Test source generators, analyzers and code fixes in-process, so a coverage collector actually attributes the generator's own code. Loads the component from the test host's output directory rather than through Roslyn's analyzer path, where instrumentation never reaches it.
The .NET measurement command for Ratchet's test suite measurement: it runs a solution's tests and writes a measurement in Ratchet's measurement contract.
# 2026-11-15 - V6 - ExcelFAST - ⛓️💥 **Breaking Change(s)** - If you want to use `Async` then create a `new Excel_PRIMEAsync()` - Same applies to `Xlsb` - Use `ReadOnlySpan<Cell>` instead of `ArraySegment<Cell>` - Using foreach over ReadOnlySpan<T> is guaranteed to be allocation-free and boxing-free. If you need LINQ-style processing on a span without allocations, iterate using for or foreach directly over the span - ✅ **ExcelFAST High-Throughput Subsystem**: - ✅ Dedicated zero-allocation sequential reader pipeline (`IExcelFast`, `ISheetFast`, `IRowFast`). - ✅ Format-specific implementations: `ExcelFastXlsx` for OpenXML (.xlsx) and `ExcelFastXlsb` for binary OpenXML (.xlsb). - ✅ Array pooling integration (`ReadOnlySpan<Cell>`) with `PooledCellView` and thread-static buffers for zero GC allocation during hot iteration loops. - ✅ Full support across .NET 8.0, .NET 9.0, and .NET 10.0. - ✅ Integrated test suite and benchmarking coverage in `ExcelFAST.Tests`. - ✅ Add .Net11 TFM support across all four target frameworks (`net8.0`, `net9.0`, `net10.0`, `net11.0`) - Replaced the previous float round-trip memory comparison in CellValue.IsExactlyDouble with a fast memory-overlay check on the underlying 64-bit decimal mantissa (lo64 <= 999_999_999_999_999UL and hi == 0). - ✅ A separate nuget for ExcelFAST - ✅ Performance Optimizations for Prime (See breaking changes above) # 2026-10-18 - V5 nuget update - Update nugets - Made benchmarks run .net**9** and .net**10** as well - **Memory & Allocation Optimizations** - Some tinkering in the way memory is used and allocated, especially in the Xlsb workflow - Eliminated String Caching Overhead: Removed the ConcurrentDictionary cache for small strings in CellValue, which introduced locking and hashing overhead during high-volume parsing. - Simplified SIMD Decoding: Reverted SimdStringDecoder.cs to use Vector<ushort> instead of explicit Vector512/Vector256/Vector128 branching, reducing overhead for common short strings. - Optimized Column Parsing: Replaced SearchValues with simple loops in ExcelColumns.cs for cell reference parsing, as the overhead of SearchValues was counterproductive for 1-3 character strings. - Tuned Method Inlining: Removed AggressiveOptimization from several complex methods to allow better JIT tiered compilation and reduce instruction cache pressure.