Test on other OSs, remove CI benchmarks (pydantic#366)

* Test on other OSes, remove benchmarks

* fix pytest-speed for 3.7

* remove windows xfail, revert pyproject changes

* uprev pytest, pytest-pretty

* add print to windows

* more eprints

* remove test-os py cache

* remove extra prints

* revert xfail

* remove pytest-speed on codspeed ci

* install ptyest-benchmark with codspeed

* clear codspeed cache

.github/workflows/ci.yml

@@ -8,6 +8,9 @@ on:
       - '**'
   pull_request: {}
 
+env:
+  COLUMNS: 120
+
 jobs:
   test-cpython:
     name: test ${{ matrix.python-version }} rust ${{ matrix.rust-toolchain }}
@@ -152,6 +155,38 @@ jobs:
 
       - run: pytest
 
+  test-os:
+    name: test on ${{ matrix.os }}
+
+    strategy:
+      fail-fast: false
+      matrix:
+        os: [ubuntu, macos, windows]
+
+    runs-on: ${{ matrix.os }}-latest
+    steps:
+      - uses: actions/checkout@v3
+
+      - name: cache rust
+        uses: Swatinem/rust-cache@v2
+        with:
+          key: ${{ matrix.os }}-v1
+
+      - name: set up python
+        uses: actions/setup-python@v4
+        with:
+          python-version: '3.11'
+
+      - run: pip install -r tests/requirements.txt
+
+      - run: pip install -e .
+        env:
+          RUST_BACKTRACE: 1
+
+      - run: pip freeze
+
+      - run: pytest
+
   lint:
     runs-on: ubuntu-latest
 
@@ -202,7 +237,7 @@ jobs:
       - run: npm run lint
 
   bench:
-    # run the rust benchmarks
+    name: rust benchmarks
     runs-on: ubuntu-latest
 
     steps:
@@ -271,7 +306,7 @@ jobs:
   # https://github.com/marketplace/actions/alls-green#why used for branch protection checks
   check:
     if: always()
-    needs: [test-cpython, test-pypy, lint, bench, build-wasm-emscripten]
+    needs: [test-cpython, test-pypy, test-os, lint, bench, build-wasm-emscripten]
     runs-on: ubuntu-latest
     steps:
       - name: Decide whether the needed jobs succeeded or failed
@@ -374,7 +409,7 @@ jobs:
 
       - run: ${{ matrix.ls || 'ls -lh' }} dist/
 
-      - run: twine check dist/*
+      - run: twine check --strict dist/*
 
       - uses: actions/upload-artifact@v3
         with:
@@ -521,7 +556,7 @@ jobs:
           name: pypi_files
           path: dist
 
-      - run: twine check dist/*
+      - run: twine check --strict dist/*
 
       - name: upload to pypi
         run: twine upload dist/*

.github/workflows/codspeed.yml

@@ -26,18 +26,22 @@ jobs:
         with:
           path: ${{ env.pythonLocation }}
           key: >
-            py
+            py-v2
             ${{ env.pythonLocation }}
             ${{ hashFiles('tests/requirements.txt') }}
             ${{ hashFiles('pyproject.toml') }}
 
       - run: pip install -r tests/requirements.txt
         if: steps.cache-py.outputs.cache-hit != 'true'
 
-      - run: pip install pytest-codspeed~=1.2.0
+      - run: pip uninstall pytest-speed -y
+        if: steps.cache-py.outputs.cache-hit != 'true'
+
+      - run: pip install pytest-codspeed~=1.2.0 pytest-benchmark==4.0.0
         if: steps.cache-py.outputs.cache-hit != 'true'
 
       - run: rm rust-toolchain
+
       - name: install rust stable
         uses: dtolnay/rust-toolchain@stable
 

README.md

@@ -71,8 +71,6 @@ except ValidationError as e:
 Pydantic-core is currently around 17x faster than pydantic standard.
 See [`tests/benchmarks/`](./tests/benchmarks/) for details.
 
-Benchmarks overtime can be seen [here](https://pydantic.github.io/pydantic-core/dev/bench/).
-
 This relative performance will be less impressive for small models but could be significantly move impressive
 for deeply nested models.
 

tests/requirements.txt

@@ -1,10 +1,10 @@
 coverage==7.0.5
 dirty-equals==0.5.0
 hypothesis==6.63.0
-pytest==7.1.2
-pytest-benchmark==4.0.0
+pytest==7.2.1
+pytest-speed==0.3.5
 pytest-mock==3.10.0
-pytest-pretty==0.0.1
+pytest-pretty==1.0.1
 pytest-timeout==2.1.0
 pydantic==1.10.4;python_version>="3.8"
 pytz==2022.7.1