,CAAC,KAAa,EAAE,GAAW,EAAE,MAAc;IAC/D,MAAM,IAAI,gBAAgB,CAAC,KAAK,EAAE,GAAG,EAAE,MAAM,CAAC,CAAC;AACjD,CAAC;AAED,mGAAmG;AACnG,SAAS,KAAK,CAAC,IAAY,EAAE,KAAyB,EAAE,WAAmB,EAAE,OAAe;IAC1F,IAAI,KAAK,KAAK,SAAS;QAAE,OAAO,EAAE,IAAI,EAAE,WAAW,EAAE,GAAG,EAAE,OAAO,EAAE,CAAC;IACpE,IAAI,CAAC,MAAM,CAAC,aAAa,CAAC,KAAK,CAAC,IAAI,KAAK,GAAG,CAAC;QAAE,MAAM,IAAI,SAAS,CAAC,GAAG,IAAI,6BAA6B,CAAC,CAAC;IACzG,OAAO,KAAK,GAAG,OAAO,CAAC,CAAC,CAAC,EAAE,IAAI,EAAE,GAAG,EAAE,KAAK,EAAE,CAAC,CAAC,CAAC,EAAE,IAAI,EAAE,WAAW,EAAE,GAAG,EAAE,OAAO,EAAE,CAAC;AACtF,CAAC;AAED,+DAA+D;AAC/D,MAAM,UAAU,kBAAkB,CAAC,OAAiD;IAClF,MAAM,YAAY,GAAG,KAAK,CAAC,cAAc,EAAE,OAAO,CAAC,YAAY,EAAE,mBAAmB,EAAE,aAAa,CAAC,iBAAiB,CAAC,CAAC;IACvH,MAAM,cAAc,GAAG,KAAK,CAC1B,gBAAgB,EAChB,OAAO,CAAC,cAAc,EACtB,qBAAqB,EACrB,aAAa,CAAC,mBAAmB,CAClC,CAAC;IACF,OAAO,EAAE,YAAY,EAAE,cAAc,EAAE,CAAC;AAC1C,CAAC;AAED,iGAAiG;AACjG,SAAS,SAAS,CAAC,IAAY,EAAE,mBAA2B;IAC1D,MAAM,KAAK,GAAG,IAAI,GAAG,EAAuB,CAAC;IAC7C,MAAM,KAAK,GAAG,EAAE,KAAK,EAAE,CAAC,EAAE,IAAI,EAAE,KAAK,EAAE,CAAC;IACxC,QAAQ,CAAC,IAAI,EAAE,CAAC,IAAI,EAAE,KAAK,EAAE,GAAG,EAAE,EAAE;QAClC,KAAK,CAAC,KAAK,IAAI,CAAC,CAAC;QACjB,mDAAmD;QACnD,IAAI,KAAK,CAAC,KAAK,GAAG,mBAAmB,EAAE,CAAC;YACtC,KAAK,CAAC,IAAI,GAAG,IAAI,CAAC;YAClB,OAAO,IAAI,CAAC;QACd,CAAC;QACD,MAAM,QAAQ,GAAG,KAAK,CAAC,GAAG,CAAC,IAAI,CAAC,CAAC;QACjC,IAAI,QAAQ;YAAE,QAAQ,CAAC,IAAI,CAAC,EAAE,KAAK,EAAE,GAAG,EAAE,CAAC,CAAC;;YACvC,KAAK,CAAC,GAAG,CAAC,IAAI,EAAE,CAAC,EAAE,KAAK,EAAE,GAAG,EAAE,CAAC,CAAC,CAAC;QACvC,OAAO,KAAK,CAAC;IACf,CAAC,CAAC,CAAC;IACH,OAAO,KAAK,CAAC,IAAI,CAAC,CAAC,CAAC,SAAS,CAAC,CAAC,CAAC,KAAK,CAAC;AACxC,CAAC;AAED;;;;;GAKG;AACH,MAAM,UAAU,kBAAkB,CAChC,SAAsC,EACtC,OAAoC;IAEpC,MAAM,GAAG,GAAG,OAAO,EAAE,GAAG,CAAC;IACzB,IAAI,OAAO,GAAG,KAAK,QAAQ,IAAI,GAAG,KAAK,EAAE;QAAE,MAAM,IAAI,SAAS,CAAC,0CAA0C,CAAC,CAAC;IAC3G,MAAM,EAAE,YAAY,EAAE,cAAc,EAAE,GAAG,kBAAkB,CAAC,OAAO,CAAC,CAAC;IACrE,MAAM,IAAI,GAAG,OAAO,CAAC,IAAI,CAAC;IAC1B,WAAW,CAAC,IAAI,CAAC,CAAC;IAClB,IAAI,SAAS,CAAC,MAAM,GAAG,YAAY,CAAC,GAAG;QAAE,aAAa,CAAC,YAAY,CAAC,IAAI,EAAE,YAAY,CAAC,GAAG,EAAE,SAAS,CAAC,MAAM,CAAC,CAAC;IAC9G,MAAM,IAAI,GAAsB,EAAE,CAAC;IACnC,MAAM,GAAG,GAAG,IAAI,GAAG,EAAU,CAAC;IAC9B,IAAI,WAAW,GAAG,CAAC,CAAC;IACpB,KAAK,MAAM,CAAC,QAAQ,EAAE,QAAQ,CAAC,IAAI,SAAS,CAAC,OAAO,EAAE,EAAE,CAAC;QACvD,IAAI,QAAQ,GAAG,eAAe,KAAK,CAAC;YAAE,WAAW,CAAC,IAAI,CAAC,CAAC;QACxD,+EAA+E;QAC/E,IAAI,GAAG,CAAC,GAAG,CAAC,QAAQ,CAAC,EAAE,CAAC;YAAE,MAAM,IAAI,SAAS,CAAC,8CAA8C,CAAC,CAAC;QAC9F,GAAG,CAAC,GAAG,CAAC,QAAQ,CAAC,EAAE,CAAC,CAAC;QACrB,8EAA8E;QAC9E,MAAM,aAAa,GAAG,mBAAmB,CAAC,QAAQ,CAAC,CAAC;QACpD,IAAI,WAAW,GAAG,aAAa,GAAG,cAAc,CAAC,GAAG,EAAE,CAAC;YACrD,aAAa,CAAC,cAAc,CAAC,IAAI,EAAE,cAAc,CAAC,GAAG,EAAE,WAAW,GAAG,aAAa,CAAC,CAAC;QACtF,CAAC;QACD,WAAW,IAAI,aAAa,CAAC;QAC7B,MAAM,IAAI,GAAG,QAAQ,CAAC,IAAI,CAAC;QAC3B,MAAM,OAAO,GAAG,SAAS,CAAC,IAAI,EAAE,aAAa,CAAC,mBAAmB,CAAC,CAAC;QACnE,IAAI,OAAO,KAAK,SAAS;YAAE,SAAS;QACpC,MAAM,SAAS,GAAG,IAAI,GAAG,EAAgC,CAAC;QAC1D,KAAK,MAAM,CAAC,IAAI,EAAE,MAAM,CAAC,IAAI,OAAO;YAAE,SAAS,CAAC,GAAG,CAAC,IAAI,EAAE,MAAM,CAAC,MAAM,CAAC,MAAM,CAAC,CAAC,CAAC;QACjF,MAAM,KAAK,GAAG,aAAa,CAAC,QAAQ,CAAC,CAAC;QACtC,MAAM,IAAI,GAAG,YAAY,CAAC,QAAQ,CAAC,CAAC;QACpC,MAAM,IAAI,GAAG,YAAY,CAAC,QAAQ,CAAC,CAAC;QACpC,MAAM,MAAM,GAAG,WAAW,CAAC,QAAQ,CAAC,CAAC;QACrC,MAAM,cAAc,GAAG,IAAI,GAAG,EAAkB,CAAC;QACjD,KAAK,MAAM,MAAM,IAAI,CAAC,MAAM,CAAC,EAAE,EAAE,MAAM,CAAC,KAAK,EAAE,MAAM,CAAC,IAAI,EAAE,MAAM,CAAC,GAAG,EAAE,MAAM,CAAC,WAAW,CAAC;YACzF,KAAK,MAAM,IAAI,IAAI,MAAM;gBAAE,cAAc,CAAC,GAAG,CAAC,IAAI,EAAE,CAAC,cAAc,CAAC,GAAG,CAAC,IAAI,CAAC,IAAI,CAAC,CAAC,GAAG,CAAC,CAAC,CAAC;QAC3F,IAAI,CAAC,IAAI,CAAC;YACR,SAAS,EAAE,QAAQ,CAAC,EAAE;YACtB,cAAc,EAAE,QAAQ,CAAC,OAAO;YAChC,QAAQ,EAAE,MAAM,CAAC,EAAE,CAAC,IAAI,CAAC,GAAG,CAAC;YAC7B,UAAU,EAAE,IAAI,GAAG,CAAC,MAAM,CAAC,EAAE,CAAC;YAC9B,WAAW,EAAE,MAAM,CAAC,KAAK,CAAC,IAAI,CAAC,GAAG,CAAC;YACnC,UAAU,EAAE,IAAI,GAAG,CAAC,MAAM,CAAC,KAAK,CAAC;YACjC,SAAS,EAAE,IAAI,GAAG,CAAC,MAAM,CAAC,IAAI,CAAC;YAC/B,QAAQ,EAAE,IAAI,GAAG,CAAC,MAAM,CAAC,GAAG,CAAC;YAC7B,gBAAgB,EAAE,IAAI,GAAG,CAAC,MAAM,CAAC,WAAW,CAAC;YAC7C,SAAS;YACT,cAAc;YACd,KAAK;YACL,SAAS,EAAE,IAAI;YACf,SAAS,EAAE,MAAM,CAAC,MAAM,CAAC,CAAC,GAAG,IAAI,CAAC,CAAC;YACnC,WAAW,EAAE,mBAAmB,CAAC,QAAQ,CAAC,IAAI,EAAE;YAChD,IAAI;SACL,CAAC,CAAC;IACL,CAAC;IACD,MAAM,CAAC,MAAM,CAAC,IAAI,CAAC,CAAC;IACpB,MAAM,IAAI,GAAG,IAAI,GAAG,CAAC,IAAI,CAAC,GAAG,CAAC,CAAC,GAAG,EAAE,EAAE,CAAC,CAAC,GAAG,CAAC,SAAS,EAAE,GAAG,CAAC,CAAC,CAAC,CAAC;IAC9D,IAAI,MAAsC,CAAC;IAC3C,MAAM,IAAI,GAAG,MAAM,CAAC,MAAM,CAAC;QACzB,SAAS,EAAE,SAAS,CAAC,MAAM;QAC3B,OAAO,EAAE,IAAI,CAAC,MAAM;QACpB,OAAO,EAAE,SAAS,CAAC,MAAM,GAAG,IAAI,CAAC,MAAM;QACvC