fastlowess (Java) (development version)#

Added#

  • Added a new Java binding (bindings/java), consuming the fastLowess Rust core via a JNI shared library (fastlowess-java crate, using the jni crate). Provides Lowess (batch), StreamingLowess, and OnlineLowess model types with an idiomatic Java API (fluent Options/StreamingOptions/OnlineOptions builders, Result/PointResult/Diagnostics records with Optional/OptionalInt/OptionalDouble accessors, AutoCloseable-based resource management for use with try-with-resources). Targets Java 25 LTS (maven.compiler.release). Ships with JUnit 5 unit tests, a Makefile (java/java-dev/java-clean targets, wired into the root Makefile), and docs (getting-started.md, api.md, api-streaming.md, api-online.md, installation.md, concepts.md, adapter-choice.md, kernels.md, robustness.md, scaling.md, boundary.md, intervals.md, cross-validation.md, custom-weights.md, merge.md, gpu-backend.md, benchmarks.md).

Changed#

  • Added a large benchmark category to benchmarks/rfastlowess.R and benchmarks/stats_lowess.R (n = 50000, delta = 0 to disable stats::lowess’s interpolation shortcut for a fair, exact-computation comparison), since every existing category completed in well under 100ms. Reliably takes ~13-16s for stats::lowess and ~3s for fastLowess (serial). Expanded it to 4 scenarios stressing different parameters: large_delta_0 (the original exact-fit case), large_delta_0.1 (same workload with delta left at its default/auto value, showing the interpolation shortcut’s speedup), large_high_iter (10 robustness iterations instead of 3), and large_high_fraction (n = 20000, fraction = 0.67, since a 0.67 fraction at n = 50000 takes over a minute). benchmarks/compare.py’s plot grid grew from 5x2 to 7x2 rows to fit the new categories (large_delta_0/large_delta_0.1 share one “large_delta” chart, matching how fraction_*/iterations_* already group into shared charts).
  • Consolidated every crate/binding README: merged the “Installation” and “Documentation” sections, replaced GitHub-only alert syntax with plain blockquotes, removed the redundant “API Reference” and “Changelog” sections (each now has its own docs-site page), shortened the “GPU Backend” blurb, and added a “Read more” link to the Concepts page. The top-level repository README is unchanged, since it’s only ever viewed on GitHub.
  • Renamed the batch adapter’s “When to Use” heading to “When to Use Batch Adapter” across every binding/crate’s API docs.
  • Vendored the doxygen-awesome-css theme (v2.4.2) for a modern, sidebar-only cpp Doxygen site with automatic dark mode.
  • Added dev/update_changelogs.py, which regenerates a per-binding/crate NEWS.md/news.md from the root CHANGELOG.md. Wired into every docs site’s navigation, the Rust crates’ rustdoc module tree, and every Makefile dev target.
  • Replaced kernels.md’s “Choosing a Kernel” mermaid flowchart (every binding/crate) with an equivalent decision table, since Doxygen and rustdoc don’t render mermaid and the deeply-nested diamond chain was hard to read even where it did render.
  • Replaced adapter-choice.md/adapters.md’s “Overview” flowchart (mermaid in most bindings/crates, ASCII art in the C++ docs) with an equivalent decision table, unifying on a single rendering-agnostic format across every binding/crate.
  • Moved the duplicated “GPU Acceleration” section (installation, usage, supported features, feature comparison) out of api.md in the C++, Node.js, and Python bindings and the fastLowess crate into each one’s dedicated gpu-backend.md guide, which also gained the “Hardware Requirements” and “Performance Considerations” subsections previously only in api.md; api.md now links to gpu-backend.md with a short blurb instead. Removed the same section from the lowess crate’s api.md entirely, since that crate has no gpu Cargo feature or GPU backend to document.
  • Consolidated parameters.md/the auto-generated @autodocs parameter reference across every binding and crate (C++, Julia, Node.js, Python, WASM, fastLowess, lowess): merged its unique content (fraction/iterations choice guidance, delta’s per-adapter default, and an inline zero_weight_fallback behavior table) into each api.md’s builder/options tables (Julia: into the Lowess/StreamingLowess/OnlineLowess docstrings), and removed parameters.md itself along with its docs-site navigation entries, doc::parameters rustdoc module, and cross-references (now pointing at api.md) — the parameter tables, kernel/robustness/boundary/scaling/merge-strategy option lists, and interval/custom-weights code examples it duplicated already live on their own dedicated pages.

