fix(NODE-5363): defer byte slicing to utf8 decoding API in nodejs (#585)

Co-authored-by: Durran Jordan <[email protected]>

Author: W-A-James

etc/benchmarks/bson_versions.json

@@ -1,10 +1,6 @@
 {
   "versions": [
-    "1.1.6",
     "4.6",
-    "5.0",
-    "5.1",
-    "5.2",
     "5.3"
   ]
 }

etc/benchmarks/install_bson_versions.sh

@@ -1,5 +1,6 @@
 #!/bin/bash
-versions=$(jq '.versions' < bson_versions.json | sed -E 's/(\[|\]|,|")//g')
+# To be run from repo root
+versions=$(jq '.versions' < etc/benchmarks/bson_versions.json | sed -E 's/(\[|\]|,|")//g')
 installVersions=''
 for bson in $versions; do
   versionNoDot=$(echo $bson | tr -d '.')

etc/benchmarks/main.mjs

@@ -129,6 +129,61 @@ await runner({
   }
 });
 
+await runner({
+  skip: true,
+  name: 'deserialize a large batch of documents each with an array of many Int32s',
+  iterations,
+  setup(libs) {
+    const bson = libs[0].lib;
+    return bson.serialize({
+      nextBatch: Array.from({ length: 1000 }, () => ({
+        _id: new bson.ObjectId(),
+        arrayField: Array.from({ length: 100 }, (_, i) => i)
+      }))
+    });
+  },
+  async run(i, bson, document) {
+    await Promise.all(
+      Array.from(
+        { length: 100 },
+        (_, i) =>
+          new Promise(resolve => {
+            setTimeout(() => {
+              resolve(bson.lib.deserialize(document, { validation: { utf8: false } }));
+            }, 20);
+          })
+      )
+    );
+  }
+});
+
+await runner({
+  skip: true,
+  name: 'deserialize a large batch of documents each with an array of many Int64s',
+  iterations,
+  setup(libs) {
+    const bson = libs[0].lib;
+    return bson.serialize({
+      nextBatch: Array.from({ length: 1000 }, () => ({
+        _id: new bson.ObjectId(),
+        arrayField: Array.from({ length: 100 }, (_, i) => bson.Long.fromInt(i))
+      }))
+    });
+  },
+  async run(i, bson, document) {
+    await Promise.all(
+      Array.from(
+        { length: 100 },
+        (_, i) =>
+          new Promise(resolve => {
+            setTimeout(() => {
+              resolve(bson.lib.deserialize(document, { validation: { utf8: false } }));
+            }, 20);
+          })
+      )
+    );
+  }
+});
 // End
 console.log(
   'Total time taken to benchmark:',

src/binary.ts

@@ -223,8 +223,9 @@ export class Binary extends BSONValue {
   toString(encoding?: 'hex' | 'base64' | 'utf8' | 'utf-8'): string {
     if (encoding === 'hex') return ByteUtils.toHex(this.buffer);
     if (encoding === 'base64') return ByteUtils.toBase64(this.buffer);
-    if (encoding === 'utf8' || encoding === 'utf-8') return ByteUtils.toUTF8(this.buffer);
-    return ByteUtils.toUTF8(this.buffer);
+    if (encoding === 'utf8' || encoding === 'utf-8')
+      return ByteUtils.toUTF8(this.buffer, 0, this.buffer.byteLength);
+    return ByteUtils.toUTF8(this.buffer, 0, this.buffer.byteLength);
   }
 
   /** @internal */

src/parser/deserializer.ts

@@ -236,7 +236,7 @@ function deserializeObject(
     if (i >= buffer.byteLength) throw new BSONError('Bad BSON Document: illegal CString');
 
     // Represents the key
-    const name = isArray ? arrayIndex++ : ByteUtils.toUTF8(buffer.subarray(index, i));
+    const name = isArray ? arrayIndex++ : ByteUtils.toUTF8(buffer, index, i);
 
     // shouldValidateKey is true if the key should be validated, false otherwise
     let shouldValidateKey = true;
@@ -476,7 +476,7 @@ function deserializeObject(
       // If are at the end of the buffer there is a problem with the document
       if (i >= buffer.length) throw new BSONError('Bad BSON Document: illegal CString');
       // Return the C string
-      const source = ByteUtils.toUTF8(buffer.subarray(index, i));
+      const source = ByteUtils.toUTF8(buffer, index, i);
       // Create the regexp
       index = i + 1;
 
