@exodus/bytes 1.0.0-rc.7 → 1.0.0-rc.9

package/utf16.node.js

@@ -0,0 +1,79 @@
+import { nativeDecoder, isDeno, isLE } from './fallback/_utils.js'
+import { E_STRICT, E_STRICT_UNICODE } from './fallback/utf16.js'
+
+if (Buffer.TYPED_ARRAY_SUPPORT) throw new Error('Unexpected Buffer polyfill')
+
+const { isWellFormed } = String.prototype
+const to8 = (a) => new Uint8Array(a.buffer, a.byteOffset, a.byteLength)
+
+// Unlike utf8, operates on Uint16Arrays by default
+
+function encode(str, loose = false, format = 'uint16') {
+  if (typeof str !== 'string') throw new TypeError('Input is not a string')
+  if (format !== 'uint16' && format !== 'uint8-le' && format !== 'uint8-be') {
+    throw new TypeError('Unknown format')
+  }
+
+  if (!isWellFormed.call(str)) {
+    if (!loose) throw new TypeError(E_STRICT_UNICODE)
+    str = nativeDecoder.decode(Buffer.from(str)) // well, let's fix up (Buffer doesn't do this with utf16 encoding)
+  }
+
+  const ble = Buffer.from(str, 'utf-16le')
+
+  if (format === 'uint8-le') return to8(ble)
+  if (format === 'uint8-be') return to8(ble.swap16())
+  if (format === 'uint16') {
+    const b = ble.byteOffset % 2 === 0 ? ble : Buffer.from(ble) // it should be already aligned, but just in case
+    if (!isLE) b.swap16()
+    return new Uint16Array(b.buffer, b.byteOffset, b.byteLength / 2)
+  }
+
+  throw new Error('Unreachable')
+}
+
+const swapped = (x, swap) =>
+  swap ? Buffer.from(x).swap16() : Buffer.from(x.buffer, x.byteOffset, x.byteLength)
+
+// We skip TextDecoder on Node.js, as it's is somewhy significantly slower than Buffer for utf16
+function decodeNode(input, loose = false, format = 'uint16') {
+  let ble
+  if (format === 'uint16') {
+    if (!(input instanceof Uint16Array)) throw new TypeError('Expected an Uint16Array')
+    ble = swapped(input, !isLE)
+  } else if (format === 'uint8-le' || format === 'uint8-be') {
+    if (!(input instanceof Uint8Array)) throw new TypeError('Expected an Uint8Array')
+    if (input.byteLength % 2 !== 0) throw new TypeError('Expected even number of bytes')
+    ble = swapped(input, format === 'uint8-be')
+  } else {
+    throw new TypeError('Unknown format')
+  }
+
+  const str = ble.ucs2Slice(0, ble.byteLength)
+  if (isWellFormed.call(str)) return str
+  if (!loose) throw new TypeError(E_STRICT)
+  return nativeDecoder.decode(Buffer.from(str)) // fixup (see above)
+}
+
+function decodeDecoder(input, loose = false, format = 'uint16') {
+  let encoding
+  if (format === 'uint16') {
+    if (!(input instanceof Uint16Array)) throw new TypeError('Expected an Uint16Array')
+    encoding = isLE ? 'utf-16le' : 'utf-16be'
+  } else if (format === 'uint8-le' || format === 'uint8-be') {
+    if (!(input instanceof Uint8Array)) throw new TypeError('Expected an Uint8Array')
+    if (input.byteLength % 2 !== 0) throw new TypeError('Expected even number of bytes')
+    encoding = format === 'uint8-le' ? 'utf-16le' : 'utf-16be'
+  } else {
+    throw new TypeError('Unknown format')
+  }
+
+  return new TextDecoder(encoding, { ignoreBOM: true, fatal: !loose }).decode(input) // TODO: cache decoder?
+}
+
+const decode = isDeno ? decodeDecoder : decodeNode
+
+export const utf16fromString = (str, format = 'uint16') => encode(str, false, format)
+export const utf16fromStringLoose = (str, format = 'uint16') => encode(str, true, format)
+export const utf16toString = (arr, format = 'uint16') => decode(arr, false, format)
+export const utf16toStringLoose = (arr, format = 'uint16') => decode(arr, true, format)

package/utf8.js

@@ -1,22 +1,21 @@
 import { assertUint8 } from './assert.js'
 import { typedView } from './array.js'
-import * as js from './fallback/utf8.js'
+import { isHermes, nativeDecoder, nativeEncoder } from './fallback/_utils.js'
 import { asciiPrefix, decodeLatin1 } from './fallback/latin1.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
+const { TextDecoder, decodeURIComponent, escape } = globalThis // Buffer is optional
 // 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 decoderLoose = nativeDecoder
+const decoderFatal = nativeDecoder
+  ? new TextDecoder('utf-8', { ignoreBOM: true, fatal: 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
+const shouldUseEscapePath = isHermes // faster only on Hermes, js path beats it on normal engines
 
 function deLoose(str, loose, res) {
   if (loose || str.length === res.length) return res // length is equal only for ascii, which is automatically fine
@@ -46,16 +45,16 @@ function deLoose(str, loose, 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
+  if (str.length === 0) return new Uint8Array() // faster than Uint8Array.of
+  if (nativeEncoder) return deLoose(str, loose, nativeEncoder.encode(str))
   // 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)
 }
 
