Unaligned bit arrays on the JavaScript target 791405

Author: richard-viney

.github/workflows/ci.yml

@@ -44,7 +44,8 @@ jobs:
       - uses: erlef/setup-beam@v1
         with:
           otp-version: "27.0"
-          gleam-version: "1.6.0"
+          gleam-version: nightly
+          version-type: strict
       - uses: actions/setup-node@v4
         with:
           node-version: ${{ matrix.node_version }}
@@ -62,7 +63,8 @@ jobs:
       - uses: erlef/setup-beam@v1
         with:
           otp-version: "27.0"
-          gleam-version: "1.6.0"
+          gleam-version: nightly
+          version-type: strict
       - uses: oven-sh/setup-bun@v2
         with:
           bun-version: ${{ matrix.bun_version }}
@@ -80,7 +82,8 @@ jobs:
       - uses: erlef/setup-beam@v1
         with:
           otp-version: "27.0"
-          gleam-version: "1.6.0"
+          gleam-version: nightly
+          version-type: strict
       - uses: denoland/setup-deno@v1
         with:
           deno-version: ${{ matrix.deno_version }}

CHANGELOG.md

@@ -2,12 +2,11 @@
 
 ## Unreleased
 
-- The performance of `dict.is_empty` has been improved.
-
-## v0.54.0 - 2025-02-04
-
 - The `uri` module gains the `empty` value, representing an empty URI which
   equivalent to `""`.
+- The performance of `dict.is_empty` has been improved.
+- Unaligned bit arrays on the JavaScript target are now supported by the
+  functions in the `bit_array` module.
 
 ## v0.54.0 - 2025-02-04
 

src/gleam/bit_array.gleam

@@ -13,22 +13,20 @@ pub fn from_string(x: String) -> BitArray
 /// Returns an integer which is the number of bits in the bit array.
 ///
 @external(erlang, "erlang", "bit_size")
-pub fn bit_size(x: BitArray) -> Int {
-  byte_size(x) * 8
-}
+@external(javascript, "../gleam_stdlib.mjs", "bit_array_bit_size")
+pub fn bit_size(x: BitArray) -> Int
 
 /// Returns an integer which is the number of bytes in the bit array.
 ///
 @external(erlang, "erlang", "byte_size")
-@external(javascript, "../gleam_stdlib.mjs", "length")
+@external(javascript, "../gleam_stdlib.mjs", "bit_array_byte_size")
 pub fn byte_size(x: BitArray) -> Int
 
 /// Pads a bit array with zeros so that it is a whole number of bytes.
 ///
 @external(erlang, "gleam_stdlib", "bit_array_pad_to_bytes")
-pub fn pad_to_bytes(x: BitArray) -> BitArray {
-  x
-}
+@external(javascript, "../gleam_stdlib.mjs", "bit_array_pad_to_bytes")
+pub fn pad_to_bytes(x: BitArray) -> BitArray
 
 /// Creates a new bit array by joining two bit arrays.
 ///
@@ -228,7 +226,6 @@ fn inspect_loop(input: BitArray, accumulator: String) -> String {
 /// // -> Eq
 /// ```
 ///
-@external(javascript, "../gleam_stdlib.mjs", "bit_array_compare")
 pub fn compare(a: BitArray, with b: BitArray) -> order.Order {
   case a, b {
     <<first_byte, first_rest:bits>>, <<second_byte, second_rest:bits>> ->
@@ -257,6 +254,7 @@ pub fn compare(a: BitArray, with b: BitArray) -> order.Order {
 }
 
 @external(erlang, "gleam_stdlib", "bit_array_to_int_and_size")
+@external(javascript, "../gleam_stdlib.mjs", "bit_array_to_int_and_size")
 fn bit_array_to_int_and_size(a: BitArray) -> #(Int, Int)
 
 /// Checks whether the first `BitArray` starts with the second one.

src/gleam_stdlib.mjs

@@ -7,6 +7,7 @@ import {
   UtfCodepoint,
   stringBits,
   toBitArray,
+  bitArraySlice,
   NonEmpty,
   CustomType,
 } from "./gleam.mjs";
@@ -316,8 +317,49 @@ export function bit_array_from_string(string) {
   return toBitArray([stringBits(string)]);
 }
 
