fix: ut8-string encode/decode (#76)

* fix: ut8-string encode/decode There was a bug by which we were storing the strings as unicode bytes instead of utf8 bytes. This was a bug since the specification clearly says that the encoding must be utf8. This commit fixes such bug using the TextEncode / TextDecode tools, which are widely supported by modern browsers and node versions. * fix: utf8 without TextEncoder/Decoder * more tests
near · Nov 23, 2023 · abe2701 · abe2701
1 parent dc9ad87
commit abe2701
Show file tree

Hide file tree

Showing 13 changed files with 221 additions and 17 deletions.
diff --git a/borsh-ts/deserialize.ts b/borsh-ts/deserialize.ts
@@ -54,7 +54,26 @@ export class BorshDeserializer {
     decode_string(): string {
         const len: number = this.decode_integer('u32') as number;
         const buffer = new Uint8Array(this.buffer.consume_bytes(len));
-        return String.fromCharCode.apply(null, buffer);
+
+        // decode utf-8 string without using TextDecoder
+        // first get all bytes to single byte code points
+        const codePoints = [];
+        for (let i = 0; i < len; ++i) {
+            const byte = buffer[i];
+            if (byte < 0x80) {
+                codePoints.push(byte);
+            } else if (byte < 0xE0) {
+                codePoints.push(((byte & 0x1F) << 6) | (buffer[++i] & 0x3F));
+            } else if (byte < 0xF0) {
+                codePoints.push(((byte & 0x0F) << 12) | ((buffer[++i] & 0x3F) << 6) | (buffer[++i] & 0x3F));
+            } else {
+                const codePoint = ((byte & 0x07) << 18) | ((buffer[++i] & 0x3F) << 12) | ((buffer[++i] & 0x3F) << 6) | (buffer[++i] & 0x3F);
+                codePoints.push(codePoint);
+            }
+        }
+
+        // then decode code points to utf-8
+        return String.fromCodePoint(...codePoints);
     }
 
     decode_boolean(): boolean {

diff --git a/borsh-ts/serialize.ts b/borsh-ts/serialize.ts
@@ -63,13 +63,32 @@ export class BorshSerializer {
         this.checkTypes && utils.expect_type(value, 'string', this.fieldPath);
         const _value = value as string;
 
-        // 4 bytes for length
-        this.encoded.store_value(_value.length, 'u32');
-
-        // string bytes
+        // encode to utf8 bytes without using TextEncoder
+        const utf8Bytes: number[] = [];
         for (let i = 0; i < _value.length; i++) {
-            this.encoded.store_value(_value.charCodeAt(i), 'u8');
+            let charCode = _value.charCodeAt(i);
+
+            if (charCode < 0x80) {
+                utf8Bytes.push(charCode);
+            } else if (charCode < 0x800) {
+                utf8Bytes.push(0xc0 | (charCode >> 6), 0x80 | (charCode & 0x3f));
+            } else if (charCode < 0xd800 || charCode >= 0xe000) {
+                utf8Bytes.push(0xe0 | (charCode >> 12), 0x80 | ((charCode >> 6) & 0x3f), 0x80 | (charCode & 0x3f));
+            } else {
+                i++;
+                charCode = 0x10000 + (((charCode & 0x3ff) << 10) | (_value.charCodeAt(i) & 0x3ff));
+                utf8Bytes.push(
+                    0xf0 | (charCode >> 18),
+                    0x80 | ((charCode >> 12) & 0x3f),
+                    0x80 | ((charCode >> 6) & 0x3f),
+                    0x80 | (charCode & 0x3f),
+                );
+            }
         }
+
+        // 4 bytes for length + string bytes
+        this.encoded.store_value(utf8Bytes.length, 'u32');
+        this.encoded.store_bytes(new Uint8Array(utf8Bytes));
     }
 
     encode_boolean(value: unknown): void {

diff --git a/borsh-ts/test/(de)serialize.test.js b/borsh-ts/test/(de)serialize.test.js
@@ -40,6 +40,10 @@ test('serialize booleans', async () => {
 
