From 4a1e7628911e0e3db41c206769714640ae0e5277 Mon Sep 17 00:00:00 2001 From: Nick Date: Fri, 14 Aug 2026 11:41:24 +0800 Subject: [PATCH 1/2] Keep notebooks and README artwork out of the sdist setuptools-scm's file finder hands the sdist every git-tracked file, so the example notebooks and the sample rasters they read were swept in automatically: 4.9 MB of a 7.2 MB tarball, none of it read by anything in the package. The README artwork under assets/ went the same way, even though the README loads it over raw.githubusercontent.com rather than from the package. Add a MANIFEST.in pruning notebooks/, assets/ and .vscode/. The sdist drops to 3.9 MB and still builds and tests after unpacking (573 passed). Wheels were never affected -- they are built from `packages` alone -- and are byte-for-byte unchanged. Same fix as OmniWatermask 0.6.1, with one difference worth recording: that project keeps tests/ because its conda-forge recipe runs the suite against the sdist. MultiClean's recipe only import-checks, so tests/ is kept here for the weaker reason that an unpacked sdist should be verifiable. Its 4.9 MB of fixture rasters are now most of what is left. Co-Authored-By: Claude Opus 5 (1M context) --- CHANGELOG.md | 7 +++++++ MANIFEST.in | 21 +++++++++++++++++++++ 2 files changed, 28 insertions(+) create mode 100644 MANIFEST.in diff --git a/CHANGELOG.md b/CHANGELOG.md index 98f8950..423667c 100644 --- a/CHANGELOG.md +++ b/CHANGELOG.md @@ -19,6 +19,13 @@ All notable changes to MultiClean are documented here. have cached results produced with an even `smooth_edge_size`, regenerate them or expect a one-pixel offset against new output. +- The source distribution no longer ships the example notebooks, their sample + rasters, or the README artwork. `setuptools-scm` hands the sdist every + git-tracked file, so these were swept in automatically — 4.9 MB of a 7.2 MB + tarball, none of it read by anything in the package. A `MANIFEST.in` now + prunes them, taking the sdist to 3.9 MB. The notebooks remain on GitHub, + where the README links them. Wheels were never affected and are unchanged. + ## [0.4.0] - 2026-07-28 ### Changed diff --git a/MANIFEST.in b/MANIFEST.in new file mode 100644 index 0000000..7cb1c77 --- /dev/null +++ b/MANIFEST.in @@ -0,0 +1,21 @@ +# setuptools-scm hands the sdist every git-tracked file, which swept in the +# example notebooks and the sample rasters they read — 4.9 MB of a 7.2 MB +# tarball, for content nobody installing the package needs. They stay on +# GitHub, where the README's tutorial badge links them. The wheel was never +# affected; it is built from `packages` alone. +prune notebooks + +# README artwork and the notebook that regenerates it. The README loads these +# over raw.githubusercontent.com rather than from the package, so nothing in +# the sdist reads them. +prune assets + +# Editor config, of no use to anyone unpacking the sdist. +prune .vscode + +# Deliberately kept: tests/ so the sdist can be verified after unpacking, and +# the CI workflows, RELEASING.md and CHANGELOG.md so a packager can see how the +# release is built and what changed. Together they are ~48 KB, excepting the +# 4.9 MB of fixture rasters under tests/data. Note that nothing downstream runs +# the suite today — the conda-forge recipe only import-checks — so if sdist size +# becomes a problem again, tests/data is the thing left to reconsider. From df94a3ab6ef7f91bdb5b6ffbeb7ad6d1d79b482c Mon Sep 17 00:00:00 2001 From: Nick Date: Fri, 14 Aug 2026 11:46:56 +0800 Subject: [PATCH 2/2] Prepare 0.5.0 release Promote the [Unreleased] changelog section to [0.5.0] - 2026-08-14 and add the comparison link for the new version. A minor bump rather than a patch: the even-kernel fix changes output for even smooth_edge_size, and the default is 2, so anyone on defaults gets different results from the same input. Co-Authored-By: Claude Opus 5 (1M context) --- CHANGELOG.md | 5 ++++- 1 file changed, 4 insertions(+), 1 deletion(-) diff --git a/CHANGELOG.md b/CHANGELOG.md index 423667c..577c29c 100644 --- a/CHANGELOG.md +++ b/CHANGELOG.md @@ -4,6 +4,8 @@ All notable changes to MultiClean are documented here. ## [Unreleased] +## [0.5.0] - 2026-08-14 + ### Fixed - Edge smoothing no longer translates the output by one pixel down and to the right when `smooth_edge_size` is even. `cv2.morphologyEx(MORPH_OPEN)` applies @@ -110,7 +112,8 @@ All notable changes to MultiClean are documented here. - Documentation: README, two example notebooks (land use, cloud shadow), and a Google Colab tutorial notebook. -[Unreleased]: https://github.com/DPIRD-DMA/MultiClean/compare/v0.4.0...HEAD +[Unreleased]: https://github.com/DPIRD-DMA/MultiClean/compare/v0.5.0...HEAD +[0.5.0]: https://github.com/DPIRD-DMA/MultiClean/compare/v0.4.0...v0.5.0 [0.4.0]: https://github.com/DPIRD-DMA/MultiClean/compare/v0.3.1...v0.4.0 [0.3.1]: https://github.com/DPIRD-DMA/MultiClean/compare/v0.3.0...v0.3.1 [0.3.0]: https://github.com/DPIRD-DMA/MultiClean/compare/v0.2.0...v0.3.0