,cAAc,EAAE,CAAC;KAClB,CAAC,CAAC;IAEH,OAAO,MAAM,CAAC,MAAM,CAAC;QACnB,GAAG;QACH,sEAAsE;QACtE,IAAI,WAAW;YACb,MAAM,KAAK,kBAAkB,CAAC,IAAI,CAAC,CAAC;YACpC,OAAO,MAAM,CAAC;QAChB,CAAC;QACD,KAAK,CAAC,QAAQ,CACZ,KAAwB,EACxB,OAAwB,EACxB,MAA8B,EAC9B,YAA2B;YAE3B,WAAW,CAAC,YAAY,CAAC,CAAC;YAC1B,IAAI,MAAM,GAAG,CAAC,CAAC;YACf,KAAK,MAAM,CAAC,QAAQ,EAAE,GAAG,CAAC,IAAI,IAAI,CAAC,OAAO,EAAE,EAAE,CAAC;gBAC7C,IAAI,QAAQ,GAAG,eAAe,KAAK,CAAC;oBAAE,WAAW,CAAC,YAAY,CAAC,CAAC;gBAChE,KAAK,MAAM,IAAI,IAAI,KAAK;oBAAE,MAAM,IAAI,WAAW,CAAC,GAAG,EAAE,IAAI,CAAC,CAAC;YAC7D,CAAC;YACD,MAAM,IAAI,GAAuB,EAAE,CAAC;YACpC,yFAAyF;YACzF,IAAI,MAAM,GAAG,MAAM,CAAC,YAAY;gBAAE,OAAO,EAAE,GAAG,EAAE,IAAI,EAAE,MAAM,EAAE,SAAS,EAAE,IAAI,EAAE,CAAC;YAChF,KAAK,MAAM,CAAC,QAAQ,EAAE,GAAG,CAAC,IAAI,IAAI,CAAC,OAAO,EAAE,EAAE,CAAC;gBAC7C,IAAI,QAAQ,GAAG,eAAe,KAAK,CAAC;oBAAE,WAAW,CAAC,YAAY,CAAC,CAAC;gBAChE,IAAI,OAAO,CAAC,IAAI,KAAK,SAAS,IAAI,GAAG,CAAC,SAAS,KAAK,OAAO,CAAC,IAAI;oBAAE,SAAS;gBAC3E,IAAI,OAAO,CAAC,GAAG,KAAK,SAAS,IAAI,CAAC,GAAG,CAAC,SAAS,CAAC,QAAQ,CAAC,OAAO,CAAC,GAAG,CAAC;oBAAE,SAAS;gBAChF,IAAI,OAAO,CAAC,MAAM,KAAK,SAAS,IAAI,CAAC,GAAG,CAAC,SAAS,CAAC,UAAU,CAAC,OAAO,CAAC,MAAM,CAAC;oBAAE,SAAS;gBACxF,MAAM,QAAQ,GAAG,UAAU,CAAC,GAAG,EAAE,KAAK,CAAC,CAAC;gBACxC,IAAI,QAAQ,KAAK,SAAS;oBAAE,IAAI,CAAC,IAAI,CAAC,QAAQ,CAAC,CAAC;YAClD,CAAC;YACD,OAAO,EAAE,GAAG,EAAE,IAAI,EAAE,MAAM,EAAE,SAAS,EAAE,IAAI,EAAE,CAAC;QAChD,CAAC;QACD,KAAK,CAAC,IAAI,CAAC,IAAkE;YAC3E,MAAM,KAAK,GAAG,IAAI,GAAG,EAAwB,CAAC;YAC9C,KAAK,MAAM,GAAG,IAAI,IAAI,EAAE,CAAC;gBACvB,MAAM,GAAG,GAAG,IAAI,CAAC,GAAG,CAAC,GAAG,CAAC,EAAE,CAAC,CAAC;gBAC7B,IAAI,GAAG,KAAK,SAAS,IAAI,GAAG,CAAC,cAAc,KAAK,GAAG,CAAC,OAAO;oBACzD,KAAK,CAAC,GAAG,CAAC,GAAG,CAAC,EAAE,EAAE,EAAE,WAAW,EAAE,GAAG,CAAC,WAAW,EAAE,IAAI,EAAE,GAAG,CAAC,IAAI,EAAE,CAAC,CAAC;YACxE,CAAC;YACD,OAAO,KAAK,CAAC;QACf,CAAC;KACF,CAAC,CAAC;AACL,CAAC;AAED,iGAAiG;AACjG,SAAS,WAAW,CAAC,GAAoB,EAAE,IAAY;IACrD,OAAO,CAAC,GAAG,CAAC,cAAc,CAAC,GAAG,CAAC,IAAI,CAAC,IAAI,CAAC,CAAC,GAAG,IAAI,CAAC,GAAG,CAAC,GAAG,CAAC,SAAS,CAAC,GAAG,CAAC,IAAI,CAAC,EAAE,MAAM,IAAI,CAAC,EAAE,mBAAmB,CAAC,CAAC;AACnH,CAAC;AAED,SAAS,aAAa,CAAC,GAAoB,EAAE,IAAY;IACvD,MAAM,MAAM,GAAyB,EAAE,CAAC;IACxC,IAAI,GAAG,CAAC,UAAU,CAAC,GAAG,CAAC,IAAI,CAAC;QAAE,MAAM,CAAC,IAAI,CAAC,OAAO,CAAC,CAAC;IACnD,IAAI,GAAG,CAAC,SAAS,CAAC,GAAG,CAAC,IAAI,CAAC;QAAE,MAAM,CAAC,IAAI,CAAC,MAAM,CAAC,CAAC;IACjD,IAAI,GAAG,CAAC,QAAQ,CAAC,GAAG,CAAC,IAAI,CAAC;QAAE,MAAM,CAAC,IAAI,CAAC,KAAK,CAAC,CAAC;IAC/C,IAAI,GAAG,CAAC,gBAAgB,CAAC,GAAG,CAAC,IAAI,CAAC;QAAE,MAAM,CAAC,IAAI,CAAC,aAAa,CAAC,CAAC;IAC/D,IAAI,GAAG,CAAC,UAAU,CAAC,GAAG,CAAC,IAAI,CAAC;QAAE,MAAM,CAAC,IAAI,CAAC,IAAI,CAAC,CAAC;IAChD,IAAI,GAAG,CAAC,SAAS,CAAC,GAAG,CAAC,IAAI,CAAC;QAAE,MAAM,CAAC,IAAI,CAAC,MAAM,CAAC,CAAC;IACjD,OAAO,MAAM,CAAC;AAChB,CAAC;AAED,wFAAwF;AACxF,SAAS,UAAU,CAAC,GAAoB,EAAE,KAAwB;IAChE,IAAI,IAAiG,CAAC;IACtG,KAAK,MAAM,IAAI,IAAI,KAAK,EAAE,CAAC;QACzB,MAAM,MAAM,GAAG,aAAa,CAAC,GAAG,EAAE,IAAI,CAAC,CAAC;QACxC,IAAI,MAAM,CAAC,MAAM,KAAK,CAAC;YAAE,SAAS;QAClC,IAAI,KAAK,IAAI,GAAG,EAAE,CAAC;QACnB,IAAI,CAAC,GAAG,CAAC,IAAI,EAAE;YACb,MAAM,EAAE,MAAM,CAAC,MAAM,CAAC,MAAM,CAAC;YAC7B,eAAe,EAAE,IAAI,CAAC,GAAG,CAAC,GAAG,CAAC,SAAS,CAAC,GAAG,CAAC,IAAI,CAAC,EAAE,MAAM,IAAI,CAAC,EAAE,mBAAmB,CAAC;SACrF,CAAC,CAAC;IACL,CAAC;IACD,IAAI,IAAI,KAAK,SAAS;QAAE,OAAO,SAAS,CAAC;IACzC,OAAO;QACL,EAAE,EAAE,GAAG,CAAC,SAAS;QACjB,OAAO,EAAE,GAAG,CAAC,cAAc;QAC3B,GAAG,CAAC,GAAG,CAAC,KAAK,KAAK,SAAS,CAAC,CAAC,CAAC,EAAE,CAAC,CAAC,CAAC,EAAE,KAAK,EAAE,GAAG,CAAC,KAAK,EAAE,CAAC;QACxD,GAAG,CAAC,GAAG,CAAC,SAAS,KAAK,SAAS,CAAC,CAAC,CAAC,EAAE,CAAC,CAAC,CAAC,EAAE,IAAI,EAAE,GAAG,CAAC,SAAS,EAAE,CAAC;QAC/D,IAAI,EAAE,GAAG,CAAC,SAAS;QACnB,WAAW,EAAE,GAAG,CAAC,WAAW;QAC5B,QAAQ,EAAE,GAAG,CAAC,QAAQ;QACtB,KAAK,EAAE,IAAI;KACZ,CAAC;AACJ,CAAC;AAED,6FAA6F;AAC7F,SAAS,kBAAkB,CAAC,IAAgC;IAC1D,MAAM,YAAY,GAAG,IAAI,CAAC,GAAG,CAAC,CAAC,GAAG,EAAE,EAAE,CACpC,eAAe,CACb,aAAa,CAAC;QACZ,EAAE,EAAE,GAAG,CAAC,SAAS;QACjB,OAAO,EAAE,GAAG,CAAC,cAAc;QAC3B,QAAQ,EAAE,GAAG,CAAC,QAAQ;QACtB,WAAW,EAAE,GAAG,CAAC,WAAW;QAC5B,IAAI,EAAE,GAAG,CAAC,SAAS,IAAI,IAAI;QAC3B,IAAI,EAAE,CAAC,GAAG,GAAG,CAAC,SAAS,CAAC;QACxB,WAAW,EAAE,CAAC,GAAG,GAAG,CAAC,gBAAgB,CAAC,CAAC,IAAI,CAAC,eAAe,CAAC;QAC5D,IAAI,EAAE,CAAC,GAAG,GAAG,CAAC,SAAS,CAAC,IAAI,EAAE,CAAC;aAC5B,IAAI,CAAC,eAAe,CAAC;aACrB,GAAG,CAAC,CAAC,IAAI,EAAE,EAAE,CAAC,CAAC,IAAI,EAAE,GAAG,CAAC,SAAS,CAAC,GAAG,CAAC,IAAI,CAAE,CAAC,GAAG,CAAC,CAAC,KAAK,EAAE,EAAE,CAAC,CAAC,KAAK,CAAC,KAAK,EAAE,KAAK,CAAC,GAAG,CAAC,CAAC,CAAC,CAAC;KAC5F,CAAC,CACH,CACF,CAAC;IACF,OAAO,UAAU,eAAe,CAAC,aAAa,CAAC,EAAE,SAAS,EAAE,gBAAgB,EAAE,SAAS,EAAE,YAAY,EAAE,CAAC,CAAC,EAAE,CAAC;AAC9G,CAAC"}

