NEWS

---

Rminibwa 0.4.0-0.0.1.9000

• Update the bundled minibwa sources to upstream v0.4 / MB_VERSION 0.4-r400 at commit b2dcea9b43c02cef47556f993fbd54bc0002cd4a.
• Include upstream's heavily clipped alignment rescoring improvement and SAM PG-CL subcommand fix.

Rminibwa 0.3.0-0.0.1.9000

• Add a MinGW/Rtools-only kthread.c dead-return guard after pthread_exit() to avoid -Wreturn-type install warnings on Windows.
• Make kseq.h's kroundup64() width-safe on 32-bit targets such as wasm32, avoiding Emscripten -Wshift-count-overflow diagnostics.
• Add mb_map_batch() for native columnar mapping of many query sequences in one call, plus mb_map_pair_batch() for explicit two-batch paired-end input.
• Add native FASTX iterators and mb_map_fastx_batch() so file-backed batches can be mapped without materializing sequence strings in R.
• Add a read-level sidecar to alignment batches with read IDs, query lengths, fragment/mate fields, and per-read hit offsets/counts, exposed through ALTREP read columns and the C API.
• Add a concise print method for native alignment batches.
• Return simd_info() as a classed rminibwa_simd_info object with a concise print method for dispatch diagnostics.
• Fold in upstream minibwa v0.3 / MB_VERSION 0.3-r391 at commit a679d17cb822c6cca67ab78c439f9c9826f1b07d.
• Switch package versions to the <minibwa-version>-<Rminibwa-version> scheme; development builds now use 0.3.0-0.0.1.9000.
• Add a MinGW/Rtools-only kalloc.c format-string guard for size_t diagnostics while leaving the non-MinGW upstream code path unchanged.

Rminibwa 0.0.0.9000

• Bootstrap Rminibwa as a GPL (>= 2) R package wrapping pinned upstream lh3/minibwa C sources.
• Add native bindings for index build/load, raw-vector query mapping, native external-pointer alignment batches, ALTREP column views, and packed CIGAR accessors.
• Install inst/include/Rminibwa.h and register C-callable accessors so downstream packages can consume alignment batches directly from C.
• Add staged KSW SIMD dispatch following the RsimdDispatch pattern: portable scalar/SIMDe fallback, SSE4, and AVX2 backends are compiled as separate objects and selected at runtime without global ISA compiler flags.
• Add AVX2 dual-gap extd2 support from the minibwa binding work and keep it as an optional runtime-dispatched backend.
• Vendor minibwa at commit f0fd108456d459e5da3755f29c8700dc7f6fe0df with a patch queue and provenance metadata.
• Add Rtinycc inline downstream-C examples in the README, vignette, and tests.
• Build and install a package-provided minibwa executable from vendored sources; CLI wrappers now use it by default.
• Add developer-only benchmarks comparing internal staged backends and optional external Python/Rust bindings from an explicit local checkout.
• Add make asm / tools/check-assembly.R to audit generated SIMD instruction families in staged objects and shared libraries.
• Add pkgdown configuration and GitHub Actions workflows for standard R CMD check, Fedora/R-devel CRAN-style check, and GitHub Pages deployment.