@exodus/bytes 1.0.0-rc.3 → 1.0.0-rc.5

package/fallback/utf8.js

@@ -0,0 +1,280 @@
+export const E_STRICT = 'Input is not well-formed utf8'
+export const E_STRICT_UNICODE = 'Input is not well-formed Unicode'
+
+const replacementPoint = 0xff_fd
+
+// https://encoding.spec.whatwg.org/#utf-8-decoder
+// We are most likely in loose mode, for non-loose escape & decodeURIComponent solved everything
+export function decode(arr, loose) {
+  const start = 0
+  const end = arr.length
+  let out = ''
+  const tmp = []
+
+  for (let i = start; i < end; i++) {
+    if (tmp.length > 0x2_00) {
+      // far below MAX_ARGUMENTS_LENGTH in npmjs.com/buffer, we use smaller chunks
+      // length can be off by a few as large code points produce two utf-16 char codes, also we overshoot in unrolled loop
+      out += String.fromCharCode.apply(String, tmp)
+      tmp.length = 0
+    }
+
+    const byte = arr[i]
+    if (byte < 0x80) {
+      // Fast path ascii
+      tmp.push(byte)
+      // Unroll the loop a bit for faster ops, overshoot by 20 chars
+      for (let j = 0; j < 5; j++) {
+        if (i + 1 >= end) break
+        const byte1 = arr[i + 1]
+        if (byte1 >= 0x80) break
+        tmp.push(byte1)
+        i++
+        if (i + 1 >= end) break
+        const byte2 = arr[i + 1]
+        if (byte2 >= 0x80) break
+        tmp.push(byte2)
+        i++
+        if (i + 1 >= end) break
+        const byte3 = arr[i + 1]
+        if (byte3 >= 0x80) break
+        tmp.push(byte3)
+        i++
+        if (i + 1 >= end) break
+        const byte4 = arr[i + 1]
+        if (byte4 >= 0x80) break
+        tmp.push(byte4)
+        i++
+      }
+    } else if (byte < 0xc2) {
+      if (!loose) throw new TypeError(E_STRICT)
+      tmp.push(replacementPoint)
+    } else if (byte < 0xe0) {
+      // need 1 more
+      if (i + 1 >= end) {
+        if (!loose) throw new TypeError(E_STRICT)
+        tmp.push(replacementPoint)
+        break
+      }
+
+      const byte1 = arr[i + 1]
+      if (byte1 < 0x80 || byte1 > 0xbf) {
+        if (!loose) throw new TypeError(E_STRICT)
+        tmp.push(replacementPoint)
+        continue
+      }
+
+      i++
+      tmp.push(((byte & 0x1f) << 6) | (byte1 & 0x3f))
+    } else if (byte < 0xf0) {
+      // need 2 more
+      if (i + 1 >= end) {
+        if (!loose) throw new TypeError(E_STRICT)
+        tmp.push(replacementPoint)
+        break
+      }
+
+      const lower = byte === 0xe0 ? 0xa0 : 0x80
+      const upper = byte === 0xed ? 0x9f : 0xbf
+      const byte1 = arr[i + 1]
+      if (byte1 < lower || byte1 > upper) {
+        if (!loose) throw new TypeError(E_STRICT)
+        tmp.push(replacementPoint)
+        continue
+      }
+
+      i++
+      if (i + 1 >= end) {
+        if (!loose) throw new TypeError(E_STRICT)
+        tmp.push(replacementPoint)
+        break
+      }
+
+      const byte2 = arr[i + 1]
+      if (byte2 < 0x80 || byte2 > 0xbf) {
+        if (!loose) throw new TypeError(E_STRICT)
+        tmp.push(replacementPoint)
+        continue
+      }
+
+      i++
+      tmp.push(((byte & 0xf) << 12) | ((byte1 & 0x3f) << 6) | (byte2 & 0x3f))
+    } else if (byte <= 0xf4) {
+      // need 3 more
+      if (i + 1 >= end) {
+        if (!loose) throw new TypeError(E_STRICT)
+        tmp.push(replacementPoint)
+        break
+      }
+
+      const lower = byte === 0xf0 ? 0x90 : 0x80
+      const upper = byte === 0xf4 ? 0x8f : 0xbf
+      const byte1 = arr[i + 1]
+      if (byte1 < lower || byte1 > upper) {
+        if (!loose) throw new TypeError(E_STRICT)
+        tmp.push(replacementPoint)
+        continue
+      }
+
+      i++
+      if (i + 1 >= end) {
+        if (!loose) throw new TypeError(E_STRICT)
+        tmp.push(replacementPoint)
+        break
+      }
+
+      const byte2 = arr[i + 1]
+      if (byte2 < 0x80 || byte2 > 0xbf) {
+        if (!loose) throw new TypeError(E_STRICT)
+        tmp.push(replacementPoint)
+        continue
+      }
+
+      i++
+      if (i + 1 >= end) {
+        if (!loose) throw new TypeError(E_STRICT)
+        tmp.push(replacementPoint)
+        break
+      }
+
+      const byte3 = arr[i + 1]
+      if (byte3 < 0x80 || byte3 > 0xbf) {
+        if (!loose) throw new TypeError(E_STRICT)
