spectral-library 0.3.0

spectral-library 0.3.0

Released 2026-04-04.

Highlights

• Rust-accelerated batch mapping via a PyO3/Rayon native extension for
  spectrum reconstruction, target-sensor projection, segment merging, and
  neighbor refinement
• LinearSpectralMapper for fixed input @ weights + bias high-throughput
  mapping without per-sample KNN retrieval
• Dense batch vectorization across the full segment retrieval, reconstruction,
  and projection pipeline
• Cross-platform compiled wheels (Linux x86_64/aarch64, macOS x86_64/arm64,
  Windows AMD64) built via cibuildwheel

What Changed

New Public API

Symbol	Purpose
LinearSpectralMapper	Compiled linear mapper for million-pixel throughput
BatchMappingArrayResult	Dense result type for array-oriented batch mapping
SpectralMapper.compile_linear_mapper()	Compile a prepared runtime into a reusable linear mapper
SpectralMapper.map_reflectance_debug()	Single-sample mapping with full diagnostic payload
SpectralMapper.map_reflectance_batch_debug()	Batch mapping returning rich MappingResult objects
SpectralMapper.map_reflectance_batch_arrays()	Explicit dense batch path returning BatchMappingArrayResult
SpectralMapper.map_reflectance_batch_ndarray()	Direct ndarray input batch mapping
SpectralMapper.map_reflectance_batch_to_zarr()	Stream large batch results to Zarr stores

Slim vs Debug Paths

map_reflectance() and map_reflectance_batch() now default to slim result
paths that skip neighbor ids, per-segment diagnostics, and review payloads.
Use the _debug() variants when you need the full payload.

CLI Additions

• --output-format zarr and --output-chunk-size flags for streaming batch output

Build and Distribution

• Wheels now include a compiled Rust extension for all major platforms
• cibuildwheel builds in the release workflow replace pure-Python wheel builds
• Package checks CI installs the Rust toolchain for editable-install tests

Full Changelog

See CHANGELOG.md
for the complete list of additions, changes, and fixes.