fix(infra): better search result (#7774)

packages/common/infra/src/sync/indexer/impl/indexeddb/inverted-index.ts

@@ -244,9 +244,12 @@ export class FullTextInvertedIndex implements InvertedIndex {
 
       // normalize score
       const maxScore = submatched.reduce((acc, s) => Math.max(acc, s.score), 0);
-      const minScore = submatched.reduce((acc, s) => Math.min(acc, s.score), 1);
+      const minScore = submatched.reduce((acc, s) => Math.min(acc, s.score), 0);
       for (const { nid, score, position } of submatched) {
-        const normalizedScore = (score - minScore) / (maxScore - minScore);
+        const normalizedScore =
+          maxScore === minScore
+            ? score
+            : (score - minScore) / (maxScore - minScore);
         const match = matched.get(nid) || {
           score: [] as number[],
           positions: new Map(),

packages/frontend/core/package.json

@@ -72,6 +72,7 @@
     "jotai-devtools": "^0.10.0",
     "jotai-effect": "^1.0.0",
     "jotai-scope": "^0.7.0",
+    "lib0": "^0.2.95",
     "lit": "^3.1.3",
     "lodash-es": "^4.17.21",
     "lottie-react": "^2.4.0",

packages/frontend/core/src/modules/docs-search/entities/docs-indexer.ts

@@ -23,11 +23,9 @@ export function isEmptyUpdate(binary: Uint8Array) {
 const logger = new DebugLogger('crawler');
 
 interface IndexerJobPayload {
-  docId: string;
   storageDocId: string;
 }
 
-// TODO(@eyhn): simplify this, it's too complex
 export class DocsIndexer extends Entity {
   private readonly jobQueue: JobQueue<IndexerJobPayload> =
     new IndexedDBJobQueue<IndexerJobPayload>(
@@ -72,13 +70,11 @@ export class DocsIndexer extends Entity {
         return;
       }
       if (event.clientId === this.workspaceEngine.doc.clientId) {
-        const docId = normalizeDocId(event.docId);
-
         this.jobQueue
           .enqueue([
             {
-              batchKey: docId,
-              payload: { docId, storageDocId: event.docId },
+              batchKey: event.docId,
+              payload: { storageDocId: event.docId },
             },
           ])
           .catch(err => {
@@ -93,18 +89,17 @@ export class DocsIndexer extends Entity {
       return;
     }
 
-    // jobs should have the same docId, so we just pick the first one
-    const docId = jobs[0].payload.docId;
+    // jobs should have the same storage docId, so we just pick the first one
     const storageDocId = jobs[0].payload.storageDocId;
 
     const worker = await this.ensureWorker(signal);
 
     const startTime = performance.now();
-    logger.debug('Start crawling job for docId:', docId);
+    logger.debug('Start crawling job for storageDocId:', storageDocId);
 
     let workerOutput;
 
-    if (docId === this.workspaceId) {
+    if (storageDocId === this.workspaceId) {
       const rootDocBuffer =
         await this.workspaceEngine.doc.storage.loadDocFromLocal(
           this.workspaceId
@@ -151,7 +146,7 @@ export class DocsIndexer extends Entity {
       workerOutput = await worker.run({
         type: 'doc',
         docBuffer,
-        docId,
+        storageDocId,
         rootDocBuffer,
       });
     }
@@ -190,13 +185,13 @@ export class DocsIndexer extends Entity {
         }
         await docIndexWriter.commit();
         const blockIndexWriter = await this.blockIndex.write();
-        for (const { blocks } of workerOutput.addedDoc) {
+        for (const { id, blocks } of workerOutput.addedDoc) {
           // delete old blocks
           const oldBlocks = await blockIndexWriter.search(
             {
               type: 'match',
               field: 'docId',
-              match: docId,
+              match: id,
             },
             {
               pagination: {
@@ -217,16 +212,20 @@ export class DocsIndexer extends Entity {
 
     if (workerOutput.reindexDoc) {
       await this.jobQueue.enqueue(
-        workerOutput.reindexDoc.map(({ docId, storageDocId }) => ({
-          batchKey: docId,
-          payload: { docId, storageDocId },
+        workerOutput.reindexDoc.map(({ storageDocId }) => ({
+          batchKey: storageDocId,
+          payload: { storageDocId },
         }))
       );
     }
 
     const duration = performance.now() - startTime;
     logger.debug(
-      'Finish crawling job for docId:' + docId + ' in ' + duration + 'ms '
+      'Finish crawling job for storageDocId:' +
+        storageDocId +
+        ' in ' +
+        duration +
+        'ms '
     );
   }
 
@@ -236,7 +235,7 @@ export class DocsIndexer extends Entity {
       .enqueue([
         {
           batchKey: this.workspaceId,
-          payload: { docId: this.workspaceId, storageDocId: this.workspaceId },
+          payload: { storageDocId: this.workspaceId },
         },
       ])
       .catch(err => {
@@ -255,47 +254,3 @@ export class DocsIndexer extends Entity {
     this.runner.stop();
   }
 }
-
-function normalizeDocId(raw: string) {
-  enum DocVariant {
-    Workspace = 'workspace',
-    Page = 'page',
-    Space = 'space',
-    Settings = 'settings',
-    Unknown = 'unknown',
-  }
-
-  try {
-    if (!raw.length) {
-      throw new Error('Invalid Empty Doc ID');
-    }
-
-    let parts = raw.split(':');
-
-    if (parts.length > 3) {
-      // special adapt case `wsId:space:page:pageId`
-      if (parts[1] === DocVariant.Space && parts[2] === DocVariant.Page) {
-        parts = [parts[0], DocVariant.Space, parts[3]];
-      } else {
-        throw new Error(`Invalid format of Doc ID: ${raw}`);
-      }
-    } else if (parts.length === 2) {
-      // `${variant}:${guid}`
-      throw new Error('not supported');
-    } else if (parts.length === 1) {
-      // ${ws} or ${pageId}
-      parts = ['', DocVariant.Unknown, parts[0]];
-    }
-
-    const docId = parts.at(2);
-
-    if (!docId) {
-      throw new Error('ID is required');
-    }
-
-    return docId;
-  } catch (err) {
-    logger.error('Error on normalize docId ' + raw, err);
-    return raw;
-  }
-}

