Test fixtures (#4241)

* fix test

* typo

* update

* update

* update

* update

* update

* update

* update

* update

* update

* update

* skip tests

* run cron only in master repo

* fix test

* update

* Add test

.github/workflows/full_testing.yml

@@ -7,13 +7,14 @@ on:  # yamllint disable-line rule:truthy
 jobs:
 
   pytest:
+    if: github.repository == 'pyg-team/pytorch_geometric'
     runs-on: ${{ matrix.os }}
 
     strategy:
       fail-fast: false
       matrix:
         os: [ubuntu-latest, windows-latest]
-        python-version: [3.7, 3.9]
+        python-version: [3.7, 3.8, 3.9, 3.10]
         torch-version: [1.10.0]
         include:
           - torch-version: 1.10.0
@@ -44,6 +45,8 @@ jobs:
       - name: Run tests
         run: |
           FULL_TEST=1 pytest --cov --cov-report=xml
+        shell:
+          bash
 
       - name: Upload coverage
         uses: codecov/codecov-action@v2

.github/workflows/testing.yml

@@ -14,8 +14,8 @@ jobs:
     strategy:
       fail-fast: false
       matrix:
-        os: [ubuntu-latest, windows-latest]
-        python-version: [3.7, 3.9]
+        os: [ubuntu-latest]
+        python-version: [3.9]
         torch-version: [1.10.0, 1.11.0]
         include:
           - torch-version: 1.10.0

CONTRIBUTING.md

@@ -80,12 +80,22 @@ Everytime you send a Pull Request, your commit will be built and checked against
 
    If you do not want to format your code manually, we recommend to use [`yapf`](https://github.com/google/yapf).
 
-2. Ensure that the entire test suite passes and that code coverage roughly stays the same. Please feel encouraged to provide a test with your submitted code.
+2. Ensure that the entire test suite passes and that code coverage roughly stays the same.
+   Please feel encouraged to provide a test with your submitted code.
+   To test, either run
 
    ```bash
    pytest --cov
    ```
 
+   or
+
+   ```bash
+   FULL_TEST=1 pytest --cov
+   ```
+
+   (which runs a set of additional but time-consuming tests) dependening on your needs.
+
 ## Building Documentation
 
 To build the documentation:

test/conftest.py

@@ -0,0 +1,47 @@
+import functools
+import os.path as osp
+import shutil
+
+import pytest
+
+from torch_geometric.data import Dataset
+
+
+def load_dataset(root: str, name: str, *args, **kwargs) -> Dataset:
+    r"""Returns a variety of datasets according to :obj:`name`."""
+    if 'karate' in name.lower():
+        from torch_geometric.datasets import KarateClub
+        return KarateClub(*args, **kwargs)
+    if name.lower() in ['cora', 'citeseer', 'pubmed']:
+        from torch_geometric.datasets import Planetoid
+        path = osp.join(root, 'Planetoid', name)
+        return Planetoid(path, name, *args, **kwargs)
+    if name in ['BZR', 'ENZYMES', 'IMDB-BINARY', 'MUTAG']:
+        from torch_geometric.datasets import TUDataset
+        path = osp.join(root, 'TUDataset')
+        return TUDataset(path, name, *args, **kwargs)
+    if name in ['ego-facebook', 'soc-Slashdot0811', 'wiki-vote']:
+        from torch_geometric.datasets import SNAPDataset
+        path = osp.join(root, 'SNAPDataset')
+        return SNAPDataset(path, name, *args, **kwargs)
+    if name.lower() in ['bashapes']:
+        from torch_geometric.datasets import BAShapes
+        return BAShapes(*args, **kwargs)
+    if name.lower() in ['dblp']:
+        from torch_geometric.datasets import DBLP
+        path = osp.join(root, 'DBLP')
+        return DBLP(path, *args, **kwargs)
+    if name in ['citationCiteseer', 'illc1850']:
+        from torch_geometric.datasets import SuiteSparseMatrixCollection
+        path = osp.join(root, 'SuiteSparseMatrixCollection')
+        return SuiteSparseMatrixCollection(path, name=name, *args, **kwargs)
+
+    raise NotImplementedError
+
+
+@pytest.fixture(scope='session')
+def get_dataset():
+    root = osp.join('/', 'tmp', 'pyg_test_datasets')
+    yield functools.partial(load_dataset, root)
+    if osp.exists(root):
+        shutil.rmtree(root)