package/dist/search/rank.d.ts

@@ -0,0 +1,34 @@
+/**
+ * Scoring, ordering, facets and excerpts over a source's evidence. The weights, the order and the
+ * excerpt window are portal-webmcp 0.3.1's (`search.ts`), read from evidence instead of an index
+ * row, so every source ranks with this one code.
+ */
+import { type SearchFacetV1, type SearchFacetValueV1, type SearchFiltersV1, type SearchMatchFieldV1, type SearchMatchModeV1, type SearchStopV1 } from "./contract.js";
+import type { SearchEvidenceV1, SearchTextV1 } from "./source.js";
+import { type NormalizedQueryV1 } from "./text.js";
+export interface SearchTermEvidenceV1 {
+    readonly term: string;
+    readonly fields: readonly SearchMatchFieldV1[];
+}
+export interface RankedHitV1 {
+    readonly evidence: SearchEvidenceV1;
+    readonly score: number;
+    /** Exact normalized terms and indexed fields that contributed to this hit. */
+    readonly matchedTerms: readonly SearchTermEvidenceV1[];
+}
+/** Deterministic ranking: score descending, then canonical id in code-unit order. With `any`,
+ * term coverage first. The exact filters are applied again here, so a source cannot widen them. */
+export declare function rankEvidence(documents: readonly SearchEvidenceV1[], query: NormalizedQueryV1, filters: SearchFiltersV1, match: SearchMatchModeV1, stop?: SearchStopV1): readonly RankedHitV1[];
+/** Value counts over every hit; a repeated tag counts once per document. */
+export declare function facetsOf(hits: readonly RankedHitV1[], wanted: ReadonlySet<SearchFacetV1>): {
+    tags?: SearchFacetValueV1[] | undefined;
+    type?: SearchFacetValueV1[] | undefined;
+};
+/**
+ * The body around the first body match of the query's terms, else the start of the description.
+ * The first token the scanner meets is the lowest byte start; tokens sharing a start come from one
+ * run and share its range, so query-term order cannot change the window. The body is scanned only
+ * when the evidence says it holds a term, and only up to the first match.
+ */
+export declare function excerptOf(hit: RankedHitV1, text: SearchTextV1 | undefined, query: NormalizedQueryV1, excerptBytes: number): string;
+//# sourceMappingURL=rank.d.ts.map

