fix: improve hashblob performance up to 100% faster (#2144)

* refactor: directly return boolean expression

* perf: avoid creating new Uint8Array if it already is

* perf: avoid wrapping result in Uint8Array

`hashObject` already returns a `Uint8Array`

* perf: only create one buffer instead of three

* refactor: reintroduce explicit encoding

Author: lukecotter

js/isomorphic-git/index.cjs

@@ -644,7 +644,7 @@ async function testSubtleSHA1() {
   // some browsers that have crypto.subtle.digest don't actually implement SHA-1.
   try {
     const hash = await subtleSHA1(new Uint8Array([]));
-    if (hash === 'da39a3ee5e6b4b0d3255bfef95601890afd80709') return true
+    return hash === 'da39a3ee5e6b4b0d3255bfef95601890afd80709'
   } catch (_) {
     // no bother
   }
@@ -2301,14 +2301,41 @@ class GitTree {
   }
 }
 
+/**
+ * Represents a Git object and provides methods to wrap and unwrap Git objects
+ * according to the Git object format.
+ */
 class GitObject {
+  /**
+   * Wraps a raw object with a Git header.
+   *
+   * @param {Object} params - The parameters for wrapping.
+   * @param {string} params.type - The type of the Git object (e.g., 'blob', 'tree', 'commit').
+   * @param {Uint8Array} params.object - The raw object data to wrap.
+   * @returns {Uint8Array} The wrapped Git object as a single buffer.
+   */
   static wrap({ type, object }) {
-    return Buffer.concat([
-      Buffer.from(`${type} ${object.byteLength.toString()}\x00`),
-      Buffer.from(object),
-    ])
+    const header = `${type} ${object.length}\x00`;
+    const headerLen = header.length;
+    const totalLength = headerLen + object.length;
+
+    // Allocate a single buffer for the header and object, rather than create multiple buffers
+    const wrappedObject = new Uint8Array(totalLength);
+    for (let i = 0; i < headerLen; i++) {
+      wrappedObject[i] = header.charCodeAt(i);
+    }
+    wrappedObject.set(object, headerLen);
+
+    return wrappedObject
   }
 
+  /**
+   * Unwraps a Git object buffer into its type and raw object data.
+   *
+   * @param {Buffer|Uint8Array} buffer - The buffer containing the wrapped Git object.
+   * @returns {{ type: string, object: Buffer }} An object containing the type and the raw object data.
+   * @throws {InternalError} If the length specified in the header does not match the actual object length.
+   */
   static unwrap(buffer) {
     const s = buffer.indexOf(32); // first space
     const i = buffer.indexOf(0); // first null value
@@ -10752,17 +10779,18 @@ async function hashBlob({ object }) {
     // Convert object to buffer
     if (typeof object === 'string') {
       object = Buffer.from(object, 'utf8');
-    } else {
-      object = Buffer.from(object);
+    } else if (!(object instanceof Uint8Array)) {
+      object = new Uint8Array(object);
     }
 
     const type = 'blob';
     const { oid, object: _object } = await hashObject({
-      type: 'blob',
+      type,
       format: 'content',
       object,
     });
-    return { oid, type, object: new Uint8Array(_object), format: 'wrapped' }
+
+    return { oid, type, object: _object, format: 'wrapped' }
   } catch (err) {
     err.caller = 'git.hashBlob';
     throw err

js/isomorphic-git/index.js

@@ -638,7 +638,7 @@ async function testSubtleSHA1() {
   // some browsers that have crypto.subtle.digest don't actually implement SHA-1.
   try {
     const hash = await subtleSHA1(new Uint8Array([]));
-    if (hash === 'da39a3ee5e6b4b0d3255bfef95601890afd80709') return true
+    return hash === 'da39a3ee5e6b4b0d3255bfef95601890afd80709'
   } catch (_) {
     // no bother
   }
@@ -2295,14 +2295,41 @@ class GitTree {
   }
 }
 
+/**
+ * Represents a Git object and provides methods to wrap and unwrap Git objects
+ * according to the Git object format.
+ */
 class GitObject {
+  /**
+   * Wraps a raw object with a Git header.
+   *
+   * @param {Object} params - The parameters for wrapping.
+   * @param {string} params.type - The type of the Git object (e.g., 'blob', 'tree', 'commit').
+   * @param {Uint8Array} params.object - The raw object data to wrap.
+   * @returns {Uint8Array} The wrapped Git object as a single buffer.
+   */
   static wrap({ type, object }) {
-    return Buffer.concat([
-      Buffer.from(`${type} ${object.byteLength.toString()}\x00`),
-      Buffer.from(object),
-    ])
+    const header = `${type} ${object.length}\x00`;
+    const headerLen = header.length;
+    const totalLength = headerLen + object.length;
+
+    // Allocate a single buffer for the header and object, rather than create multiple buffers
+    const wrappedObject = new Uint8Array(totalLength);
+    for (let i = 0; i < headerLen; i++) {
+      wrappedObject[i] = header.charCodeAt(i);
+    }
+    wrappedObject.set(object, headerLen);
+
+    return wrappedObject
   }
 
+  /**
+   * Unwraps a Git object buffer into its type and raw object data.
+   *
+   * @param {Buffer|Uint8Array} buffer - The buffer containing the wrapped Git object.
+   * @returns {{ type: string, object: Buffer }} An object containing the type and the raw object data.
+   * @throws {InternalError} If the length specified in the header does not match the actual object length.
+   */
   static unwrap(buffer) {
     const s = buffer.indexOf(32); // first space
     const i = buffer.indexOf(0); // first null value
@@ -10746,17 +10773,18 @@ async function hashBlob({ object }) {
     // Convert object to buffer
     if (typeof object === 'string') {
       object = Buffer.from(object, 'utf8');
-    } else {
-      object = Buffer.from(object);
+    } else if (!(object instanceof Uint8Array)) {
+      object = new Uint8Array(object);
     }
 
     const type = 'blob';
     const { oid, object: _object } = await hashObject({
-      type: 'blob',
+      type,
       format: 'content',
       object,
     });
-    return { oid, type, object: new Uint8Array(_object), format: 'wrapped' }
+
+    return { oid, type, object: _object, format: 'wrapped' }
   } catch (err) {
     err.caller = 'git.hashBlob';
     throw err