adhara 0.1.0__tar.gz

adhara-0.1.0/.github/workflows/ci.yml

@@ -0,0 +1,86 @@
+name: CI
+
+on:
+  push:
+    branches: [main]
+    tags: ['v*']
+  pull_request:
+  workflow_dispatch:
+
+permissions:
+  contents: read
+
+jobs:
+  test:
+    runs-on: ubuntu-latest
+    steps:
+      - uses: actions/checkout@v4
+      - uses: dtolnay/rust-toolchain@stable
+      - uses: actions/setup-python@v5
+        with:
+          python-version: '3.12'
+      - run: cargo test --release
+      - run: pip install -e '.[dev]'
+      - run: pytest -q
+
+  swift:
+    runs-on: macos-15
+    steps:
+      - uses: actions/checkout@v4
+      - uses: dtolnay/rust-toolchain@stable
+      - run: scripts/xcframework.sh
+      - run: swift test
+
+  build:
+    needs: test
+    strategy:
+      fail-fast: false
+      matrix:
+        os: [ubuntu-latest, ubuntu-24.04-arm, macos-14, macos-15-intel]
+    runs-on: ${{ matrix.os }}
+    steps:
+      - uses: actions/checkout@v4
+      - uses: PyO3/maturin-action@v1
+        with:
+          args: --profile dist --out dist
+          manylinux: 2_28   # usearch's numkong needs a newer gcc than manylinux2014 has
+      - uses: actions/upload-artifact@v4
+        with:
+          name: wheels-${{ matrix.os }}
+          path: dist
+
+  sdist:
+    runs-on: ubuntu-latest
+    steps:
+      - uses: actions/checkout@v4
+      - uses: PyO3/maturin-action@v1
+        with:
+          command: sdist
+          args: -o dist
+      - uses: actions/upload-artifact@v4
+        with:
+          name: wheels-sdist
+          path: dist
+
+  publish:
+    if: startsWith(github.ref, 'refs/tags/v')
+    needs: [build, sdist, swift]
+    runs-on: ubuntu-latest
+    permissions:
+      contents: write
+    steps:
+      - uses: actions/download-artifact@v4
+        with:
+          pattern: wheels-*
+          path: dist
+          merge-multiple: true
+      - uses: softprops/action-gh-release@v2
+        with:
+          files: dist/*
+          generate_release_notes: true
+      - uses: PyO3/maturin-action@v1
+        env:
+          MATURIN_PYPI_TOKEN: ${{ secrets.PYPI_TOKEN }}
+        with:
+          command: upload
+          args: --non-interactive --skip-existing dist/*

adhara-0.1.0/.gitignore

@@ -0,0 +1,11 @@
+target/
+*.so
+*.dylib
+__pycache__/
+.venv/
+*.sqlite*
+*.usearch*
+.pytest_cache/
+swift/.build/
+swift/AdharaCore.xcframework/
+.build/

adhara-0.1.0/CLAUDE.md

@@ -0,0 +1,31 @@
+# Working in this repo
+
+The Rust code in `src/` is the source. The Python package (`python/adhara`) and the Swift package (`swift/`, `Package.swift`) are thin layers over it. `scripts/gen_tables.py` writes `src/tables.rs`. Never edit `src/tables.rs` by hand.
+
+## The one rule
+
+adhara gives the results of litesearch. A faster method that changes a result is a bug. The tests in `tests/` compare adhara with litesearch directly.
+
+- The folds equal `litesearch/sanskrit.py` on every word of the test corpus.
+- The tokens equal the litesearch chain. One apsw quirk is the exception. `RESULTS.md` describes it.
+- `adhara_hybrid` returns the rows of litesearch `search` in the same order. `ann` selects the HNSW search or the exact search, as `ann` does in litesearch.
+
+## Three parts
+
+- `fold.rs` is the port of `sanskrit.py`. It does not use SQLite.
+- `tok.rs` is the FTS5 tokenizer `adhara`. Porter goes outside it (`porter adhara`), as litesearch puts porter outside its chain.
+- `hybrid.rs` is the table function `adhara_hybrid`. It reads the litesearch schema: `<store>`, `<store>_fts` and `usearch_indices`.
+
+## Builds
+
+- `--features python` builds the Python module. The same file is the SQLite extension (`sqlite3_adhara_init`).
+- `--features static` builds SQLite into the library, for Swift (`adhara_register`). This build also registers `simplify` for the litesearch chain. The extension does not, because apsw registers its own `simplify` on a Python connection.
+- A change in tokens needs a new tokenizer name. Do not change the behaviour of an existing name. FTS5 tokenizes old content again on delete, so a changed token corrupts an index.
+
+## Numbers
+
+Each speed claim has a row in `RESULTS.md` and a script in `bench/`. Change the number before you change the claim.
+
+## Licence
+
+MIT.