 test('serialize strings', async () => {
     check_roundtrip('h"i', 'string', [3, 0, 0, 0, 104, 34, 105]);
+    check_roundtrip('Chévere', 'string', [8, 0, 0, 0, 67, 104, 195, 169, 118, 101, 114, 101]);
+    check_roundtrip('!ǬЇЉي࠺👍ઠ൧࿄ሒᘻᏠᬅᡝ࠻', 'string', [43, 0, 0, 0, 33, 199, 172, 208, 135, 208, 137, 217, 138, 224, 160, 186, 240, 159, 145, 141, 224, 170, 160, 224, 181, 167, 224, 191, 132, 225, 136, 146, 225, 152, 187, 225, 143, 160, 225, 172, 133, 225, 161, 157, 224, 160, 187]);
+    check_roundtrip('óñ@‡؏ث 漢࠶⭐🔒􀀀', 'string', [30, 0, 0, 0, 195, 179, 195, 177, 64, 226, 128, 161, 216, 143, 216, 171, 32, 230, 188, 162, 224, 160, 182, 226, 173, 144, 240, 159, 148, 146, 244, 128, 128, 128]);
+    check_roundtrip('f © bar 𝌆 baz ☃ qux', 'string', [25, 0, 0, 0, 102, 32, 194, 169, 32, 98, 97, 114, 32, 240, 157, 140, 134, 32, 98, 97, 122, 32, 226, 152, 131, 32, 113, 117, 120]);
 });
 
 test('serialize floats', async () => {

diff --git a/borsh-ts/utils.ts b/borsh-ts/utils.ts
@@ -118,4 +118,34 @@ function validate_struct_schema(schema: { [key: string]: Schema }): void {
     for (const key in schema) {
         validate_schema(schema[key]);
     }
+}
+
+// utf-8 encode
+export function encodeCodePoint(codePoint): number[] {
+    if ((codePoint & 0xFFFFFF80) == 0) { // 1-byte sequence
+        return [codePoint];
+    }
+    let symbol: number[] = [];
+    if ((codePoint & 0xFFFFF800) == 0) { // 2-byte sequence
+        symbol = [((codePoint >> 6) & 0x1F) | 0xC0];
+    }
+    else if ((codePoint & 0xFFFF0000) == 0) { // 3-byte sequence
+        symbol = [
+            ((codePoint >> 12) & 0x0F) | 0xE0,
+            createByte(codePoint, 6)
+        ];
+    }
+    else if ((codePoint & 0xFFE00000) == 0) { // 4-byte sequence
+        symbol = [
+            ((codePoint >> 18) & 0x07) | 0xF0,
+            createByte(codePoint, 12),
+            createByte(codePoint, 6)
+        ];
+    }
+    symbol.push((codePoint & 0x3F) | 0x80);
+    return symbol;
+}
+
+function createByte(codePoint, shift): number {
+    return ((codePoint >> shift) & 0x3F) | 0x80;
 }
diff --git a/lib/cjs/deserialize.js b/lib/cjs/deserialize.js
@@ -55,7 +55,27 @@ var BorshDeserializer = /** @class */ (function () {
     BorshDeserializer.prototype.decode_string = function () {
         var len = this.decode_integer('u32');
         var buffer = new Uint8Array(this.buffer.consume_bytes(len));
-        return String.fromCharCode.apply(null, buffer);
+        // decode utf-8 string without using TextDecoder
+        // first get all bytes to single byte code points
+        var codePoints = [];
+        for (var i = 0; i < len; ++i) {
+            var byte = buffer[i];
+            if (byte < 0x80) {
+                codePoints.push(byte);
+            }
+            else if (byte < 0xE0) {
+                codePoints.push(((byte & 0x1F) << 6) | (buffer[++i] & 0x3F));
+            }
+            else if (byte < 0xF0) {
+                codePoints.push(((byte & 0x0F) << 12) | ((buffer[++i] & 0x3F) << 6) | (buffer[++i] & 0x3F));
+            }
+            else {
+                var codePoint = ((byte & 0x07) << 18) | ((buffer[++i] & 0x3F) << 12) | ((buffer[++i] & 0x3F) << 6) | (buffer[++i] & 0x3F);
+                codePoints.push(codePoint);
+            }
+        }
+        // then decode code points to utf-8
+        return String.fromCodePoint.apply(String, codePoints);
     };
     BorshDeserializer.prototype.decode_boolean = function () {
         return this.buffer.consume_value('u8') > 0;