package/dist/search/rank.d.ts.map

@@ -0,0 +1 @@
+{"version":3,"file":"rank.d.ts","sourceRoot":"","sources":["../../src/search/rank.ts"],"names":[],"mappings":"AAAA;;;;GAIG;AACH,OAAO,EAML,KAAK,aAAa,EAClB,KAAK,kBAAkB,EACvB,KAAK,eAAe,EACpB,KAAK,kBAAkB,EACvB,KAAK,iBAAiB,EACtB,KAAK,YAAY,EAClB,MAAM,eAAe,CAAC;AAGvB,OAAO,KAAK,EAAE,gBAAgB,EAAE,YAAY,EAAE,MAAM,aAAa,CAAC;AAClE,OAAO,EAOL,KAAK,iBAAiB,EACvB,MAAM,WAAW,CAAC;AAEnB,MAAM,WAAW,oBAAoB;IACnC,QAAQ,CAAC,IAAI,EAAE,MAAM,CAAC;IACtB,QAAQ,CAAC,MAAM,EAAE,SAAS,kBAAkB,EAAE,CAAC;CAChD;AAED,MAAM,WAAW,WAAW;IAC1B,QAAQ,CAAC,QAAQ,EAAE,gBAAgB,CAAC;IACpC,QAAQ,CAAC,KAAK,EAAE,MAAM,CAAC;IACvB,8EAA8E;IAC9E,QAAQ,CAAC,YAAY,EAAE,SAAS,oBAAoB,EAAE,CAAC;CACxD;AAwCD;mGACmG;AACnG,wBAAgB,YAAY,CAC1B,SAAS,EAAE,SAAS,gBAAgB,EAAE,EACtC,KAAK,EAAE,iBAAiB,EACxB,OAAO,EAAE,eAAe,EACxB,KAAK,EAAE,iBAAiB,EACxB,IAAI,CAAC,EAAE,YAAY,GAClB,SAAS,WAAW,EAAE,CAmBxB;AASD,4EAA4E;AAC5E,wBAAgB,QAAQ,CAAC,IAAI,EAAE,SAAS,WAAW,EAAE,EAAE,MAAM,EAAE,WAAW,CAAC,aAAa,CAAC;;;EAWxF;AAgBD;;;;;GAKG;AACH,wBAAgB,SAAS,CACvB,GAAG,EAAE,WAAW,EAChB,IAAI,EAAE,YAAY,GAAG,SAAS,EAC9B,KAAK,EAAE,iBAAiB,EACxB,YAAY,EAAE,MAAM,GACnB,MAAM,CAcR"}

package/dist/search/rank.js

@@ -0,0 +1,129 @@
+/**
+ * Scoring, ordering, facets and excerpts over a source's evidence. The weights, the order and the
+ * excerpt window are portal-webmcp 0.3.1's (`search.ts`), read from evidence instead of an index
+ * row, so every source ranks with this one code.
+ */
+import { BODY_OCCURRENCE_CAP, EXCERPT_LEAD_BYTES, SEARCH_LIMITS, SEARCH_MATCH_FIELDS, SIGNAL_INTERVAL, } from "./contract.js";
+import { stopIfAsked } from "./errors.js";
+import { codeUnitCompare } from "./identity.js";
+import { backwardBoundary, collapseWhitespace, decodeUtf8, forwardBoundary, scanBody, utf8, } from "./text.js";
+/** The fields holding `term`, in the contract's field order whatever order the source used. */
+function fieldsForTerm(evidence, term) {
+    const held = evidence.terms.get(term);
+    if (held === undefined)
+        return [];
+    return SEARCH_MATCH_FIELDS.filter((field) => held.fields.includes(field));
+}
+function score(evidence, query, match) {
+    const matchedTerms = [];
+    for (const term of query.terms) {
+        const fields = fieldsForTerm(evidence, term);
+        if (fields.length === 0) {
+            if (match === "all")
+                return undefined;
+            continue;
+        }
+        matchedTerms.push(Object.freeze({ term, fields: Object.freeze(fields) }));
+    }
+    if (matchedTerms.length === 0)
+        return undefined;
+    let total = 0;
+    if (evidence.titlePhrase === query.phrase)
+        total += 10_000;
+    else if (evidence.titlePhrase.startsWith(`${query.phrase} `))
+        total += 5_000;
+    if (evidence.idPhrase === query.phrase)
+        total += 4_000;
+    for (const { term, fields } of matchedTerms) {
+        if (fields.includes("title"))
+            total += 500;
+        if (fields.includes("type"))
+            total += 250;
+        if (fields.includes("tag"))
+            total += 250;
+        if (fields.includes("description"))
+            total += 100;
+        if (fields.includes("id"))
+            total += 100;
+        const occurrences = evidence.terms.get(term)?.bodyOccurrences ?? 0;
+        total += Math.min(occurrences, BODY_OCCURRENCE_CAP) * 10;
+    }
+    return { score: total, matchedTerms: Object.freeze(matchedTerms) };
+}
+/** Deterministic ranking: score descending, then canonical id in code-unit order. With `any`,
+ * term coverage first. The exact filters are applied again here, so a source cannot widen them. */
+export function rankEvidence(documents, query, filters, match, stop) {
+    stopIfAsked(stop);
+    const hits = [];
+    for (const [position, evidence] of documents.entries()) {
+        if (position % SIGNAL_INTERVAL === 0)
+            stopIfAsked(stop);
+        if (filters.type !== undefined && evidence.type !== filters.type)
+            continue;
+        if (filters.tag !== undefined && !evidence.tags.includes(filters.tag))
+            continue;
+        if (filters.prefix !== undefined && !evidence.id.startsWith(filters.prefix))
+            continue;
+        const value = score(evidence, query, match);
+        if (value === undefined)
+            continue;
+        hits.push(Object.freeze({ evidence, score: value.score, matchedTerms: value.matchedTerms }));
+    }
+    hits.sort((left, right) => (match === "any" ? right.matchedTerms.length - left.matchedTerms.length : 0) ||
+        right.score - left.score ||
+        codeUnitCompare(left.evidence.id, right.evidence.id));
+    return Object.freeze(hits);
+}
+function facetValues(counts) {
+    return [...counts.entries()]
+        .sort(([leftValue, left], [rightValue, right]) => right - left || codeUnitCompare(leftValue, rightValue))
+        .slice(0, SEARCH_LIMITS.maxFacetValues)
+        .map(([value, count]) => ({ value, count }));
+}
+/** Value counts over every hit; a repeated tag counts once per document. */
+export function facetsOf(hits, wanted) {
+    const types = new Map();
+    const tags = new Map();
+    for (const { evidence } of hits) {
+        if (wanted.has("type") && evidence.type !== undefined)
+            types.set(evidence.type, (types.get(evidence.type) ?? 0) + 1);
+        if (wanted.has("tags"))
+            for (const tag of new Set(evidence.tags))
+                tags.set(tag, (tags.get(tag) ?? 0) + 1);
+    }
+    return {
+        ...(wanted.has("type") ? { type: facetValues(types) } : {}),
+        ...(wanted.has("tags") ? { tags: facetValues(tags) } : {}),
+    };
+}
+function excerptFrom(bytes, match, excerptBytes) {
+    const matchBytes = match.end - match.start;
+    const leadBudget = Math.max(0, excerptBytes - Math.min(excerptBytes, matchBytes));
+    const lead = Math.min(EXCERPT_LEAD_BYTES, leadBudget);
+    const start = forwardBoundary(bytes, Math.max(0, match.start - lead));
+    const end = backwardBoundary(bytes, Math.min(bytes.byteLength, start + excerptBytes));
+    return collapseWhitespace(decodeUtf8(bytes, start, Math.max(start, end)));
+}
+/**
+ * The body around the first body match of the query's terms, else the start of the description.
+ * The first token the scanner meets is the lowest byte start; tokens sharing a start come from one
+ * run and share its range, so query-term order cannot change the window. The body is scanned only
+ * when the evidence says it holds a term, and only up to the first match.
+ */
+export function excerptOf(hit, text, query, excerptBytes) {
+    const body = text?.body ?? "";
+    const found = {};
+    if (hit.matchedTerms.some(({ fields }) => fields.includes("body"))) {
+        const terms = new Set(query.terms);
+        scanBody(body, (term, start, end) => {
+            if (!terms.has(term))
+                return false;
+            found.range = { start, end };
+            return true;
+        });
+    }
+    return found.range !== undefined
+        ? excerptFrom(utf8(body), found.range, excerptBytes)
+        : excerptFrom(utf8(text?.description ?? ""), { start: 0, end: 0 }, excerptBytes);
+}
+//# sourceMappingURL=rank.js.map

