• Actually publish the contents of 4.0.0-alpha+3.

New features:

• Added <Int>.maxInt and <Int>.minInt as static (int) constants.

Breaking changes:

• <FixedInt>, which was pointless, is now just int. That means that any fixed integer representation can be provided somewhere an int is expected, which cuts down on boilerplate without much value.

New features:

• Added BitList, a compact List<bool> implementation that stores every element as a single bit, with implementations that are fixed-size and growable.

• Added <FixedInt>.zero and <FixedInt>.one as static constants.

• Added collectBytes(), a utility to convert a Stream<List<int>> into a Uint8List.

Breaking changes:

• Replaced <FixedInt>.bits with <FixedInt>.toBitList():

  - final bits = Int8(0).bits;
  + final bits = Int8(0).toBitList();
  

New features:

Lots and lots. It will be easier to just read the API documentation.

Breaking changes:

Basically everything. The entire API has been restructured to use extension types, and some APIs removed entirely that were either not well-thought out (oops) or were unnecessary:

• Integral, which was a base type for defining integers, has been removed in favor of a helper class, IntDescriptor, which is used to define new integer types, and acts sort of a meta type or poor man's macro for defining features:

  - class Int4 extends Integral<Int4> {
  -   Int4(int value) : super.checked(value, signed: true, size: 4);
  -
  -   @override
  -   Int4 wrapSafeValue(int value) => Int4(value);
  - }
  
  + extension type const Int4._(int _) implements FixedInt {
  +   static const _descriptor = _IntDescriptor<Int8>.signed(
  +     Int4.fromUnchecked,
  +     width: 4,
  +     max: 7,
  +   );
  +
  +   factory Int4(int v) => _descriptor.fit(v);
  +
  +   // ...
  + }
  

  In practice, it is much more difficult to implement a custom type, as many methods have to be hand-written, but it is also a much better future-proof approach. In the near-term, it's possible we could expose the code generator used by this package internally as a tool for others to use, and longer-term Dart macros can be used to simplify this process for users.

• Fixed-size integers still exist, but with an updated API. Replacements are a follows:

  • .bitChunk(l, r) -> .chunk(l, [s?])
  • .bitRange(l, r) -> .slice(l, [r?])
  • .bitsSet -> .countOnes()
  • .clearBit(n) -> .setNthBit(n, false)
  • .getBit(n) -> .nthBit(n) or operator [n]
  • .replaceBitRange(l, r, b) -> .replace(l, r?, b)
  • .rotateLeftShift(n) -> .rotateLeft(n)
  • .rotateRightShift(n) -> .rotateRight(n)
  • .setBit(n) -> .setNthBit(n)
  • .signExtend(n) -> removed.
  • .size has been removed in favor of a static .width; as extension types are non-virtual.
  • .toggleBit(n) -> .toggleNthBit(n)
  • .value -> .toInt()

• The extension methods BinaryInt were removed. Instead, use the extension types directly. A small subset of helper methods are available on IntExtension but have little in common with the previous API (mostly convenience methods).

• Every other extension method set was removed.

• Downgrade meta version to v1.7.0 for Flutter compatibility.

Wooho: 🎉 Added support for Dart SDK 2.17.0 and Null Safety 🎉

Changes are relatively minor, some code that threw an ArgumentError on a null value now fails statically, but otherwise there are no behavioral changes to well-behaving code.

Special thanks to https://github.com/leynier for driving this update!

Highlights:

Added limited support for operations on integer values that exceed 32-bits. Before 2.0.0 most of the methods provided by this package had undefined behavior when accessing a bit > the 31st bit when compiled to JavaScript:

• msb
• getBit
• setBit and isSet
• clearBit and isCleared
• toggleBit
• countSetBits
• bitRange and bitChunk
• hiLo
• toBinary and toBinaryPadded

In addition, these operations will now throw UnsupportedError when compiled to JavaScript when attempting operations on integer values that exceed 52-bits, which is the maximum integer that is supported in JavaScript VMs. The remaining methods (i.e. signedRightShift, so on), unless otherwise documented are assumed to have undefined behavior when compiled to JavaScript.

Tip: Don't want to consider all of that? You can always just use Uint32!