test/data/test_lightning_datamodule.py

@@ -1,16 +1,11 @@
 import math
-import os.path as osp
-import random
-import shutil
-import sys
 import warnings
 
 import pytest
 import torch
 import torch.nn.functional as F
 
 from torch_geometric.data import LightningDataset, LightningNodeData
-from torch_geometric.datasets import DBLP, Planetoid, TUDataset
 from torch_geometric.nn import global_mean_pool
 
 try:
@@ -61,15 +56,13 @@ def configure_optimizers(self):
 @pytest.mark.skipif(no_pytorch_lightning, reason='PL not available')
 @pytest.mark.skipif(not torch.cuda.is_available(), reason='CUDA not available')
 @pytest.mark.parametrize('strategy', [None, 'ddp_spawn'])
-def test_lightning_dataset(strategy):
+def test_lightning_dataset(get_dataset, strategy):
     import pytorch_lightning as pl
 
-    root = osp.join('/', 'tmp', str(random.randrange(sys.maxsize)))
-    dataset = TUDataset(root, name='MUTAG').shuffle()
+    dataset = get_dataset(name='MUTAG').shuffle()
     train_dataset = dataset[:50]
     val_dataset = dataset[50:80]
     test_dataset = dataset[80:90]
-    shutil.rmtree(root)
 
     gpus = 1 if strategy is None else torch.cuda.device_count()
     if strategy == 'ddp_spawn':
@@ -146,15 +139,13 @@ def configure_optimizers(self):
 @pytest.mark.skipif(not torch.cuda.is_available(), reason='CUDA not available')
 @pytest.mark.parametrize('loader', ['full', 'neighbor'])
 @pytest.mark.parametrize('strategy', [None, 'ddp_spawn'])
-def test_lightning_node_data(strategy, loader):
+def test_lightning_node_data(get_dataset, strategy, loader):
     import pytorch_lightning as pl
 
-    root = osp.join('/', 'tmp', str(random.randrange(sys.maxsize)))
-    dataset = Planetoid(root, name='Cora')
+    dataset = get_dataset(name='Cora')
     data = dataset[0]
     data_repr = ('Data(x=[2708, 1433], edge_index=[2, 10556], y=[2708], '
                  'train_mask=[2708], val_mask=[2708], test_mask=[2708])')
-    shutil.rmtree(root)
 
     model = LinearNodeModule(dataset.num_features, dataset.num_classes)
 
@@ -237,13 +228,11 @@ def configure_optimizers(self):
 
 @pytest.mark.skipif(no_pytorch_lightning, reason='PL not available')
 @pytest.mark.skipif(not torch.cuda.is_available(), reason='CUDA not available')
-def test_lightning_hetero_node_data():
+def test_lightning_hetero_node_data(get_dataset):
     import pytorch_lightning as pl
 
-    root = osp.join('/', 'tmp', str(random.randrange(sys.maxsize)))
-    dataset = DBLP(root)
+    dataset = get_dataset(name='DBLP')
     data = dataset[0]
-    shutil.rmtree(root)
 
     model = LinearHeteroNodeModule(data['author'].num_features,
                                    int(data['author'].y.max()) + 1)

test/datasets/test_ba_shapes.py

@@ -1,10 +1,6 @@
-from torch_geometric.datasets import BAShapes
-
-
-def test_ba_shapes():
-    dataset = BAShapes()
+def test_ba_shapes(get_dataset):
+    dataset = get_dataset(name='BAShapes')
     assert str(dataset) == 'BAShapes()'
-
     assert len(dataset) == 1
     assert dataset.num_features == 10
     assert dataset.num_classes == 4

test/datasets/test_bzr.py

@@ -1,26 +1,21 @@
-import os.path as osp
-import random
-import shutil
-import sys
+from torch_geometric.testing import onlyFullTest
 