packages/frontend/core/src/modules/docs-search/worker/in-worker.ts

@@ -1,6 +1,8 @@
 import type { AffineTextAttributes } from '@blocksuite/blocks';
 import type { DeltaInsert } from '@blocksuite/inline';
 import { Document } from '@toeverything/infra';
+import { toHexString } from 'lib0/buffer.js';
+import { digest as lib0Digest } from 'lib0/hash/sha256';
 import { difference } from 'lodash-es';
 import {
   applyUpdate,
@@ -18,24 +20,66 @@ import type {
   WorkerOutput,
 } from './types';
 
-function crawlingDocData({
+let cachedRootDoc: { doc: YDoc; hash: string } | null = null;
+
+async function digest(data: Uint8Array) {
+  if (
+    globalThis.crypto &&
+    globalThis.crypto.subtle &&
+    typeof globalThis.crypto.subtle.digest === 'function'
+  ) {
+    return new Uint8Array(
+      await globalThis.crypto.subtle.digest('SHA-256', data)
+    );
+  }
+  return lib0Digest(data);
+}
+
+async function crawlingDocData({
   docBuffer,
-  docId,
+  storageDocId,
   rootDocBuffer,
-}: WorkerInput & { type: 'doc' }): WorkerOutput {
-  const yRootDoc = new YDoc();
-  applyUpdate(yRootDoc, rootDocBuffer);
+}: WorkerInput & { type: 'doc' }): Promise<WorkerOutput> {
+  if (isEmptyUpdate(rootDocBuffer)) {
+    console.warn('[worker]: Empty root doc buffer');
+    return {};
+  }
+
+  const rootDocBufferHash = toHexString(await digest(rootDocBuffer));
+
+  let yRootDoc;
+  if (cachedRootDoc && cachedRootDoc.hash === rootDocBufferHash) {
+    yRootDoc = cachedRootDoc.doc;
+  } else {
+    yRootDoc = new YDoc();
+    applyUpdate(yRootDoc, rootDocBuffer);
+    cachedRootDoc = { doc: yRootDoc, hash: rootDocBufferHash };
+  }
+
+  let docId = null;
+  for (const [id, subdoc] of yRootDoc.getMap('spaces')) {
+    if (subdoc instanceof YDoc && storageDocId === subdoc.guid) {
+      docId = id;
+      break;
+    }
+  }
+
+  if (docId === null) {
+    return {};
+  }
 
   const ydoc = new YDoc();
 
-  applyUpdate(ydoc, docBuffer);
+  if (!isEmptyUpdate(docBuffer)) {
+    applyUpdate(ydoc, docBuffer);
+  }
 
   let docExists: boolean | null = null;
 
   (
     yRootDoc.getMap('meta').get('pages') as YArray<YMap<any>> | undefined
   )?.forEach(page => {
-    if (page.get('id') === docId) {
+    if (page.get('id') === storageDocId) {
       docExists = !(page.get('trash') ?? false);
     }
   });
@@ -283,7 +327,7 @@ function crawlingRootDocData({
   };
 }
 
-globalThis.onmessage = (event: MessageEvent<WorkerIngoingMessage>) => {
+globalThis.onmessage = async (event: MessageEvent<WorkerIngoingMessage>) => {
   const message = event.data;
   if (message.type === 'init') {
     postMessage({ type: 'init', msgId: message.msgId });
@@ -296,7 +340,7 @@ globalThis.onmessage = (event: MessageEvent<WorkerIngoingMessage>) => {
       if (input.type === 'rootDoc') {
         data = crawlingRootDocData(input);
       } else {
-        data = crawlingDocData(input);
+        data = await crawlingDocData(input);
       }
 
       postMessage({ type: 'done', msgId: message.msgId, output: data });
@@ -311,3 +355,10 @@ globalThis.onmessage = (event: MessageEvent<WorkerIngoingMessage>) => {
 };
 
 declare function postMessage(message: WorkerOutgoingMessage): void;
+
+function isEmptyUpdate(binary: Uint8Array) {
+  return (
+    binary.byteLength === 0 ||
+    (binary.byteLength === 2 && binary[0] === 0 && binary[1] === 0)
+  );
+}

packages/frontend/core/src/modules/docs-search/worker/types.ts

@@ -34,7 +34,7 @@ export type WorkerInput =
     }
   | {
       type: 'doc';
-      docId: string;
+      storageDocId: string;
       rootDocBuffer: Uint8Array;
       docBuffer: Uint8Array;
     };

yarn.lock

@@ -452,6 +452,7 @@ __metadata:
     jotai-devtools: "npm:^0.10.0"
     jotai-effect: "npm:^1.0.0"
     jotai-scope: "npm:^0.7.0"
+    lib0: "npm:^0.2.95"
     lit: "npm:^3.1.3"
     lodash-es: "npm:^4.17.21"
     lottie-react: "npm:^2.4.0"