 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
+  if (arr.byteLength === 0) return ''
+  if (nativeDecoder) return loose ? decoderLoose.decode(arr) : decoderFatal.decode(arr) // Node.js and browsers
 
   // Fast path for ASCII prefix, this is faster than all alternatives below
   const prefix = decodeLatin1(arr, 0, asciiPrefix(arr))

package/utf8.node.js

@@ -5,23 +5,44 @@ import { isAscii } from 'node:buffer'
 
 if (Buffer.TYPED_ARRAY_SUPPORT) throw new Error('Unexpected Buffer polyfill')
 
-const decoderFatal = new TextDecoder('utf8', { ignoreBOM: true, fatal: true })
-const decoderLoose = new TextDecoder('utf8', { ignoreBOM: true })
+const decoderFatal = new TextDecoder('utf-8', { ignoreBOM: true, fatal: true })
+const decoderLoose = new TextDecoder('utf-8', { ignoreBOM: true })
 const { isWellFormed } = String.prototype
+const isDeno = Boolean(globalThis.Deno)
 
 function encode(str, loose = false) {
   if (typeof str !== 'string') throw new TypeError('Input is not a string')
-  const res = Buffer.from(str)
-  if (loose || str.length === res.length || isWellFormed.call(str)) return res // length is equal only for ascii, which is automatically fine
-  throw new TypeError(E_STRICT_UNICODE)
+  const strLength = str.length
+  if (strLength === 0) return new Uint8Array() // faster than Uint8Array.of
+  let res
+  if (strLength > 0x4_00 && !isDeno) {
+    // Faster for large strings
+    const byteLength = Buffer.byteLength(str)
+    res = Buffer.allocUnsafe(byteLength)
+    const ascii = byteLength === strLength
+    const written = ascii ? res.latin1Write(str) : res.utf8Write(str)
+    if (written !== byteLength) throw new Error('Failed to write all bytes') // safeguard just in case
+    if (ascii || loose) return res // no further checks needed
+  } else {
+    res = Buffer.from(str)
+    if (res.length === strLength || loose) return res
+  }
+
+  if (!isWellFormed.call(str)) throw new TypeError(E_STRICT_UNICODE)
+  return res
 }
 
 function decode(arr, loose = false) {
   assertUint8(arr)
-  if (isAscii(arr)) {
+  const byteLength = arr.byteLength
+  if (byteLength === 0) return ''
+  if (byteLength > 0x6_00 && !(isDeno && loose) && isAscii(arr)) {
     // On non-ascii strings, this loses ~10% * [relative position of the first non-ascii byte] (up to 10% total)
     // On ascii strings, this wins 1.5x on loose = false and 1.3x on loose = true
-    return Buffer.from(arr.buffer, arr.byteOffset, arr.byteLength).latin1Slice(0, arr.byteLength) // .latin1Slice is faster than .asciiSlice
+    // Only makes sense for large enough strings
+    const buf = Buffer.from(arr.buffer, arr.byteOffset, arr.byteLength)
+    if (isDeno) return buf.toString() // Deno suffers from .latin1Slice
+    return buf.latin1Slice(0, arr.byteLength) // .latin1Slice is faster than .asciiSlice
   }
 
   return loose ? decoderLoose.decode(arr) : decoderFatal.decode(arr)

package/wif.js

@@ -0,0 +1,42 @@
+import { toBase58checkSync, fromBase58checkSync } from './base58check.js'
+import { assertUint8 } from './assert.js'
+
+// Mostly matches npmjs.com/wif, but with extra checks + using our base58check
+// Also no inconsistent behavior on Buffer/Uint8Array input
+
+function from(arr, expectedVersion) {
+  assertUint8(arr)
+  const version = arr[0]
+  if (expectedVersion !== undefined && version !== expectedVersion) {
+    throw new Error('Invalid network version')
+  }
+
+  // Makes a copy, regardless of input being a Buffer or a Uint8Array (unlike .slice)
+  const privateKey = Uint8Array.from(arr.subarray(1, 33))
+  if (arr.length === 33) return { version, privateKey, compressed: false }
+  if (arr.length !== 34) throw new Error('Invalid WIF length')
+  if (arr[33] !== 1) throw new Error('Invalid compression flag')
+  return { version, privateKey, compressed: true }
+}
+
+function to({ version: v, privateKey, compressed }) {
+  if (!Number.isSafeInteger(v) || v < 0 || v > 0xff) throw new Error('Missing or invalid version')
+  assertUint8(privateKey, { length: 32, name: 'privateKey' })
+  if (privateKey.length !== 32) throw new TypeError('Invalid privateKey length')
+  const out = new Uint8Array(compressed ? 34 : 33)
+  out[0] = v
+  out.set(privateKey, 1)
+  if (compressed) out[33] = 1
+  return out
+}
+
+// Async performance is worse here, so expose the same internal methods as sync for now
+// ./base58check is sync internally anyway for now, so doesn't matter until that is changed
+
+export const fromWifStringSync = (string, version) => from(fromBase58checkSync(string), version)
+// export const fromWifString = async (string, version) => from(await fromBase58check(string), version)
+export const fromWifString = async (string, version) => from(fromBase58checkSync(string), version)
+
+export const toWifStringSync = (wif) => toBase58checkSync(to(wif))
+// export const toWifString = async (wif) => toBase58check(to(wif))
+export const toWifString = async (wif) => toBase58checkSync(to(wif))