Fixed#

  • Fixed docs.yml triggering GitHub’s “pages build and deployment” once per docs job; per-language jobs now upload artifacts, and a single final deploy job pushes to gh-pages once per run.
  • Fixed docs.yml’s reliance on GitHub’s legacy branch-based Pages deployment, which auto-triggers an unpinned, GitHub-managed “pages build and deployment” job on every gh-pages push (surfacing deprecation warnings, e.g. for Node.js 20, that aren’t fixable from this repo). The former deploy job is now build, which still pushes the merged _site to gh-pages as a cache for future incremental runs, but publishing now goes through actions/upload-pages-artifact and a new deploy job using the official actions/deploy-pages, which this repo pins directly. Requires the repository’s Pages source to be switched to “GitHub Actions” in settings.
  • Fixed every benchmark category in benchmarks/rfastlowess.R failing with attempt to apply non-function: it called the R6-style model$fit(x, y), but fit is an S3 generic (fit(model, x, y)), not a field on the Lowess object. Also fixed benchmarks/stats_lowess.R resolving its output/ directory relative to the current working directory instead of the script’s own location (unlike rfastlowess.R, which already did this correctly), so results could land outside benchmarks/output/ depending on how the script was invoked.
  • Fixed the “Handling Outliers” quickstart example (every binding and the lowess/fastLowess crates) printing nothing: with only 6 points and fraction = 0.5, the local window is small enough that tricube weighting drives the farthest neighbor’s weight to ~0, leaving just 2 effectively-weighted points, which a degree-1 fit reproduces exactly (zero residual, no downweighting) — confirmed directly against the lowess/loess core, not binding-specific. Bumped to fraction = 0.7, which correctly downweights the injected outlier.
  • Fixed the R OnlineLowess() roxygen example printing one line per point (48 lines for a 50-point loop); it now collects the smoothed values and prints only head(smoothed, 5).
  • Fixed the R add_point() roxygen example always printing NULL, since a single call never reaches the default min_points = 3; it now uses min_points = 2L and shows the second (non-NULL) call’s result.
  • Fixed the R robustness.Rmd “Detecting Outliers” example printing 22 lines at the weight < 0.5 threshold, most of them incidental noise rather than the 3 deliberately injected outliers; tightened to weight < 0.05, which isolates the points effectively excluded by the fit.
  • Fixed the R merge.Rmd “Choosing Chunk Size and Overlap” example constructing a StreamingLowess model but never printing anything; it now prints the computed overlap size and its percentage of chunk_size.
  • Fixed the R use-case-genomics.Rmd ChIP-seq example never calling fit(), so result referenced a stale variable from an earlier chunk and the smoothed line either failed to plot or didn’t align with the current example’s x-range; added the missing result <- fit(model, positions, signal_noisy) call.
  • Fixed the R use-case-real-time.Rmd “Update Modes” example constructing a "full"-mode OnlineLowess model but never feeding it data or plotting a result; it now runs the same accumulate-and-plot pattern as the preceding example.
  • Fixed the “Detecting Outliers” example’s robustness.md page (C++, Node.js, WASM, and the lowess/fastLowess crates) printing an unbounded number of “is likely an outlier” lines; capped output at 5 lines, matching the already-capped Julia and Python versions and the R vignette fix above.
  • Fixed the Julia intervals.md “Confidence Intervals” and “Standard Errors” examples each looping over all 100 points instead of a short sample; switched to result.y[1:5]/result.confidence_lower[1:5]/result.standard_errors[1:5]-style slicing, matching the already-concise Python version.

fastlowess (Java) 3.1.0#

Added#

  • Added a GitHub Pages landing page at the repository root, built from README.md via pandoc and deployed by docs.yml.
  • Added a GitHub workflow for running validation scripts.

Changed#

  • Split the monolithic .github/workflows/ci.yml into seven per-language workflow files: ci-rust.yml, ci-python.yml, ci-julia.yml, ci-nodejs.yml, ci-wasm.yml, ci-cpp.yml, and ci-r.yml. Each file carries the relevant ci (multi-OS matrix), asan, and gpu jobs for its language.
  • Each crate/binding sub-Makefile now runs dev/verify_snippets.py --lang <lang> for its own language as the final step of make default. The root docs-test target remains as a convenience to run all languages at once.
  • Split dev/verify_snippets.py into a lean orchestrator and a dev/runners/ package. Each language has its own module (python.py, julia.py, nodejs.py, r.py, wasm.py, rust.py, cpp.py) containing its run_<lang>() function and a skip_reason() predicate. Shared types (Snippet, RunResult) and utilities live in runners/base.py; the registry (RUNNERS, SKIP_CHECKS) is exported from runners/__init__.py.
  • Moved CHANGELOG and CONTRIBUTING guides to project root.
  • Updated README files to be binding/crate specific instead of one generic README for all bindings/crates.