+export function bit_array_bit_size(bit_array) {
+  return bit_array.bitSize;
+}
+
+export function bit_array_byte_size(bit_array) {
+  return bit_array.byteSize;
+}
+
+export function bit_array_pad_to_bytes(bit_array) {
+  const trailingBitsCount = bit_array.bitSize % 8;
+
+  // If the bit array is a whole number of bytes  it can be returned unchanged
+  if (trailingBitsCount === 0) {
+    return bit_array;
+  }
+
+  const finalByte = bit_array.byteAt(bit_array.byteSize - 1);
+
+  const unusedBitsCount = 8 - trailingBitsCount;
+  const correctFinalByte = (finalByte >> unusedBitsCount) << unusedBitsCount;
+
+  // If the unused bits in the final byte are already set to zero then the
+  // existing buffer can be re-used, avoiding a copy
+  if (finalByte === correctFinalByte) {
+    return new BitArray(
+      bit_array.rawBuffer,
+      bit_array.byteSize * 8,
+      bit_array.bitOffset,
+    );
+  }
+
+  // Copy the bit array into a new aligned buffer and set the correct final byte
+  const buffer = new Uint8Array(bit_array.byteSize);
+  for (let i = 0; i < buffer.length - 1; i++) {
+    buffer[i] = bit_array.byteAt(i);
+  }
+  buffer[buffer.length - 1] = correctFinalByte;
+
+  return new BitArray(buffer);
+}
+
 export function bit_array_concat(bit_arrays) {
-  return toBitArray(bit_arrays.toArray().map((b) => b.buffer));
+  return toBitArray(bit_arrays.toArray());
 }
 
 export function console_log(term) {
@@ -333,9 +375,25 @@ export function crash(message) {
 }
 
 export function bit_array_to_string(bit_array) {
+  // If the bit array isn't a whole number of bytes then return an error
+  if (bit_array.bitSize % 8 !== 0) {
+    return new Error(Nil);
+  }
+
   try {
     const decoder = new TextDecoder("utf-8", { fatal: true });
-    return new Ok(decoder.decode(bit_array.buffer));
+
+    if (bit_array.bitOffset === 0) {
+      return new Ok(decoder.decode(bit_array.rawBuffer));
+    } else {
+      // The input data isn't aligned, so copy it into a new aligned buffer so
+      // that TextDecoder can be used
+      const buffer = new Uint8Array(bit_array.byteSize);
+      for (let i = 0; i < buffer.length; i++) {
+        buffer[i] = bit_array.byteAt(i);
+      }
+      return new Ok(decoder.decode(buffer));
+    }
   } catch {
     return new Error(Nil);
   }
@@ -413,16 +471,14 @@ export function random_uniform() {
 }
 
 export function bit_array_slice(bits, position, length) {
-  const start = Math.min(position, position + length);
-  const end = Math.max(position, position + length);
-  if (start < 0 || end > bits.length) return new Error(Nil);
-  const byteOffset = bits.buffer.byteOffset + start;
-  const buffer = new Uint8Array(
-    bits.buffer.buffer,
-    byteOffset,
-    Math.abs(length),
-  );
-  return new Ok(new BitArray(buffer));
+  let start = Math.min(position, position + length);
+  let end = Math.max(position, position + length);
+
+  if (start < 0 || end * 8 > bits.bitSize) {
+    return new Error(Nil);
+  }
+
+  return new Ok(bitArraySlice(bits, start * 8, end * 8));
 }
 
 export function codepoint(int) {
@@ -522,16 +578,20 @@ let b64TextDecoder;
 export function encode64(bit_array, padding) {
   b64TextDecoder ??= new TextDecoder();
 
-  const bytes = bit_array.buffer;
+  bit_array = bit_array_pad_to_bytes(bit_array);
 
-  const m = bytes.length;
+  const m = bit_array.byteSize;
   const k = m % 3;
   const n = Math.floor(m / 3) * 4 + (k && k + 1);
   const N = Math.ceil(m / 3) * 4;
   const encoded = new Uint8Array(N);
 