diff --git a/lib/cjs/serialize.js b/lib/cjs/serialize.js
@@ -84,12 +84,28 @@ var BorshSerializer = /** @class */ (function () {
     BorshSerializer.prototype.encode_string = function (value) {
         this.checkTypes && utils.expect_type(value, 'string', this.fieldPath);
         var _value = value;
-        // 4 bytes for length
-        this.encoded.store_value(_value.length, 'u32');
-        // string bytes
+        // encode to utf8 bytes without using TextEncoder
+        var utf8Bytes = [];
         for (var i = 0; i < _value.length; i++) {
-            this.encoded.store_value(_value.charCodeAt(i), 'u8');
+            var charCode = _value.charCodeAt(i);
+            if (charCode < 0x80) {
+                utf8Bytes.push(charCode);
+            }
+            else if (charCode < 0x800) {
+                utf8Bytes.push(0xc0 | (charCode >> 6), 0x80 | (charCode & 0x3f));
+            }
+            else if (charCode < 0xd800 || charCode >= 0xe000) {
+                utf8Bytes.push(0xe0 | (charCode >> 12), 0x80 | ((charCode >> 6) & 0x3f), 0x80 | (charCode & 0x3f));
+            }
+            else {
+                i++;
+                charCode = 0x10000 + (((charCode & 0x3ff) << 10) | (_value.charCodeAt(i) & 0x3ff));
+                utf8Bytes.push(0xf0 | (charCode >> 18), 0x80 | ((charCode >> 12) & 0x3f), 0x80 | ((charCode >> 6) & 0x3f), 0x80 | (charCode & 0x3f));
+            }
         }
+        // 4 bytes for length + string bytes
+        this.encoded.store_value(utf8Bytes.length, 'u32');
+        this.encoded.store_bytes(new Uint8Array(utf8Bytes));
     };
     BorshSerializer.prototype.encode_boolean = function (value) {
         this.checkTypes && utils.expect_type(value, 'boolean', this.fieldPath);

diff --git a/lib/cjs/utils.d.ts b/lib/cjs/utils.d.ts
@@ -8,3 +8,4 @@ export declare class ErrorSchema extends Error {
     constructor(schema: Schema, expected: string);
 }
 export declare function validate_schema(schema: Schema): void;
+export declare function encodeCodePoint(codePoint: any): number[];
diff --git a/lib/cjs/utils.js b/lib/cjs/utils.js
@@ -15,7 +15,7 @@ var __extends = (this && this.__extends) || (function () {
     };
 })();
 exports.__esModule = true;
-exports.validate_schema = exports.ErrorSchema = exports.expect_enum = exports.expect_same_size = exports.expect_bigint = exports.expect_type = exports.isArrayLike = void 0;
+exports.encodeCodePoint = exports.validate_schema = exports.ErrorSchema = exports.expect_enum = exports.expect_same_size = exports.expect_bigint = exports.expect_type = exports.isArrayLike = void 0;
 var types_js_1 = require("./types.js");
 function isArrayLike(value) {
     // source: https://stackoverflow.com/questions/24048547/checking-if-an-object-is-array-like
@@ -132,3 +132,32 @@ function validate_struct_schema(schema) {
         validate_schema(schema[key]);
     }
 }
+// utf-8 encode
+function encodeCodePoint(codePoint) {
+    if ((codePoint & 0xFFFFFF80) == 0) { // 1-byte sequence
+        return [codePoint];
+    }
+    var symbol = [];
+    if ((codePoint & 0xFFFFF800) == 0) { // 2-byte sequence
+        symbol = [((codePoint >> 6) & 0x1F) | 0xC0];
+    }
+    else if ((codePoint & 0xFFFF0000) == 0) { // 3-byte sequence
+        symbol = [
+            ((codePoint >> 12) & 0x0F) | 0xE0,
+            createByte(codePoint, 6)
+        ];
+    }
+    else if ((codePoint & 0xFFE00000) == 0) { // 4-byte sequence
+        symbol = [
+            ((codePoint >> 18) & 0x07) | 0xF0,
+            createByte(codePoint, 12),
+            createByte(codePoint, 6)
+        ];
+    }
+    symbol.push((codePoint & 0x3F) | 0x80);
+    return symbol;
+}
+exports.encodeCodePoint = encodeCodePoint;
+function createByte(codePoint, shift) {
+    return ((codePoint >> shift) & 0x3F) | 0x80;
+}
diff --git a/lib/esm/deserialize.js b/lib/esm/deserialize.js
@@ -52,7 +52,27 @@ var BorshDeserializer = /** @class */ (function () {
     BorshDeserializer.prototype.decode_string = function () {
         var len = this.decode_integer('u32');
         var buffer = new Uint8Array(this.buffer.consume_bytes(len));
-        return String.fromCharCode.apply(null, buffer);
+        // decode utf-8 string without using TextDecoder
+        // first get all bytes to single byte code points
+        var codePoints = [];
+        for (var i = 0; i < len; ++i) {
+            var byte = buffer[i];
+            if (byte < 0x80) {
+                codePoints.push(byte);
+            }
+            else if (byte < 0xE0) {
+                codePoints.push(((byte & 0x1F) << 6) | (buffer[++i] & 0x3F));
+            }
+            else if (byte < 0xF0) {
+                codePoints.push(((byte & 0x0F) << 12) | ((buffer[++i] & 0x3F) << 6) | (buffer[++i] & 0x3F));
+            }
+            else {
+                var codePoint = ((byte & 0x07) << 18) | ((buffer[++i] & 0x3F) << 12) | ((buffer[++i] & 0x3F) << 6) | (buffer[++i] & 0x3F);
+                codePoints.push(codePoint);
+            }
+        }
+        // then decode code points to utf-8
+        return String.fromCodePoint.apply(String, codePoints);
     };
     BorshDeserializer.prototype.decode_boolean = function () {
         return this.buffer.consume_value('u8') > 0;