package/dist/search/rank.js.map

@@ -0,0 +1 @@
+{"version":3,"file":"rank.js","sourceRoot":"","sources":["../../src/search/rank.ts"],"names":[],"mappings":"AAAA;;;;GAIG;AACH,OAAO,EACL,mBAAmB,EACnB,kBAAkB,EAClB,aAAa,EACb,mBAAmB,EACnB,eAAe,GAOhB,MAAM,eAAe,CAAC;AACvB,OAAO,EAAE,WAAW,EAAE,MAAM,aAAa,CAAC;AAC1C,OAAO,EAAE,eAAe,EAAE,MAAM,eAAe,CAAC;AAEhD,OAAO,EACL,gBAAgB,EAChB,kBAAkB,EAClB,UAAU,EACV,eAAe,EACf,QAAQ,EACR,IAAI,GAEL,MAAM,WAAW,CAAC;AAcnB,+FAA+F;AAC/F,SAAS,aAAa,CAAC,QAA0B,EAAE,IAAY;IAC7D,MAAM,IAAI,GAAG,QAAQ,CAAC,KAAK,CAAC,GAAG,CAAC,IAAI,CAAC,CAAC;IACtC,IAAI,IAAI,KAAK,SAAS;QAAE,OAAO,EAAE,CAAC;IAClC,OAAO,mBAAmB,CAAC,MAAM,CAAC,CAAC,KAAK,EAAE,EAAE,CAAC,IAAI,CAAC,MAAM,CAAC,QAAQ,CAAC,KAAK,CAAC,CAAC,CAAC;AAC5E,CAAC;AAED,SAAS,KAAK,CACZ,QAA0B,EAC1B,KAAwB,EACxB,KAAwB;IAExB,MAAM,YAAY,GAA2B,EAAE,CAAC;IAChD,KAAK,MAAM,IAAI,IAAI,KAAK,CAAC,KAAK,EAAE,CAAC;QAC/B,MAAM,MAAM,GAAG,aAAa,CAAC,QAAQ,EAAE,IAAI,CAAC,CAAC;QAC7C,IAAI,MAAM,CAAC,MAAM,KAAK,CAAC,EAAE,CAAC;YACxB,IAAI,KAAK,KAAK,KAAK;gBAAE,OAAO,SAAS,CAAC;YACtC,SAAS;QACX,CAAC;QACD,YAAY,CAAC,IAAI,CAAC,MAAM,CAAC,MAAM,CAAC,EAAE,IAAI,EAAE,MAAM,EAAE,MAAM,CAAC,MAAM,CAAC,MAAM,CAAC,EAAE,CAAC,CAAC,CAAC;IAC5E,CAAC;IACD,IAAI,YAAY,CAAC,MAAM,KAAK,CAAC;QAAE,OAAO,SAAS,CAAC;IAChD,IAAI,KAAK,GAAG,CAAC,CAAC;IACd,IAAI,QAAQ,CAAC,WAAW,KAAK,KAAK,CAAC,MAAM;QAAE,KAAK,IAAI,MAAM,CAAC;SACtD,IAAI,QAAQ,CAAC,WAAW,CAAC,UAAU,CAAC,GAAG,KAAK,CAAC,MAAM,GAAG,CAAC;QAAE,KAAK,IAAI,KAAK,CAAC;IAC7E,IAAI,QAAQ,CAAC,QAAQ,KAAK,KAAK,CAAC,MAAM;QAAE,KAAK,IAAI,KAAK,CAAC;IACvD,KAAK,MAAM,EAAE,IAAI,EAAE,MAAM,EAAE,IAAI,YAAY,EAAE,CAAC;QAC5C,IAAI,MAAM,CAAC,QAAQ,CAAC,OAAO,CAAC;YAAE,KAAK,IAAI,GAAG,CAAC;QAC3C,IAAI,MAAM,CAAC,QAAQ,CAAC,MAAM,CAAC;YAAE,KAAK,IAAI,GAAG,CAAC;QAC1C,IAAI,MAAM,CAAC,QAAQ,CAAC,KAAK,CAAC;YAAE,KAAK,IAAI,GAAG,CAAC;QACzC,IAAI,MAAM,CAAC,QAAQ,CAAC,aAAa,CAAC;YAAE,KAAK,IAAI,GAAG,CAAC;QACjD,IAAI,MAAM,CAAC,QAAQ,CAAC,IAAI,CAAC;YAAE,KAAK,IAAI,GAAG,CAAC;QACxC,MAAM,WAAW,GAAG,QAAQ,CAAC,KAAK,CAAC,GAAG,CAAC,IAAI,CAAC,EAAE,eAAe,IAAI,CAAC,CAAC;QACnE,KAAK,IAAI,IAAI,CAAC,GAAG,CAAC,WAAW,EAAE,mBAAmB,CAAC,GAAG,EAAE,CAAC;IAC3D,CAAC;IACD,OAAO,EAAE,KAAK,EAAE,KAAK,EAAE,YAAY,EAAE,MAAM,CAAC,MAAM,CAAC,YAAY,CAAC,EAAE,CAAC;AACrE,CAAC;AAED;mGACmG;AACnG,MAAM,UAAU,YAAY,CAC1B,SAAsC,EACtC,KAAwB,EACxB,OAAwB,EACxB,KAAwB,EACxB,IAAmB;IAEnB,WAAW,CAAC,IAAI,CAAC,CAAC;IAClB,MAAM,IAAI,GAAkB,EAAE,CAAC;IAC/B,KAAK,MAAM,CAAC,QAAQ,EAAE,QAAQ,CAAC,IAAI,SAAS,CAAC,OAAO,EAAE,EAAE,CAAC;QACvD,IAAI,QAAQ,GAAG,eAAe,KAAK,CAAC;YAAE,WAAW,CAAC,IAAI,CAAC,CAAC;QACxD,IAAI,OAAO,CAAC,IAAI,KAAK,SAAS,IAAI,QAAQ,CAAC,IAAI,KAAK,OAAO,CAAC,IAAI;YAAE,SAAS;QAC3E,IAAI,OAAO,CAAC,GAAG,KAAK,SAAS,IAAI,CAAC,QAAQ,CAAC,IAAI,CAAC,QAAQ,CAAC,OAAO,CAAC,GAAG,CAAC;YAAE,SAAS;QAChF,IAAI,OAAO,CAAC,MAAM,KAAK,SAAS,IAAI,CAAC,QAAQ,CAAC,EAAE,CAAC,UAAU,CAAC,OAAO,CAAC,MAAM,CAAC;YAAE,SAAS;QACtF,MAAM,KAAK,GAAG,KAAK,CAAC,QAAQ,EAAE,KAAK,EAAE,KAAK,CAAC,CAAC;QAC5C,IAAI,KAAK,KAAK,SAAS;YAAE,SAAS;QAClC,IAAI,CAAC,IAAI,CAAC,MAAM,CAAC,MAAM,CAAC,EAAE,QAAQ,EAAE,KAAK,EAAE,KAAK,CAAC,KAAK,EAAE,YAAY,EAAE,KAAK,CAAC,YAAY,EAAE,CAAC,CAAC,CAAC;IAC/F,CAAC;IACD,IAAI,CAAC,IAAI,CACP,CAAC,IAAI,EAAE,KAAK,EAAE,EAAE,CACd,CAAC,KAAK,KAAK,KAAK,CAAC,CAAC,CAAC,KAAK,CAAC,YAAY,CAAC,MAAM,GAAG,IAAI,CAAC,YAAY,CAAC,MAAM,CAAC,CAAC,CAAC,CAAC,CAAC;QAC5E,KAAK,CAAC,KAAK,GAAG,IAAI,CAAC,KAAK;QACxB,eAAe,CAAC,IAAI,CAAC,QAAQ,CAAC,EAAE,EAAE,KAAK,CAAC,QAAQ,CAAC,EAAE,CAAC,CACvD,CAAC;IACF,OAAO,MAAM,CAAC,MAAM,CAAC,IAAI,CAAC,CAAC;AAC7B,CAAC;AAED,SAAS,WAAW,CAAC,MAA2B;IAC9C,OAAO,CAAC,GAAG,MAAM,CAAC,OAAO,EAAE,CAAC;SACzB,IAAI,CAAC,CAAC,CAAC,SAAS,EAAE,IAAI,CAAC,EAAE,CAAC,UAAU,EAAE,KAAK,CAAC,EAAE,EAAE,CAAC,KAAK,GAAG,IAAI,IAAI,eAAe,CAAC,SAAS,EAAE,UAAU,CAAC,CAAC;SACxG,KAAK,CAAC,CAAC,EAAE,aAAa,CAAC,cAAc,CAAC;SACtC,GAAG,CAAC,CAAC,CAAC,KAAK,EAAE,KAAK,CAAC,EAAE,EAAE,CAAC,CAAC,EAAE,KAAK,EAAE,KAAK,EAAE,CAAC,CAAC,CAAC;AACjD,CAAC;AAED,4EAA4E;AAC5E,MAAM,UAAU,QAAQ,CAAC,IAA4B,EAAE,MAAkC;IACvF,MAAM,KAAK,GAAG,IAAI,GAAG,EAAkB,CAAC;IACxC,MAAM,IAAI,GAAG,IAAI,GAAG,EAAkB,CAAC;IACvC,KAAK,MAAM,EAAE,QAAQ,EAAE,IAAI,IAAI,EAAE,CAAC;QAChC,IAAI,MAAM,CAAC,GAAG,CAAC,MAAM,CAAC,IAAI,QAAQ,CAAC,IAAI,KAAK,SAAS;YAAE,KAAK,CAAC,GAAG,CAAC,QAAQ,CAAC,IAAI,EAAE,CAAC,KAAK,CAAC,GAAG,CAAC,QAAQ,CAAC,IAAI,CAAC,IAAI,CAAC,CAAC,GAAG,CAAC,CAAC,CAAC;QACrH,IAAI,MAAM,CAAC,GAAG,CAAC,MAAM,CAAC;YAAE,KAAK,MAAM,GAAG,IAAI,IAAI,GAAG,CAAC,QAAQ,CAAC,IAAI,CAAC;gBAAE,IAAI,CAAC,GAAG,CAAC,GAAG,EAAE,CAAC,IAAI,CAAC,GAAG,CAAC,GAAG,CAAC,IAAI,CAAC,CAAC,GAAG,CAAC,CAAC,CAAC;IAC5G,CAAC;IACD,OAAO;QACL,GAAG,CAAC,MAAM,CAAC,GAAG,CAAC,MAAM,CAAC,CAAC,CAAC,CAAC,EAAE,IAAI,EAAE,WAAW,CAAC,KAAK,CAAC,EAAE,CAAC,CAAC,CAAC,EAAE,CAAC;QAC3D,GAAG,CAAC,MAAM,CAAC,GAAG,CAAC,MAAM,CAAC,CAAC,CAAC,CAAC,EAAE,IAAI,EAAE,WAAW,CAAC,IAAI,CAAC,EAAE,CAAC,CAAC,CAAC,EAAE,CAAC;KAC3D,CAAC;AACJ,CAAC;AAOD,SAAS,WAAW,CAAC,KAAiB,EAAE,KAAgB,EAAE,YAAoB;IAC5E,MAAM,UAAU,GAAG,KAAK,CAAC,GAAG,GAAG,KAAK,CAAC,KAAK,CAAC;IAC3C,MAAM,UAAU,GAAG,IAAI,CAAC,GAAG,CAAC,CAAC,EAAE,YAAY,GAAG,IAAI,CAAC,GAAG,CAAC,YAAY,EAAE,UAAU,CAAC,CAAC,CAAC;IAClF,MAAM,IAAI,GAAG,IAAI,CAAC,GAAG,CAAC,kBAAkB,EAAE,UAAU,CAAC,CAAC;IACtD,MAAM,KAAK,GAAG,eAAe,CAAC,KAAK,EAAE,IAAI,CAAC,GAAG,CAAC,CAAC,EAAE,KAAK,CAAC,KAAK,GAAG,IAAI,CAAC,CAAC,CAAC;IACtE,MAAM,GAAG,GAAG,gBAAgB,CAAC,KAAK,EAAE,IAAI,CAAC,GAAG,CAAC,KAAK,CAAC,UAAU,EAAE,KAAK,GAAG,YAAY,CAAC,CAAC,CAAC;IACtF,OAAO,kBAAkB,CAAC,UAAU,CAAC,KAAK,EAAE,KAAK,EAAE,IAAI,CAAC,GAAG,CAAC,KAAK,EAAE,GAAG,CAAC,CAAC,CAAC,CAAC;AAC5E,CAAC;AAED;;;;;GAKG;AACH,MAAM,UAAU,SAAS,CACvB,GAAgB,EAChB,IAA8B,EAC9B,KAAwB,EACxB,YAAoB;IAEpB,MAAM,IAAI,GAAG,IAAI,EAAE,IAAI,IAAI,EAAE,CAAC;IAC9B,MAAM,KAAK,GAA0B,EAAE,CAAC;IACxC,IAAI,GAAG,CAAC,YAAY,CAAC,IAAI,CAAC,CAAC,EAAE,MAAM,EAAE,EAAE,EAAE,CAAC,MAAM,CAAC,QAAQ,CAAC,MAAM,CAAC,CAAC,EAAE,CAAC;QACnE,MAAM,KAAK,GAAG,IAAI,GAAG,CAAC,KAAK,CAAC,KAAK,CAAC,CAAC;QACnC,QAAQ,CAAC,IAAI,EAAE,CAAC,IAAI,EAAE,KAAK,EAAE,GAAG,EAAE,EAAE;YAClC,IAAI,CAAC,KAAK,CAAC,GAAG,CAAC,IAAI,CAAC;gBAAE,OAAO,KAAK,CAAC;YACnC,KAAK,CAAC,KAAK,GAAG,EAAE,KAAK,EAAE,GAAG,EAAE,CAAC;YAC7B,OAAO,IAAI,CAAC;QACd,CAAC,CAAC,CAAC;IACL,CAAC;IACD,OAAO,KAAK,CAAC,KAAK,KAAK,SAAS;QAC9B,CAAC,CAAC,WAAW,CAAC,IAAI,CAAC,IAAI,CAAC,EAAE,KAAK,CAAC,KAAK,EAAE,YAAY,CAAC;QACpD,CAAC,CAAC,WAAW,CAAC,IAAI,CAAC,IAAI,EAAE,WAAW,IAAI,EAAE,CAAC,EAAE,EAAE,KAAK,EAAE,CAAC,EAAE,GAAG,EAAE,CAAC,EAAE,EAAE,YAAY,CAAC,CAAC;AACrF,CAAC"}

