fix: search with exact: true

packages/orama/package.json

@@ -81,7 +81,9 @@
     }
   },
   "types": "./dist/commonjs/index.d.ts",
-  "files": ["dist"],
+  "files": [
+    "dist"
+  ],
   "repository": {
     "type": "git",
     "url": "https://github.com/oramasearch/orama"
@@ -169,7 +171,10 @@
       "./components": "./src/components.ts",
       "./trees": "./src/trees.ts"
     },
-    "esmDialects": ["deno", "browser"]
+    "esmDialects": [
+      "deno",
+      "browser"
+    ]
   },
   "module": "./dist/esm/index.js"
 }

packages/orama/src/components/index.ts

@@ -459,8 +459,8 @@ export function search(
   tokenizer: Tokenizer,
   language: string | undefined,
   propertiesToSearch: string[],
-  exact: boolean,
   tolerance: number,
+  exactToken: boolean,
   boost: Record<string, number>,
   relevance: Required<BM25Params>,
   docsCount: number,
@@ -500,7 +500,7 @@ export function search(
     const tokenLength = tokens.length
     for (let i = 0; i < tokenLength; i++) {
       const token = tokens[i]
-      const searchResult = tree.node.find({ term: token, exact, tolerance })
+      const searchResult = tree.node.find({ term: token, exact: exactToken, tolerance })
 
       // See if this token was found (for threshold=0 filtering)
       const termsFound = Object.keys(searchResult)

packages/orama/src/methods/search-fulltext.ts

@@ -1,9 +1,10 @@
 import { getFacets } from '../components/facets.js'
 import { getGroups } from '../components/groups.js'
 import { runAfterSearch, runBeforeSearch } from '../components/hooks.js'
-import { getInternalDocumentId } from '../components/internal-document-id-store.js'
+import { getInternalDocumentId, InternalDocumentID } from '../components/internal-document-id-store.js'
 import { Language } from '../components/tokenizer/languages.js'
 import { createError } from '../errors.js'
+import { getNested } from '../utils.js'
 import type {
   AnyOrama,
   BM25Params,
@@ -22,7 +23,7 @@ export function innerFullTextSearch<T extends AnyOrama>(
   orama: T,
   params: Pick<
     SearchParamsFullText<T>,
-    'term' | 'properties' | 'where' | 'exact' | 'tolerance' | 'boost' | 'relevance' | 'threshold'
+    'term' | 'properties' | 'where' | 'exact' | 'tolerance' | 'boost' | 'relevance' | 'threshold' | 'exactToken'
   >,
   language: Language | undefined
 ) {
@@ -66,16 +67,35 @@ export function innerFullTextSearch<T extends AnyOrama>(
   //   in this case, we need to return all the documents that contains at least one of the given properties
   const threshold = params.threshold !== undefined && params.threshold !== null ? params.threshold : 1
 
-  if (term || properties) {
+  if (params.exact && term) {
+    const docs = orama.documentsStore.getAll(orama.data.docs) as Record<InternalDocumentID, TypedDocument<T>>
+    const normalizeTerm= term.toLowerCase()
+
+    uniqueDocsIDs = Object.entries(docs)
+    .filter(([, doc]) => {
+      return propertiesToSearch.some((prop) => {
+        const value = getNested(doc, prop)
+        if (typeof value === 'string') {
+          return value.toLowerCase() === normalizeTerm
+        }
+        if (Array.isArray(value)) {
+          return value.some((v) => typeof v === 'string' && v.toLowerCase() === normalizeTerm)
+        }
+        return false
+      })
+    })
+    .map(([id,]) => [+id, 0] as TokenScore)
+  }
+  else if (term || properties) {
     const docsCount = count(orama)
     uniqueDocsIDs = orama.index.search(
       index,
       term || '',
       orama.tokenizer,
       language,
       propertiesToSearch,
-      params.exact || false,
       params.tolerance || 0,
+      params.exactToken || false,
       params.boost || {},
       applyDefault(params.relevance),
       docsCount,

packages/orama/src/types.ts

@@ -318,6 +318,11 @@ export interface SearchParamsFullText<T extends AnyOrama, ResultDocument = Typed
    */
   exact?: boolean
 