diff --git a/lib/esm/serialize.js b/lib/esm/serialize.js
@@ -58,12 +58,28 @@ var BorshSerializer = /** @class */ (function () {
     BorshSerializer.prototype.encode_string = function (value) {
         this.checkTypes && utils.expect_type(value, 'string', this.fieldPath);
         var _value = value;
-        // 4 bytes for length
-        this.encoded.store_value(_value.length, 'u32');
-        // string bytes
+        // encode to utf8 bytes without using TextEncoder
+        var utf8Bytes = [];
         for (var i = 0; i < _value.length; i++) {
-            this.encoded.store_value(_value.charCodeAt(i), 'u8');
+            var charCode = _value.charCodeAt(i);
+            if (charCode < 0x80) {
+                utf8Bytes.push(charCode);
+            }
+            else if (charCode < 0x800) {
+                utf8Bytes.push(0xc0 | (charCode >> 6), 0x80 | (charCode & 0x3f));
+            }
+            else if (charCode < 0xd800 || charCode >= 0xe000) {
+                utf8Bytes.push(0xe0 | (charCode >> 12), 0x80 | ((charCode >> 6) & 0x3f), 0x80 | (charCode & 0x3f));
+            }
+            else {
+                i++;
+                charCode = 0x10000 + (((charCode & 0x3ff) << 10) | (_value.charCodeAt(i) & 0x3ff));
+                utf8Bytes.push(0xf0 | (charCode >> 18), 0x80 | ((charCode >> 12) & 0x3f), 0x80 | ((charCode >> 6) & 0x3f), 0x80 | (charCode & 0x3f));
+            }
         }
+        // 4 bytes for length + string bytes
+        this.encoded.store_value(utf8Bytes.length, 'u32');
+        this.encoded.store_bytes(new Uint8Array(utf8Bytes));
     };
     BorshSerializer.prototype.encode_boolean = function (value) {
         this.checkTypes && utils.expect_type(value, 'boolean', this.fieldPath);

diff --git a/lib/esm/utils.d.ts b/lib/esm/utils.d.ts
@@ -8,3 +8,4 @@ export declare class ErrorSchema extends Error {
     constructor(schema: Schema, expected: string);
 }
 export declare function validate_schema(schema: Schema): void;
+export declare function encodeCodePoint(codePoint: any): number[];
diff --git a/lib/esm/utils.js b/lib/esm/utils.js
@@ -123,3 +123,31 @@ function validate_struct_schema(schema) {
         validate_schema(schema[key]);
     }
 }
+// utf-8 encode
+export function encodeCodePoint(codePoint) {
+    if ((codePoint & 0xFFFFFF80) == 0) { // 1-byte sequence
+        return [codePoint];
+    }
+    var symbol = [];
+    if ((codePoint & 0xFFFFF800) == 0) { // 2-byte sequence
+        symbol = [((codePoint >> 6) & 0x1F) | 0xC0];
+    }
+    else if ((codePoint & 0xFFFF0000) == 0) { // 3-byte sequence
+        symbol = [
+            ((codePoint >> 12) & 0x0F) | 0xE0,
+            createByte(codePoint, 6)
+        ];
+    }
+    else if ((codePoint & 0xFFE00000) == 0) { // 4-byte sequence
+        symbol = [
+            ((codePoint >> 18) & 0x07) | 0xF0,
+            createByte(codePoint, 12),
+            createByte(codePoint, 6)
+        ];
+    }
+    symbol.push((codePoint & 0x3F) | 0x80);
+    return symbol;
+}
+function createByte(codePoint, shift) {
+    return ((codePoint >> shift) & 0x3F) | 0x80;
+}
diff --git a/lib/types/utils.d.ts b/lib/types/utils.d.ts
@@ -8,3 +8,4 @@ export declare class ErrorSchema extends Error {
     constructor(schema: Schema, expected: string);
 }
 export declare function validate_schema(schema: Schema): void;
+export declare function encodeCodePoint(codePoint: any): number[];