package/dist/search/row-text.d.ts

@@ -0,0 +1,38 @@
+import { type SearchDocumentV1 } from "./contract.js";
+/** UTF-8 bytes of the text one document contributes to an index: id, title, description, type, tags, body. */
+export declare function documentSourceBytes(document: SearchDocumentV1): number;
+/** The normalized tokens of every field except the body, in document order. */
+export declare function fieldTokens(document: SearchDocumentV1): {
+    readonly id: readonly string[];
+    readonly title: readonly string[];
+    readonly type: readonly string[];
+    readonly tag: readonly string[];
+    readonly description: readonly string[];
+};
+export interface SearchRowTextV1 {
+    /** Each field's normalized tokens joined by one ASCII space, in document order. */
+    readonly idTerms: string;
+    readonly titleTerms: string;
+    readonly typeTerms: string;
+    readonly tagTerms: string;
+    readonly descriptionTerms: string;
+    /** The body's tokens in document order, keeping the first BODY_OCCURRENCE_CAP occurrences of each term. */
+    readonly bodyTerms: string;
+    readonly titlePhrase: string;
+    readonly idPhrase: string;
+    /** term -> its occurrences across the six column texts above (the body capped), in first-seen order. */
+    readonly instances: ReadonlyMap<string, number>;
+    /** Body tokens before the cap; over SEARCH_LIMITS.maxTermsPerDocument the index cannot hold the document. */
+    readonly bodyTermCount: number;
+}
+/** What one document contributes to any index: per-field token text, the two phrases, and the
+ * instance count per term. */
+export declare function searchRowText(document: SearchDocumentV1): SearchRowTextV1;
+export declare const FINGERPRINT_SAMPLE: readonly string[];
+/** The normalizer fingerprint of the running runtime: a short hash of normalize() and tokenize()
+ * over a fixed sample. */
+export declare function normalizerFingerprint(): string;
+/** `${SEARCH_ALGORITHM}/${SEARCH_TEXT_REVISION}/${normalizerFingerprint()}`: what a stored index row
+ * was derived with. */
+export declare function textVersion(): string;
+//# sourceMappingURL=row-text.d.ts.map