+        tmp.push(replacementPoint)
+        continue
+      }
+
+      i++
+      const codePoint =
+        ((byte & 0xf) << 18) | ((byte1 & 0x3f) << 12) | ((byte2 & 0x3f) << 6) | (byte3 & 0x3f)
+      if (codePoint > 0xff_ff) {
+        // split into char codes as String.fromCharCode is faster than String.fromCodePoint
+        const u = codePoint - 0x1_00_00
+        tmp.push(0xd8_00 + ((u >> 10) & 0x3_ff), 0xdc_00 + (u & 0x3_ff))
+      } else {
+        tmp.push(codePoint)
+      }
+      // eslint-disable-next-line sonarjs/no-duplicated-branches
+    } else {
+      if (!loose) throw new TypeError(E_STRICT)
+      tmp.push(replacementPoint)
+    }
+  }
+
+  if (tmp.length > 0) out += String.fromCharCode.apply(String, tmp)
+  return out
+}
+
+export function encode(string, loose) {
+  const length = string.length
+  let lead = null
+  let small = true
+  let bytes = new Uint8Array(length) // assume ascii
+  let p = 0
+
+  for (let i = 0; i < length; i++) {
+    const code = string.charCodeAt(i)
+    if (code < 0x80) {
+      // Fast path for ascii
+      if (lead) {
+        if (!loose) throw new TypeError(E_STRICT_UNICODE)
+        bytes[p++] = 0xef
+        bytes[p++] = 0xbf
+        bytes[p++] = 0xbd
+        lead = null
+      }
+
+      bytes[p++] = code
+      // Unroll the loop a bit for faster ops
+      for (let j = 0; j < 5; j++) {
+        if (i + 1 >= length) break
+        const c1 = string.charCodeAt(i + 1)
+        if (c1 >= 0x80) break
+        bytes[p++] = c1
+        i++
+        if (i + 1 >= length) break
+        const c2 = string.charCodeAt(i + 1)
+        if (c2 >= 0x80) break
+        bytes[p++] = c2
+        i++
+        if (i + 1 >= length) break
+        const c3 = string.charCodeAt(i + 1)
+        if (c3 >= 0x80) break
+        bytes[p++] = c3
+        i++
+        if (i + 1 >= length) break
+        const c4 = string.charCodeAt(i + 1)
+        if (c4 >= 0x80) break
+        bytes[p++] = c4
+        i++
+      }
+
+      continue
+    }
+
+    if (small) {
+      // TODO: use resizable array buffers? will have to return a non-resizeable one
+      const bytesNew = new Uint8Array(length * 3) // maximium can be 3x of the string length in charcodes
+      bytesNew.set(bytes)
+      bytes = bytesNew
+      small = false
+    }
+
+    // surrogate, charcodes = [d800 + a & 3ff, dc00 + b & 3ff]; codePoint = 0x1_00_00 | (a << 10) | b
+    // lead: d800 - dbff
+    // trail: dc00 - dfff
+    if (code >= 0xd8_00 && code < 0xe0_00) {
+      if (lead && code < 0xdc_00) {
+        // a second lead, meaning the previous one was unpaired
+        if (!loose) throw new TypeError(E_STRICT_UNICODE)
+        bytes[p++] = 0xef
+        bytes[p++] = 0xbf
+        bytes[p++] = 0xbd
+        lead = null
+        // code is still processed as a new lead
+      }
+
+      if (!lead) {
+        if (code > 0xdb_ff || i + 1 >= length) {
+          // lead out of range || unpaired
+          if (!loose) throw new TypeError(E_STRICT_UNICODE)
+          bytes[p++] = 0xef
+          bytes[p++] = 0xbf
+          bytes[p++] = 0xbd
+          continue
+        }
+
+        lead = code
+        continue
+      }
+
+      // here, codePoint is always between 0x1_00_00 and 0x11_00_00, we encode as 4 bytes
+      const codePoint = (((lead - 0xd8_00) << 10) | (code - 0xdc_00)) + 0x1_00_00
+      bytes[p++] = (codePoint >> 18) | 0xf0
+      bytes[p++] = ((codePoint >> 12) & 0x3f) | 0x80
+      bytes[p++] = ((codePoint >> 6) & 0x3f) | 0x80
+      bytes[p++] = (codePoint & 0x3f) | 0x80
+      lead = null
+      continue
+    } else if (lead) {
+      if (!loose) throw new TypeError(E_STRICT_UNICODE)
+      bytes[p++] = 0xef
+      bytes[p++] = 0xbf
+      bytes[p++] = 0xbd
+      lead = null
+      // code is still processed
+    }
+
+    // We are left with a non-pair char code above ascii, it gets encoded to 2 or 3 bytes
+    if (code < 0x8_00) {
+      bytes[p++] = (code >> 6) | 0xc0
+      bytes[p++] = (code & 0x3f) | 0x80
+    } else {
+      bytes[p++] = (code >> 12) | 0xe0
+      bytes[p++] = ((code >> 6) & 0x3f) | 0x80
+      bytes[p++] = (code & 0x3f) | 0x80
+    }
+  }
+
+  return bytes.length === p ? bytes : bytes.slice(0, p)
+}

