- Use curl with -fSL flags for fail-fast behavior and proper redirects
- Check curl exit code immediately after download
- Replace piped echo with temporary file for sha256sum verification
- Use sha256sum -c --status for cleaner verification
- Add detailed error logging with expected checksum
- Clean up temp file and binary on any failure
- Exit 1 immediately on download or verification failure
- Updated .coderabbit.yaml to exclude additional file types from reviews, improving the focus on relevant code changes.
- Modified scanner.sh to handle test failures more gracefully, ensuring that analysis stops on test failures and improving logging.
- Improved sonarscanner.sh to ensure build and test failures are properly reported, enhancing CI reliability.
- Refined SimdExtensions.cs documentation for clarity on variance calculation methods.
- Cleaned up TSeries.Tests.cs by simplifying the test structure and ensuring proper namespace usage.
- Fixed potential issues in tseries.cs by ensuring correct handling of DateTime values.
- Enhanced CsvFeed.cs to improve error handling during CSV parsing, ensuring robustness against malformed data.
- Updated GBM.cs to correctly calculate volume in the current bar, ensuring accurate simulation.
- Adjusted index.html to use globalThis for better compatibility across environments.
- Refined quantalib.csproj to exclude unnecessary files from compilation, streamlining the build process.
- Added comprehensive tests for the Mama class to ensure correct behavior during updates and state management.
- Improved error handling in various trend classes (Kama, Dema, Ema, T3, Tema, Wma) to ensure NaN values are managed correctly.
- Removed redundant Mama.Repro.Tests.cs file and consolidated tests into Mama.Tests.cs for better organization.
- Enhanced T3 and Tema classes to maintain state integrity during updates, particularly with NaN values.
- Introduced `TrimaVector` class for multi-period Triangular Moving Average (TRIMA) calculations, optimized for SIMD.
- Implemented last-value substitution for invalid inputs in TRIMA.
- Added methods for calculating TRIMA for entire series and individual updates.
- Enhanced `Wma` class with periodic resync to prevent floating-point drift and introduced SIMD optimizations for performance.
- Updated benchmark suite to include TRIMA calculations alongside existing SMA, EMA, and WMA benchmarks.
Miha Kralj