package/dist/search/row-text.d.ts.map

@@ -0,0 +1 @@
+{"version":3,"file":"row-text.d.ts","sourceRoot":"","sources":["../../src/search/row-text.ts"],"names":[],"mappings":"AAMA,OAAO,EAQL,KAAK,gBAAgB,EACtB,MAAM,eAAe,CAAC;AAGvB,8GAA8G;AAC9G,wBAAgB,mBAAmB,CAAC,QAAQ,EAAE,gBAAgB,GAAG,MAAM,CAStE;AAED,+EAA+E;AAC/E,wBAAgB,WAAW,CAAC,QAAQ,EAAE,gBAAgB,GAAG;IACvD,QAAQ,CAAC,EAAE,EAAE,SAAS,MAAM,EAAE,CAAC;IAC/B,QAAQ,CAAC,KAAK,EAAE,SAAS,MAAM,EAAE,CAAC;IAClC,QAAQ,CAAC,IAAI,EAAE,SAAS,MAAM,EAAE,CAAC;IACjC,QAAQ,CAAC,GAAG,EAAE,SAAS,MAAM,EAAE,CAAC;IAChC,QAAQ,CAAC,WAAW,EAAE,SAAS,MAAM,EAAE,CAAC;CACzC,CASA;AAED,MAAM,WAAW,eAAe;IAC9B,mFAAmF;IACnF,QAAQ,CAAC,OAAO,EAAE,MAAM,CAAC;IACzB,QAAQ,CAAC,UAAU,EAAE,MAAM,CAAC;IAC5B,QAAQ,CAAC,SAAS,EAAE,MAAM,CAAC;IAC3B,QAAQ,CAAC,QAAQ,EAAE,MAAM,CAAC;IAC1B,QAAQ,CAAC,gBAAgB,EAAE,MAAM,CAAC;IAClC,2GAA2G;IAC3G,QAAQ,CAAC,SAAS,EAAE,MAAM,CAAC;IAC3B,QAAQ,CAAC,WAAW,EAAE,MAAM,CAAC;IAC7B,QAAQ,CAAC,QAAQ,EAAE,MAAM,CAAC;IAC1B,wGAAwG;IACxG,QAAQ,CAAC,SAAS,EAAE,WAAW,CAAC,MAAM,EAAE,MAAM,CAAC,CAAC;IAChD,6GAA6G;IAC7G,QAAQ,CAAC,aAAa,EAAE,MAAM,CAAC;CAChC;AAED;8BAC8B;AAC9B,wBAAgB,aAAa,CAAC,QAAQ,EAAE,gBAAgB,GAAG,eAAe,CA+BzE;AAMD,eAAO,MAAM,kBAAkB,mBAM7B,CAAC;AAIH;0BAC0B;AAC1B,wBAAgB,qBAAqB,IAAI,MAAM,CAK9C;AAED;uBACuB;AACvB,wBAAgB,WAAW,IAAI,MAAM,CAEpC"}

package/dist/search/row-text.js

@@ -0,0 +1,88 @@
+/**
+ * What one document contributes to any search index, one owner. The in-memory source builds its
+ * term sets and volume from `fieldTokens` and the body scanner; a native index stores
+ * `searchRowText`'s column texts. Both read the same tokens, so their instance counts agree.
+ */
+import { sha256HexOfUtf8 } from "../sha256.js";
+import { BODY_OCCURRENCE_CAP, SEARCH_ALGORITHM, SEARCH_TEXT_REVISION, documentDescription, documentTags, documentTitle, documentType, } from "./contract.js";
+import { normalize, scanBody, tokenize, utf8 } from "./text.js";
+/** UTF-8 bytes of the text one document contributes to an index: id, title, description, type, tags, body. */
+export function documentSourceBytes(document) {
+    return (utf8(document.id).byteLength +
+        utf8(documentTitle(document) ?? "").byteLength +
+        utf8(documentDescription(document) ?? "").byteLength +
+        utf8(documentType(document) ?? "").byteLength +
+        documentTags(document).reduce((total, tag) => total + utf8(tag).byteLength, 0) +
+        utf8(document.body).byteLength);
+}
+/** The normalized tokens of every field except the body, in document order. */
+export function fieldTokens(document) {
+    const type = documentType(document);
+    return {
+        id: tokenize(document.id),
+        title: tokenize(documentTitle(document) ?? ""),
+        type: type === undefined ? [] : tokenize(type),
+        tag: documentTags(document).flatMap((tag) => tokenize(tag)),
+        description: tokenize(documentDescription(document) ?? ""),
+    };
+}
+/** What one document contributes to any index: per-field token text, the two phrases, and the
+ * instance count per term. */
+export function searchRowText(document) {
+    const fields = fieldTokens(document);
+    const instances = new Map();
+    const count = (term) => void instances.set(term, (instances.get(term) ?? 0) + 1);
+    for (const tokens of [fields.id, fields.title, fields.type, fields.tag, fields.description])
+        for (const term of tokens)
+            count(term);
+    const body = [];
+    const seen = new Map();
+    let bodyTermCount = 0;
+    scanBody(document.body, (term) => {
+        bodyTermCount += 1;
+        const occurrence = (seen.get(term) ?? 0) + 1;
+        seen.set(term, occurrence);
+        if (occurrence > BODY_OCCURRENCE_CAP)
+            return;
+        body.push(term);
+        count(term);
+    });
+    const idTerms = fields.id.join(" ");
+    const titleTerms = fields.title.join(" ");
+    return Object.freeze({
+        idTerms,
+        titleTerms,
+        typeTerms: fields.type.join(" "),
+        tagTerms: fields.tag.join(" "),
+        descriptionTerms: fields.description.join(" "),
+        bodyTerms: body.join(" "),
+        titlePhrase: titleTerms,
+        idPhrase: idTerms,
+        instances,
+        bodyTermCount,
+    });
+}
+// Compatibility characters, combining sequences, case pairs across scripts, scripts whose marks
+// split tokens, and characters from recent Unicode versions (Glagolitic U+2C2F and U+2C5F from 14.0,
+// U+1E030 from 15.0, Cyrillic TJE U+1C89 and U+1C8A from 16.0), so a runtime whose Unicode tables
+// differ prints another fingerprint.
+export const FINGERPRINT_SAMPLE = Object.freeze([
+    "fi fl ＡＢ ½ ℀ ㎃ ① Ⅸ ⁵ µ ㍿ ｶ ﾞ",
+    "café Å Å ñ 가 क़",
+    "ΟΔΟΣ Σίσυφος Dž DŽ İ ẞ Ꭰ ꭰ Ა Ⴀ Ꙁ",
+    "खोज ค้นหา البحث 中文 검색 եւ և",
+    "Ⱟⱟ \u{1e030} Ᲊᲊ",
+]);
+let fingerprint;
+/** The normalizer fingerprint of the running runtime: a short hash of normalize() and tokenize()
+ * over a fixed sample. */
+export function normalizerFingerprint() {
+    fingerprint ??= sha256HexOfUtf8(FINGERPRINT_SAMPLE.map((sample) => `${normalize(sample)}\u0001${tokenize(sample).join(" ")}`).join("\u0000")).slice(0, 16);
+    return fingerprint;
+}
+/** `${SEARCH_ALGORITHM}/${SEARCH_TEXT_REVISION}/${normalizerFingerprint()}`: what a stored index row
+ * was derived with. */
+export function textVersion() {
+    return `${SEARCH_ALGORITHM}/${SEARCH_TEXT_REVISION}/${normalizerFingerprint()}`;
+}
+//# sourceMappingURL=row-text.js.map

