radix_plus 1.0.6

in place Radix Sort implementation for Dart and Flutter, providing significant speed improvements over standard sorting for number lists.

Changelog

1.0.6

• fix(core): Resolve a critical bug in radixSortCore64 where maxValue tracking on negative signed integers (representing values >= 2^63 unsigned) led to underestimating actualPasses. The algorithm now correctly detects when any MSB is set, ensuring all 8 passes are executed. This fixes sorting behavior for lists containing both large negative and positive doubles.
• perf(core): Optimize radixSortCore64 pass determination scan by breaking early as soon as any value with its MSB set is encountered, avoiding $O(n)$ scanning overhead.
• perf(parallel): Optimize the parallel sorting k-way merge by representing the min-heap using parallel flat arrays (heapValues and heapChunkIndices) instead of instantiating millions of _HeapNode objects. This completely eliminates object allocation overhead during the merge loop and reduces GC pressure.
• Optimized Parallel: k-way merge, reducing parallel sorting time by ~14–15% and significantly reducing allocation/GC overhead.
• test(core): Add unit tests to prevent regression in float sorting with large positive and negative numbers.

1.0.5

• build(sdk): Relax Dart SDK version constraint

1.0.4

• perf(parallel): Overhaul parallel sorting for improved performance and memory usage.
  • Isolates are now spawned in parallel using Future.wait to reduce startup overhead.
  • radixSortParallelSigned now avoids intermediate list allocations, significantly reducing GC pressure.
  • Created a core _radixSortParallelUint32 engine to streamline parallel logic.
• perf(bigint): Optimize radixSortBigInt memory management.
  • The auxiliary buffer is now allocated only once, reducing GC pressure during sorting.
  • Refactored and unified internal BigInt sorting functions to reduce code duplication.
• perf(utils): Enhance buffer pool logic.
  • The get method in buffer pools now finds the best-fit buffer (>= requested size) to increase hit rate.
• style(core): Minor code cleanup using fillRange for array initialization.

1.0.3

• build(sdk): bump SDK requirement to 3.10.1
• change(package): Rename project from Radix Pulse to Radix Plus Update package name, version to 1.0.3, and repository URLs. Adjust all internal and external references. Remove outdated API documentation.

1.0.2

• fix(core): Resolve linting issues and refine parallel export logic
• Update benchmark documentation in README and results.md with improved methodology and new performance data.
• Add publishing metadata and topics to pubspec.yaml. Update CHANGELOG for version 1.0.1.
  • fix(core): Resolve linting issues and refine parallel export logic
  • Update benchmark documentation in README and results.md with improved methodology and new performance data.
  • Add publishing metadata and topics to pubspec.yaml. Update CHANGELOG for version 1.0.2.

1.0.1

feat(core): Introduce BigInt sorting, parallel enhancements, and perf optimizations

• Add comprehensive BigInt sorting support, including radixSortBigInt and radixSortBigIntWithRange.
• Enhance parallel sorting with radixSortParallelSigned, adaptive thread management, and k-way merge optimization.
• Introduce zero-copy sorting APIs for typed lists: radixSortFloat64, radixSortInt32, and radixSortUint32.
• Provide advanced float handling with radixSortFloat64WithNaN for NaN placement.
• Implement adaptive core algorithms that fallback to insertion sort for small lists.
• Optimize all core radix sort algorithms (int, double, BigInt) with skip passes, efficient data transformations, and improved buffer reuse for performance.
• Overhaul README.md with new sections, detailed performance benchmarks, and expanded usage examples.
• Add dedicated benchmark/results.md and update CONTRIBUTING.md links.
• Implement web-platform stub for parallel sorting functions.
• Update .gitignore to ignore example build files and add pubspec.yaml metadata.

1.0.0

• Initial release.