   for (let i = 0, j = 0; j < m; i += 4, j += 3) {
-    const y = (bytes[j] << 16) + (bytes[j + 1] << 8) + (bytes[j + 2] | 0);
+    const y =
+      (bit_array.byteAt(j) << 16) +
+      (bit_array.byteAt(j + 1) << 8) +
+      (bit_array.byteAt(j + 2) | 0);
+
     encoded[i] = b64EncodeLookup[y >> 18];
     encoded[i + 1] = b64EncodeLookup[(y >> 12) & 0x3f];
     encoded[i + 2] = b64EncodeLookup[(y >> 6) & 0x3f];
@@ -804,7 +864,7 @@ export function inspect(v) {
   if (Array.isArray(v)) return `#(${v.map(inspect).join(", ")})`;
   if (v instanceof List) return inspectList(v);
   if (v instanceof UtfCodepoint) return inspectUtfCodepoint(v);
-  if (v instanceof BitArray) return inspectBitArray(v);
+  if (v instanceof BitArray) return `<<${bit_array_inspect(v, "")}>>`;
   if (v instanceof CustomType) return inspectCustomType(v);
   if (v instanceof Dict) return inspectDict(v);
   if (v instanceof Set) return `//js(Set(${[...v].map(inspect).join(", ")}))`;
@@ -895,19 +955,26 @@ export function inspectList(list) {
   return `[${list.toArray().map(inspect).join(", ")}]`;
 }
 
-export function inspectBitArray(bits) {
-  return `<<${Array.from(bits.buffer).join(", ")}>>`;
-}
-
 export function inspectUtfCodepoint(codepoint) {
   return `//utfcodepoint(${String.fromCodePoint(codepoint.value)})`;
 }
 
 export function base16_encode(bit_array) {
+  const trailingBitsCount = bit_array.bitSize % 8;
+
   let result = "";
-  for (const byte of bit_array.buffer) {
+
+  for (let i = 0; i < bit_array.byteSize; i++) {
+    let byte = bit_array.byteAt(i);
+
+    if (i === bit_array.byteSize - 1 && trailingBitsCount !== 0) {
+      const unusedBitsCount = 8 - trailingBitsCount;
+      byte = (byte >> unusedBitsCount) << unusedBitsCount;
+    }
+
     result += byte.toString(16).padStart(2, "0").toUpperCase();
   }
+
   return result;
 }
 
@@ -923,38 +990,53 @@ export function base16_decode(string) {
 }
 
 export function bit_array_inspect(bits, acc) {
-  return `${acc}${[...bits.buffer].join(", ")}`;
-}
+  if (bits.bitSize === 0) {
+    return acc;
+  }
 
-export function bit_array_compare(first, second) {
-  for (let i = 0; i < first.length; i++) {
-    if (i >= second.length) {
-      return new Gt(); // first has more items
-    }
-    const f = first.buffer[i];
-    const s = second.buffer[i];
-    if (f > s) {
-      return new Gt();
-    }
-    if (f < s) {
-      return new Lt();
-    }
+  for (let i = 0; i < bits.byteSize - 1; i++) {
+    acc += bits.byteAt(i).toString();
+    acc += ", ";
   }
-  // This means that either first did not have any items
-  // or all items in first were equal to second.
-  if (first.length === second.length) {
-    return new Eq();
+
+  if (bits.byteSize * 8 === bits.bitSize) {
+    acc += bits.byteAt(bits.byteSize - 1).toString();
+  } else {
+    const trailingBitsCount = bits.bitSize % 8;
+    acc += bits.byteAt(bits.byteSize - 1) >> (8 - trailingBitsCount);
+    acc += `:size(${trailingBitsCount})`;
   }
-  return new Lt(); // second has more items
+
+  return acc;
+}
+
+export function bit_array_to_int_and_size(bits) {
+  const trailingBitsCount = bits.bitSize % 8;
+  const unusedBitsCount = trailingBitsCount === 0 ? 0 : 8 - trailingBitsCount;
+
+  return [bits.byteAt(0) >> unusedBitsCount, bits.bitSize];
 }
 
 export function bit_array_starts_with(bits, prefix) {
-  if (prefix.length > bits.length) {
+  if (prefix.bitSize > bits.bitSize) {
     return false;
   }
 
-  for (let i = 0; i < prefix.length; i++) {
-    if (bits.buffer[i] !== prefix.buffer[i]) {
+  // Check any whole bytes
+  const byteCount = Math.trunc(prefix.bitSize / 8);
+  for (let i = 0; i < byteCount; i++) {
+    if (bits.byteAt(i) !== prefix.byteAt(i)) {
+      return false;
+    }
+  }
+
+  // Check any trailing bits at the end of the prefix
+  if (prefix.bitSize % 8 !== 0) {
+    const unusedBitsCount = 8 - (prefix.bitSize % 8);
+    if (
+      bits.byteAt(byteCount) >> unusedBitsCount !==
+      prefix.byteAt(byteCount) >> unusedBitsCount
+    ) {
       return false;
     }
   }