package/dist/search/row-text.js.map

@@ -0,0 +1 @@
+{"version":3,"file":"row-text.js","sourceRoot":"","sources":["../../src/search/row-text.ts"],"names":[],"mappings":"AAAA;;;;GAIG;AACH,OAAO,EAAE,eAAe,EAAE,MAAM,cAAc,CAAC;AAC/C,OAAO,EACL,mBAAmB,EACnB,gBAAgB,EAChB,oBAAoB,EACpB,mBAAmB,EACnB,YAAY,EACZ,aAAa,EACb,YAAY,GAEb,MAAM,eAAe,CAAC;AACvB,OAAO,EAAE,SAAS,EAAE,QAAQ,EAAE,QAAQ,EAAE,IAAI,EAAE,MAAM,WAAW,CAAC;AAEhE,8GAA8G;AAC9G,MAAM,UAAU,mBAAmB,CAAC,QAA0B;IAC5D,OAAO,CACL,IAAI,CAAC,QAAQ,CAAC,EAAE,CAAC,CAAC,UAAU;QAC5B,IAAI,CAAC,aAAa,CAAC,QAAQ,CAAC,IAAI,EAAE,CAAC,CAAC,UAAU;QAC9C,IAAI,CAAC,mBAAmB,CAAC,QAAQ,CAAC,IAAI,EAAE,CAAC,CAAC,UAAU;QACpD,IAAI,CAAC,YAAY,CAAC,QAAQ,CAAC,IAAI,EAAE,CAAC,CAAC,UAAU;QAC7C,YAAY,CAAC,QAAQ,CAAC,CAAC,MAAM,CAAC,CAAC,KAAK,EAAE,GAAG,EAAE,EAAE,CAAC,KAAK,GAAG,IAAI,CAAC,GAAG,CAAC,CAAC,UAAU,EAAE,CAAC,CAAC;QAC9E,IAAI,CAAC,QAAQ,CAAC,IAAI,CAAC,CAAC,UAAU,CAC/B,CAAC;AACJ,CAAC;AAED,+EAA+E;AAC/E,MAAM,UAAU,WAAW,CAAC,QAA0B;IAOpD,MAAM,IAAI,GAAG,YAAY,CAAC,QAAQ,CAAC,CAAC;IACpC,OAAO;QACL,EAAE,EAAE,QAAQ,CAAC,QAAQ,CAAC,EAAE,CAAC;QACzB,KAAK,EAAE,QAAQ,CAAC,aAAa,CAAC,QAAQ,CAAC,IAAI,EAAE,CAAC;QAC9C,IAAI,EAAE,IAAI,KAAK,SAAS,CAAC,CAAC,CAAC,EAAE,CAAC,CAAC,CAAC,QAAQ,CAAC,IAAI,CAAC;QAC9C,GAAG,EAAE,YAAY,CAAC,QAAQ,CAAC,CAAC,OAAO,CAAC,CAAC,GAAG,EAAE,EAAE,CAAC,QAAQ,CAAC,GAAG,CAAC,CAAC;QAC3D,WAAW,EAAE,QAAQ,CAAC,mBAAmB,CAAC,QAAQ,CAAC,IAAI,EAAE,CAAC;KAC3D,CAAC;AACJ,CAAC;AAmBD;8BAC8B;AAC9B,MAAM,UAAU,aAAa,CAAC,QAA0B;IACtD,MAAM,MAAM,GAAG,WAAW,CAAC,QAAQ,CAAC,CAAC;IACrC,MAAM,SAAS,GAAG,IAAI,GAAG,EAAkB,CAAC;IAC5C,MAAM,KAAK,GAAG,CAAC,IAAY,EAAQ,EAAE,CAAC,KAAK,SAAS,CAAC,GAAG,CAAC,IAAI,EAAE,CAAC,SAAS,CAAC,GAAG,CAAC,IAAI,CAAC,IAAI,CAAC,CAAC,GAAG,CAAC,CAAC,CAAC;IAC/F,KAAK,MAAM,MAAM,IAAI,CAAC,MAAM,CAAC,EAAE,EAAE,MAAM,CAAC,KAAK,EAAE,MAAM,CAAC,IAAI,EAAE,MAAM,CAAC,GAAG,EAAE,MAAM,CAAC,WAAW,CAAC;QACzF,KAAK,MAAM,IAAI,IAAI,MAAM;YAAE,KAAK,CAAC,IAAI,CAAC,CAAC;IACzC,MAAM,IAAI,GAAa,EAAE,CAAC;IAC1B,MAAM,IAAI,GAAG,IAAI,GAAG,EAAkB,CAAC;IACvC,IAAI,aAAa,GAAG,CAAC,CAAC;IACtB,QAAQ,CAAC,QAAQ,CAAC,IAAI,EAAE,CAAC,IAAI,EAAE,EAAE;QAC/B,aAAa,IAAI,CAAC,CAAC;QACnB,MAAM,UAAU,GAAG,CAAC,IAAI,CAAC,GAAG,CAAC,IAAI,CAAC,IAAI,CAAC,CAAC,GAAG,CAAC,CAAC;QAC7C,IAAI,CAAC,GAAG,CAAC,IAAI,EAAE,UAAU,CAAC,CAAC;QAC3B,IAAI,UAAU,GAAG,mBAAmB;YAAE,OAAO;QAC7C,IAAI,CAAC,IAAI,CAAC,IAAI,CAAC,CAAC;QAChB,KAAK,CAAC,IAAI,CAAC,CAAC;IACd,CAAC,CAAC,CAAC;IACH,MAAM,OAAO,GAAG,MAAM,CAAC,EAAE,CAAC,IAAI,CAAC,GAAG,CAAC,CAAC;IACpC,MAAM,UAAU,GAAG,MAAM,CAAC,KAAK,CAAC,IAAI,CAAC,GAAG,CAAC,CAAC;IAC1C,OAAO,MAAM,CAAC,MAAM,CAAC;QACnB,OAAO;QACP,UAAU;QACV,SAAS,EAAE,MAAM,CAAC,IAAI,CAAC,IAAI,CAAC,GAAG,CAAC;QAChC,QAAQ,EAAE,MAAM,CAAC,GAAG,CAAC,IAAI,CAAC,GAAG,CAAC;QAC9B,gBAAgB,EAAE,MAAM,CAAC,WAAW,CAAC,IAAI,CAAC,GAAG,CAAC;QAC9C,SAAS,EAAE,IAAI,CAAC,IAAI,CAAC,GAAG,CAAC;QACzB,WAAW,EAAE,UAAU;QACvB,QAAQ,EAAE,OAAO;QACjB,SAAS;QACT,aAAa;KACd,CAAC,CAAC;AACL,CAAC;AAED,gGAAgG;AAChG,qGAAqG;AACrG,kGAAkG;AAClG,qCAAqC;AACrC,MAAM,CAAC,MAAM,kBAAkB,GAAG,MAAM,CAAC,MAAM,CAAC;IAC9C,4BAA4B;IAC5B,qBAAqB;IACrB,gCAAgC;IAChC,4BAA4B;IAC5B,iBAAiB;CAClB,CAAC,CAAC;AAEH,IAAI,WAA+B,CAAC;AAEpC;0BAC0B;AAC1B,MAAM,UAAU,qBAAqB;IACnC,WAAW,KAAK,eAAe,CAC7B,kBAAkB,CAAC,GAAG,CAAC,CAAC,MAAM,EAAE,EAAE,CAAC,GAAG,SAAS,CAAC,MAAM,CAAC,SAAS,QAAQ,CAAC,MAAM,CAAC,CAAC,IAAI,CAAC,GAAG,CAAC,EAAE,CAAC,CAAC,IAAI,CAAC,QAAQ,CAAC,CAC7G,CAAC,KAAK,CAAC,CAAC,EAAE,EAAE,CAAC,CAAC;IACf,OAAO,WAAW,CAAC;AACrB,CAAC;AAED;uBACuB;AACvB,MAAM,UAAU,WAAW;IACzB,OAAO,GAAG,gBAAgB,IAAI,oBAAoB,IAAI,qBAAqB,EAAE,EAAE,CAAC;AAClF,CAAC"}