package/hex.js

@@ -1,4 +1,4 @@
-import { assertTypedArray } from './assert.js'
+import { assertUint8 } from './assert.js'
 import { typedView } from './array.js'
 import * as js from './fallback/hex.js'
 
@@ -6,9 +6,10 @@ const { Buffer } = globalThis // Buffer is optional, only used when native
 const haveNativeBuffer = Buffer && !Buffer.TYPED_ARRAY_SUPPORT
 const { toHex: webHex } = Uint8Array.prototype // Modern engines have this
 
+const { E_HEX } = js
+
 export function toHex(arr) {
-  assertTypedArray(arr)
-  if (!(arr instanceof Uint8Array)) arr = new Uint8Array(arr.buffer, arr.byteOffset, arr.byteLength)
+  assertUint8(arr)
   if (arr.length === 0) return ''
   if (webHex && arr.toHex === webHex) return arr.toHex()
   if (!haveNativeBuffer) return js.toHex(arr)
@@ -19,4 +20,12 @@ export function toHex(arr) {
 // Unlike Buffer.from(), throws on invalid input
 export const fromHex = Uint8Array.fromHex
   ? (str, format = 'uint8') => typedView(Uint8Array.fromHex(str), format)
-  : (str, format = 'uint8') => typedView(js.fromHex(str), format)
+  : haveNativeBuffer
+    ? (str, format = 'uint8') => {
+        if (typeof str !== 'string') throw new TypeError('Input is not a string')
+        if (str.length % 2 !== 0) throw new SyntaxError(E_HEX)
+        const buf = Buffer.from(str, 'hex') // will stop on first non-hex character, so we can just validate length
+        if (buf.length * 2 !== str.length) throw new SyntaxError(E_HEX)
+        return typedView(buf, format)
+      }
+    : (str, format = 'uint8') => typedView(js.fromHex(str), format)