@@ -489,7 +489,7 @@ function deserializeObject(
       // If are at the end of the buffer there is a problem with the document
       if (i >= buffer.length) throw new BSONError('Bad BSON Document: illegal CString');
       // Return the C string
-      const regExpOptions = ByteUtils.toUTF8(buffer.subarray(index, i));
+      const regExpOptions = ByteUtils.toUTF8(buffer, index, i);
       index = i + 1;
 
       // For each option add the corresponding one for javascript
@@ -521,7 +521,7 @@ function deserializeObject(
       // If are at the end of the buffer there is a problem with the document
       if (i >= buffer.length) throw new BSONError('Bad BSON Document: illegal CString');
       // Return the C string
-      const source = ByteUtils.toUTF8(buffer.subarray(index, i));
+      const source = ByteUtils.toUTF8(buffer, index, i);
       index = i + 1;
 
       // Get the start search index
@@ -533,7 +533,7 @@ function deserializeObject(
       // If are at the end of the buffer there is a problem with the document
       if (i >= buffer.length) throw new BSONError('Bad BSON Document: illegal CString');
       // Return the C string
-      const regExpOptions = ByteUtils.toUTF8(buffer.subarray(index, i));
+      const regExpOptions = ByteUtils.toUTF8(buffer, index, i);
       index = i + 1;
 
       // Set the object
@@ -678,7 +678,7 @@ function deserializeObject(
           throw new BSONError('Invalid UTF-8 string in BSON document');
         }
       }
-      const namespace = ByteUtils.toUTF8(buffer.subarray(index, index + stringSize - 1));
+      const namespace = ByteUtils.toUTF8(buffer, index, index + stringSize - 1);
       // Update parse index position
       index = index + stringSize;
 
@@ -735,7 +735,7 @@ function getValidatedString(
   end: number,
   shouldValidateUtf8: boolean
 ) {
-  const value = ByteUtils.toUTF8(buffer.subarray(start, end));
+  const value = ByteUtils.toUTF8(buffer, start, end);
   // if utf8 validation is on, do the check
   if (shouldValidateUtf8) {
     for (let i = 0; i < value.length; i++) {

src/utils/byte_utils.ts

@@ -26,7 +26,7 @@ export type ByteUtils = {
   /** Create a Uint8Array containing utf8 code units from a string */
   fromUTF8: (text: string) => Uint8Array;
   /** Create a string from utf8 code units */
-  toUTF8: (buffer: Uint8Array) => string;
+  toUTF8: (buffer: Uint8Array, start: number, end: number) => string;
   /** Get the utf8 code unit count from a string if it were to be transformed to utf8 */
   utf8ByteLength: (input: string) => number;
   /** Encode UTF8 bytes generated from `source` string into `destination` at byteOffset. Returns the number of bytes encoded. */

src/utils/node_byte_utils.ts

@@ -5,7 +5,7 @@ type NodeJsBuffer = ArrayBufferView &
   Uint8Array & {
     write(string: string, offset: number, length: undefined, encoding: 'utf8'): number;
     copy(target: Uint8Array, targetStart: number, sourceStart: number, sourceEnd: number): number;
-    toString: (this: Uint8Array, encoding: NodeJsEncoding) => string;
+    toString: (this: Uint8Array, encoding: NodeJsEncoding, start?: number, end?: number) => string;
     equals: (this: Uint8Array, other: Uint8Array) => boolean;
   };
 type NodeJsBufferConstructor = Omit<Uint8ArrayConstructor, 'from'> & {
@@ -125,8 +125,8 @@ export const nodeJsByteUtils = {
     return Buffer.from(text, 'utf8');
   },
 
-  toUTF8(buffer: Uint8Array): string {
-    return nodeJsByteUtils.toLocalBufferType(buffer).toString('utf8');
+  toUTF8(buffer: Uint8Array, start: number, end: number): string {
+    return nodeJsByteUtils.toLocalBufferType(buffer).toString('utf8', start, end);
   },
 
   utf8ByteLength(input: string): number {

src/utils/web_byte_utils.ts

@@ -172,8 +172,8 @@ export const webByteUtils = {
     return new TextEncoder().encode(text);
   },
 
-  toUTF8(uint8array: Uint8Array): string {
-    return new TextDecoder('utf8', { fatal: false }).decode(uint8array);
+  toUTF8(uint8array: Uint8Array, start: number, end: number): string {
+    return new TextDecoder('utf8', { fatal: false }).decode(uint8array.slice(start, end));
   },
 
   utf8ByteLength(input: string): number {