Operating on Larger Ints:

Added <int>.hiLo(), which returns a fixed 2-length Uint32List where element 0 is the "hi" (upper bits) and element 1 is the "lo" (lower bits). Because of platform limitations, "hi" may only include up to the 52nd bit.

Additionally, Uint32List has received new extension methods through the extension class BinaryUint64HiLo (e.g. the result of <int>.hiLo()), which simplifies operations for integers larger than 32 bits:

• .hi and .lo getters.
• ~ and & and | operators.
• equals(Uint32List) and toInt() methods.

These methods allow treating a Uint32List roughly as a 64-bit integer with limited boxing. We may still consider adding a Uint64 boxed Integral in a future release, but you may also try using these other implemntations:

• BigInt
• Int64

Additional Changes:

• Added <int>.pow(n), which is like <dart:math>.pow with a return of int.
• Added <Integral>.signExtend(startSize).
• Removed all methods deprecated up to this point.
• Removed <Integral>.[un]signed, which was misleading for unsigned integers.
• Fixed a bug where <int>.replaceBitRange often emitted an incorrect result.
• Fixed a bug where <int>.signExtend often emitted an incorrect result.

• Deprecated <*>.shiftRight in favor of <*>.signedShiftRight>.
• Deprecated <*>.rotateRight, which was not correctly implemeted.
• Added <*>.rotateRightShift to replace rotateRight.
• Updated some doc comments that referred to incorrect JavaScript operators.

• Added <BinaryInt|BinaryList|Integral>.toggleBit.
• Deprecated Integral.setBits in favor of .bitsSet.

• Added BitPatternGroup's constructor, deprecating .toGroup().
• Added BitPart.zero and BitPart.one and deprecated BitPart(int).

• Added List<int>.toBits() as a replacement for List<int>.parseBits().
• Added String.bits as a replacement for String.parseBits().
• Deprecated List<int>.parseBits() and String.parseBits().
• Deprecated int.as[U]Int{N} functions in favor of manual wrapping.

• Added comparison operators (>, >=, <, <=) to Integral.
• Added <Integral>.checkRange and <Integral>.assertRange static methods.
• Added the abiltiy to extend Integral to create custom-sized integers.

• Fixed a bug where _InterpretedBitPattern (the BitPattern generated from BitPatternBuilder) was sorted in an incorrect order (ascending instead of descending), which would not match correctly in some scenarios.

• Fixed a bug where BitPatternBuilder.parse('00AA_AABB') incorrectly threw a FormatException assuming that _ deliniated the end of the A variable segment and the subsequent A was a new segment (which is invalid). It now correctly parses the above as just two variable segments (AAAA, BB).

• Added BitPatternBuilder.parse, a simplified-format for building BitPattern from a string of alpha-numeric characters, where 0 and 1 are pre-defined (static) flags for matching, and charaters are variable segments:

  // Create a BitPattern (Data Structures).
  final $01V = BitPatternBuilder([
    BitPart(0),
    BitPart(1),
    BitPart.v(1, 'A'),
  ]).build();
  
  // Create a BitPattern (Parse a String).
  final $01Vs = BitPatternBuilder.parse('01A').build();
  print($01V == $01Vs); // true
  

• Fixed a bug where it was not possible to capture variables that were >8-bit.

• Added BitPatternBuilder, BitPattern, BitPart: a new API in order to build bit-based patterns and match against arbitrary sets of bits, optionally extracting variable names. This API is intended to make it easier to build apps and packages around implementing emulators and other decoders.

A large update to bring into line for Dart 2, as well take advantage of newer langauge features like extension methods over top-level methods. As a result, the new API is not compatible with previous versions, but migration should be trivial.

• Added arithmeticShiftRight

• Moved into a standalone repository (outside of gba.dart).
• Added signExtend as a method to Integral.
• Added areSet.
• Added msb.

• Added signExtend

• Fixed a bug where int128 and uint128 only had a length of 64.

• Updated the documentation and README.

• Added isZero.

• Added isNegative, hasCarryBit, doesAddOverflow, doesSubOverflow, mask.
• Added parseBits.

• Add top-level isSet and isClear, Integral#isSet, Integral#isClear.
• Add checked-mode range checks to bitChunk and bitRange.
• Fix a bug in the implementation of bitChunk and bitRange.
• Added a top-level fromBits and Integral#fromBits

• Initial commit, feedback welcome!