package/package.json

@@ -1,12 +1,11 @@
 {
   "name": "@exodus/bytes",
-  "version": "1.0.0-rc.3",
+  "version": "1.0.0-rc.5",
   "description": "Various operations on Uint8Array data",
   "scripts": {
     "lint": "eslint .",
-    "test:v8": "npm run test:d8 --",
     "test:javascriptcore": "npm run test:jsc --",
-    "test:d8": "exodus-test --engine=d8:bundle",
+    "test:v8": "exodus-test --engine=v8:bundle",
     "test:jsc": "exodus-test --engine=jsc:bundle",
     "test:spidermonkey": "exodus-test --engine=spidermonkey:bundle",
     "test:hermes": "exodus-test --engine=hermes:bundle",
@@ -39,32 +38,62 @@
   },
   "type": "module",
   "files": [
+    "/fallback/_utils.js",
+    "/fallback/base32.js",
     "/fallback/base64.js",
     "/fallback/hex.js",
+    "/fallback/utf8.js",
     "/array.js",
     "/assert.js",
+    "/base32.js",
+    "/base58.js",
+    "/base58check.js",
     "/base64.js",
-    "/hex.js"
+    "/hex.js",
+    "/utf8.js"
   ],
   "exports": {
     "./array.js": "./array.js",
+    "./base32.js": "./base32.js",
+    "./base58.js": "./base58.js",
+    "./base58check.js": "./base58check.js",
     "./base64.js": "./base64.js",
-    "./hex.js": "./hex.js"
+    "./hex.js": "./hex.js",
+    "./utf8.js": "./utf8.js"
+  },
+  "peerDependencies": {
+    "@exodus/crypto": "^1.0.0-rc.4"
+  },
+  "peerDependenciesMeta": {
+    "@exodus/crypto": {
+      "optional": true
+    }
   },
-  "dependencies": {},
   "devDependencies": {
+    "@ethersproject/strings": "^5.8.0",
+    "@exodus/crypto": "1.0.0-rc.29",
     "@exodus/eslint-config": "^5.24.0",
     "@exodus/prettier": "^1.0.0",
-    "@exodus/test": "^1.0.0-rc.105",
+    "@exodus/test": "^1.0.0-rc.107",
+    "@noble/hashes": "^2.0.1",
     "@scure/base": "^1.2.6",
+    "@stablelib/base64": "^2.0.1",
+    "@stablelib/hex": "^2.0.1",
     "@types/node": "^24.0.10",
     "base-x": "^5.0.1",
     "base32.js": "^0.1.0",
     "base64-js": "^1.5.1",
+    "bs58": "^6.0.0",
+    "bs58check": "^4.0.0",
+    "bstring": "^0.3.9",
     "buffer": "^6.0.3",
     "electron": "36.5.0",
     "eslint": "^8.44.0",
-    "jsvu": "^3.0.0"
+    "fast-base64-decode": "^2.0.0",
+    "fast-base64-encode": "^1.0.0",
+    "hi-base32": "^0.5.1",
+    "jsvu": "^3.0.0",
+    "text-encoding": "^0.7.0"
   },
   "prettier": "@exodus/prettier",
   "packageManager": "pnpm@10.12.1+sha256.889bac470ec93ccc3764488a19d6ba8f9c648ad5e50a9a6e4be3768a5de387a3"