Fixed#

  • Fixed .cargo/config.toml hardcoding absolute c:/rtools45/... paths for the x86_64-pc-windows-gnu linker and ar tool. Replaced with bare tool names resolved via PATH, matching the existing fix in bindings/r/src/cargo-config.toml.

fastlowess (Java) 3.0.0#

Changed#

  • Removed dev/isolate_cargo.py, dev/check_root_cargo.py, dev/fix_doc_snippets.py, and check_js_licenses.js — workspace isolation, doc-snippet transformation, and license checks are no longer needed.
  • Split the monolithic root Makefile into per-crate/binding sub-Makefiles (e.g. crates/lowess/Makefile, bindings/r/Makefile), each invokable directly via make -f path/Makefile. The root Makefile now only aggregates (docs, check-msrv, all*).
  • Moved Rust and binding tests into their respective crate/binding directories (e.g. tests/lowess/ → crates/lowess/tests/lowess/, tests/cpp/ → bindings/cpp/tests/). Removed the standalone tests/ workspace packages and bindings/r/demo/.

fastlowess (Java) 2.0.0#

Changed#

  • Renamed all public API method and option names from camelCase to snake_case across every binding and all documentation. This is a breaking change for all consumers of the C++, Node.js, and WASM APIs.
  • Converted all documentation tables to compact single-space format.
  • Updated .clang-tidy to configure lower_case as the required naming convention for functions and member functions, matching the new snake_case public API.
  • Moved BENCHMARKS.md, CHANGELOG.md, and CONTRIBUTING.md from the repository root into docs/ and added them to the documentation site navigation.
  • Added a [patch.crates-io] section to the root Cargo.toml so all workspace bindings resolve fastLowess and lowess to the local workspace crates during development, replacing the previously-used registry (crates.io) versions.
  • Eliminated all local parse_* functions that each binding previously duplicated independently. Option parsing and builder application now delegates to fastLowess::binding_support, ensuring consistent aliases, validation messages, and behaviour across every language frontend.
  • Replaced direct use of KFold / LOOCV constructor types in the cross-validation path with binding_support::apply_cross_validation.
  • Split the Julia release CI workflow into two separate workflows: release-julia-jll.yml (triggered on release, opens the Yggdrasil PR) and release-julia-register.yml (manual dispatch, triggers JuliaRegistrator once the JLL PR is merged).
  • Major documentation improvements.

fastlowess (Java) 1.3.0#

Added#

  • Added prerequisites for different bindings and platforms to CONTRIBUTING.md
  • Updated docs/assets/diagrams/lowess_smoothing_concept.svg to correctly illustrate LOWESS concepts (robustness iterations, bisquare re-weighting, outlier downweighting) instead of the generic LOESS algorithm it previously depicted.
  • Modified docs/requirements.txt to update the versions of the documentation dependencies.
  • Improved CI tests and coverage.
  • Modified Makefile to be truely cross-platform.
  • Added sanitizer check for all bindings and crates.

Fixed#

  • Fixed R ASAN tests failing to compile vignettes by passing --no-build-vignettes to rcmdcheck.
  • Upgraded ASAN test environment to use modern rocker/r-devel-san:latest image and RDscript to resolve outdated readelf warnings.
  • Fixed Makefile idempotency checks on Linux by providing a default /tmp fallback for the TEMP directory variable.
  • Fixed accidental root Cargo.toml workspace isolation leaks by adding checked-in pre-commit and pre-push git hook guards that restore Cargo.toml.bak when present and fail loudly if required workspace members are still commented out.
  • Added a repo-local .cargo/config.toml that sets CC=clang-cl for x86_64-pc-windows-msvc, fixing Criterion 0.8 benchmark builds on Windows when cc-rs would otherwise pick clang.exe and fail to link alloca.

fastlowess (Java) 1.2.0#

Fixed#

  • Fixed project logo.

fastlowess (Java) 0.99.9#

Changed#

  • Bump rust version to 1.88 for better stability
  • Change function-based builder pattern in the bindings to class-based builder pattern, allowing true streaming and online processing
  • Improve API docs

fastlowess (Java) 0.99.7#

Fixed#

  • Fix README file links
  • Fix Makefile bug with R versioning

fastlowess (Java) 0.99.6#

Fixed#

  • Fix README file formats and links

fastlowess (Java) 0.99.5#

Changed#

  • Reduced package size significantly by removing unnecessary dev files and docs from the final package.
  • Implemented comprehensive Cargo workspace inheritance pattern
  • Unified MSRV to 1.85.0
  • Centralized all metadata (version, authors, edition, license, etc.) in root Cargo.toml
  • All crates now use workspace = true for shared configuration
  • Created unified README.md for all crates/packages
  • Created unified CHANGELOG.md for all crates/packages
  • Created unified LICENSE for all crates/packages
  • Created unified .gitignore for all crates/packages
  • Added comprehensive badges from all packages

For the full changelog, see: https://github.com/thisisamirv/lowess-project/blob/main/CHANGELOG.md