+  /**
+   * Whether each token should be matched exactly.
+   */
+  exactToken?: boolean
+
   /**
    * The maximum [levenshtein distance](https://en.wikipedia.org/wiki/Levenshtein_distance)
    * between the term and the searchable property.
@@ -982,8 +987,8 @@ export interface IIndex<I extends AnyIndexStore> {
     tokenizer: Tokenizer,
     language: string | undefined,
     propertiesToSearch: string[],
-    exact: boolean,
     tolerance: number,
+    exactToken: boolean,
     boost: Partial<Record<OnlyStrings<FlattenSchemaProperty<T>[]>, number>>,
     relevance: Required<BM25Params>,
     docsCount: number,

packages/orama/tests/dataset.test.ts

@@ -108,9 +108,9 @@ t.test("orama.dataset", async (t) => {
       Object.keys((db.data.docs as DocumentsStore).docs).length,
       (dataset as EventJson).result.events.length,
     );
-    t.equal(s1.count, 1117);
-    t.equal(s2.count, 7314);
-    t.equal(s3.count, 7314);
+    t.equal(s1.count, 1081);
+    t.equal(s2.count, 0);
+    t.equal(s3.count, 1842);
 
     t.end();
   });
@@ -143,7 +143,7 @@ t.test("orama.dataset", async (t) => {
     const s1 = removeVariadicData(
       await search(db, {
         term: "war",
-        exact: true,
+        exactToken: true,
         // eslint-disable-next-line
         // @ts-ignore
         properties: ["description"],
@@ -155,7 +155,7 @@ t.test("orama.dataset", async (t) => {
     const s2 = removeVariadicData(
       await search(db, {
         term: "war",
-        exact: true,
+        exactToken: true,
         properties: ["description"],
         limit: 10,
         offset: 10,
@@ -165,7 +165,7 @@ t.test("orama.dataset", async (t) => {
     const s3 = removeVariadicData(
       await search(db, {
         term: "war",
-        exact: true,
+        exactToken: true,
         properties: ["description"],
         limit: 10,
         offset: 20,
@@ -174,15 +174,15 @@ t.test("orama.dataset", async (t) => {
 
     const s4 = await search(db, {
       term: "war",
-      exact: true,
+      exactToken: true,
       properties: ["description"],
       limit: 2240,
       offset: 0,
     });
 
     const s5 = await search(db, {
       term: "war",
-      exact: true,
+      exactToken: true,
       properties: ["description"],
       limit: 10,
       offset: 2239,
@@ -223,7 +223,7 @@ t.test("orama.dataset", async (t) => {
   t.test("should correctly delete documents", async (t) => {
     const documentsToDelete = await search(db, {
       term: "war",
-      exact: true,
+      exactToken: true,
       properties: ["description"],
       limit: 10,
       offset: 0,
@@ -235,13 +235,12 @@ t.test("orama.dataset", async (t) => {
 
     const newSearch = await search(db, {
       term: "war",
-      exact: true,
       properties: ["description"],
       limit: 10,
       offset: 0,
     });
 
-    t.equal(newSearch.count, 2347);
+    t.equal(newSearch.count, 2743);
 
     t.end();
   });

packages/orama/tests/search.test.ts

@@ -84,11 +84,11 @@ t.test('search method', async (t) => {
       await insert(db, { quote: 'I like cats. They are the best.', author: 'Jane Doe' })
 
       // Exact search
-      const result1 = await search(db, { term: 'fox', exact: true })
+      const result1 = await search(db, { term: 'John Doe', properties: ["author"], exact: true })
       const result2 = await search(db, { term: 'dog', exact: true })
 
       t.equal(result1.count, 2)
-      t.equal(result2.count, 3)
+      t.equal(result2.count, 0)
 
       // Prefix search
       const result3 = await search(db, { term: 'fox', exact: false })
@@ -193,15 +193,15 @@ t.test('search method', async (t) => {
 
       const partialSearch = await search(db, {
         term: 'alr',
-        exact: true
+        exactToken: true
       })
 
       t.equal(partialSearch.count, 0)
       t.strictSame(partialSearch.hits, [])
 
       const exactSearch = await search(db, {
         term: 'already',
-        exact: true
+        exactToken: true
       })
 
       t.equal(exactSearch.count, 1)
@@ -719,11 +719,11 @@ t.test('search method', async (t) => {
 
     const result1 = await search(db, { term: 'foxes', exact: true })
     const result2 = await search(db, { term: 'cats', exact: true })
-    const result3 = await search(db, { term: 'brown', exact: true })
+    const result3 = await search(db, { term: 'John Doe', exact: true })
 
     t.equal(result1.count, 0)
     t.equal(result2.count, 0)
-    t.equal(result3.count, 1)
+    t.equal(result3.count, 2)
     t.end()
   })
 

packages/tokenizers/package.json

@@ -30,7 +30,9 @@
   "dependencies": {
     "@orama/orama": "workspace:*"
   },
-  "files": ["dist"],
+  "files": [
+    "dist"
+  ],
   "repository": {
     "type": "git",
     "url": "https://github.com/oramasearch/orama"
@@ -65,7 +67,10 @@
     "node": ">= 20.0.0"
   },
   "tshy": {
-    "dialects": ["esm", "commonjs"],
+    "dialects": [
+      "esm",
+      "commonjs"
+    ],
     "exports": {
       "./japanese": "./src/japanese.ts",
       "./mandarin": "./src/mandarin.ts",