package/utf8.js

@@ -0,0 +1,117 @@
+import { assertUint8 } from './assert.js'
+import { typedView } from './array.js'
+import * as js from './fallback/utf8.js'
+
+const { Buffer, TextEncoder, TextDecoder, decodeURIComponent, escape } = globalThis // Buffer is optional
+const haveNativeBuffer = Buffer && !Buffer.TYPED_ARRAY_SUPPORT
+const isNative = (x) => x && (haveNativeBuffer || `${x}`.includes('[native code]')) // we consider Node.js TextDecoder/TextEncoder native
+const haveDecoder = isNative(TextDecoder)
+const nativeEncoder = isNative(TextEncoder) ? new TextEncoder() : null
+// ignoreBOM: true means that BOM will be left as-is, i.e. will be present in the output
+// We don't want to strip anything unexpectedly
+const decoderFatal = haveDecoder ? new TextDecoder('utf8', { ignoreBOM: true, fatal: true }) : null
+const decoderLoose = haveDecoder ? new TextDecoder('utf8', { ignoreBOM: true }) : null
+const { isWellFormed } = String.prototype
+
+const { E_STRICT, E_STRICT_UNICODE } = js
+
+const shouldUseEscapePath = Boolean(globalThis.HermesInternal) // faster only on Hermes, js path beats it on normal engines
+
+function deLoose(str, loose, res) {
+  if (loose) return res
+  if (isWellFormed) {
+    // We have a fast native method
+    if (isWellFormed.call(str)) return res
+    throw new TypeError(E_STRICT_UNICODE)
+  }
+
+  // Recheck if the string was encoded correctly
+  let start = 0
+  const last = res.length - 2
+  // Search for EFBFBD
+  while (start < last) {
+    const pos = res.indexOf(0xef, start)
+    if (pos === -1) break
+    start = pos + 1
+    if (res[pos + 1] === 0xbf && res[pos + 2] === 0xbd) {
+      // Found a replacement char in output, need to recheck if we encoded the input correctly
+      if (str !== decode(res)) throw new TypeError(E_STRICT_UNICODE)
+      return res
+    }
+  }
+
+  return res
+}
+
+function encode(str, loose = false) {
+  if (typeof str !== 'string') throw new TypeError('Input is not a string')
+  if (haveNativeBuffer) return deLoose(str, loose, Buffer.from(str)) // faster on ascii on Node.js
+  if (nativeEncoder) return deLoose(str, loose, nativeEncoder.encode(str)) // Node.js, browsers, and Hermes
+  // No reason to use unescape + encodeURIComponent: it's slower than JS on normal engines, and modern Hermes already has TextEncoder
+  return js.encode(str, loose)
+}
+
+let escapes
+
+function toEscapesPart(arr, start, end) {
+  let o = ''
+  let i = start
+  const last3 = end - 3
+  // Unrolled loop is faster
+  while (i < last3) {
+    const a = arr[i++]
+    const b = arr[i++]
+    const c = arr[i++]
+    const d = arr[i++]
+    o += escapes[a]
+    o += escapes[b]
+    o += escapes[c]
+    o += escapes[d]
+  }
+
+  while (i < end) o += escapes[arr[i++]]
+  return o
+}
+
+function decode(arr, loose = false) {
+  assertUint8(arr)
+  if (haveDecoder) return loose ? decoderLoose.decode(arr) : decoderFatal.decode(arr) // Node.js and browsers
+  // No reason to use native Buffer: it's not faster than TextDecoder, needs rechecks in non-loose mode, and Node.js has TextDecoder
+
+  // This codepath gives a ~2x perf boost on Hermes
+  if (shouldUseEscapePath && escape && decodeURIComponent) {
+    if (!escapes) escapes = Array.from({ length: 256 }, (_, i) => escape(String.fromCharCode(i)))
+    const length = arr.length
+    let o
+    if (length > 30_000) {
+      // Limit concatenation to avoid excessive GC
+      // TODO: recheck thresholds on Hermes (taken from hex)
+      const concat = []
+      for (let i = 0; i < length; ) {
+        const step = i + 500
+        const end = step > length ? length : step
+        concat.push(toEscapesPart(arr, i, end))
+        i = end
+      }
+
+      o = concat.join('')
+      concat.length = 0
+    } else {
+      o = toEscapesPart(arr, 0, length)
+    }
+
+    try {
+      return decodeURIComponent(o) // asci to utf8, escape() is precalucated
+    } catch {
+      if (!loose) throw new TypeError(E_STRICT)
+      // Ok, we have to use manual implementation for loose decoder
+    }
+  }
+
+  return js.decode(arr, loose)
+}
+
+export const utf8fromString = (str, format = 'uint8') => typedView(encode(str, false), format)
+export const utf8fromStringLoose = (str, format = 'uint8') => typedView(encode(str, true), format)
+export const utf8toString = (arr) => decode(arr, false)
+export const utf8toStringLoose = (arr) => decode(arr, true)