-from torch_geometric.datasets import TUDataset
-
-
-def test_bzr():
-    root = osp.join('/', 'tmp', str(random.randrange(sys.maxsize)))
-    dataset = TUDataset(root, 'BZR')
 
+@onlyFullTest
+def test_bzr(get_dataset):
+    dataset = get_dataset(name='BZR')
     assert len(dataset) == 405
     assert dataset.num_features == 53
     assert dataset.num_node_labels == 53
     assert dataset.num_node_attributes == 0
     assert dataset.num_classes == 2
-    assert dataset.__repr__() == 'BZR(405)'
+    assert str(dataset) == 'BZR(405)'
     assert len(dataset[0]) == 3
 
-    dataset = TUDataset(root, 'BZR', use_node_attr=True)
+
+@onlyFullTest
+def test_bzr_with_node_attr(get_dataset):
+    dataset = get_dataset(name='BZR', use_node_attr=True)
     assert dataset.num_features == 56
     assert dataset.num_node_labels == 53
     assert dataset.num_node_attributes == 3
-
-    shutil.rmtree(root)

test/datasets/test_enzymes.py

@@ -1,24 +1,16 @@
-import os.path as osp
-import random
-import shutil
-import sys
-
 import pytest
 import torch
 
-from torch_geometric.datasets import TUDataset
 from torch_geometric.loader import DataListLoader, DataLoader, DenseDataLoader
 from torch_geometric.transforms import ToDense
 
 
-def test_enzymes():
-    root = osp.join('/', 'tmp', str(random.randrange(sys.maxsize)))
-    dataset = TUDataset(root, 'ENZYMES')
-
+def test_enzymes(get_dataset):
+    dataset = get_dataset(name='ENZYMES')
     assert len(dataset) == 600
     assert dataset.num_features == 3
     assert dataset.num_classes == 6
-    assert dataset.__repr__() == 'ENZYMES(600)'
+    assert str(dataset) == 'ENZYMES(600)'
 
     assert len(dataset[0]) == 3
     assert len(dataset.shuffle()) == 600
@@ -63,18 +55,14 @@ def test_enzymes():
         assert list(data.mask.size()) == [600, 126]
         assert list(data.y.size()) == [600, 1]
 
-    dataset = TUDataset(root, 'ENZYMES', use_node_attr=True)
+
+def test_enzymes_with_node_attr(get_dataset):
+    dataset = get_dataset(name='ENZYMES', use_node_attr=True)
     assert dataset.num_node_features == 21
     assert dataset.num_features == 21
     assert dataset.num_edge_features == 0
 
-    shutil.rmtree(root)
-
-
-def test_cleaned_enzymes():
-    root = osp.join('/', 'tmp', str(random.randrange(sys.maxsize)))
-    dataset = TUDataset(root, 'ENZYMES', cleaned=True)
 
+def test_cleaned_enzymes(get_dataset):
+    dataset = get_dataset(name='ENZYMES', cleaned=True)
     assert len(dataset) == 595
-
-    shutil.rmtree(root)

test/datasets/test_imdb_binary.py

@@ -1,15 +1,9 @@
-import os.path as osp
-import random
-import shutil
-import sys
+from torch_geometric.testing import onlyFullTest
 
-from torch_geometric.datasets import TUDataset
-
-
-def test_imdb_binary():
-    root = osp.join('/', 'tmp', str(random.randrange(sys.maxsize)))
-    dataset = TUDataset(root, 'IMDB-BINARY')
 
+@onlyFullTest
+def test_imdb_binary(get_dataset):
+    dataset = get_dataset(name='IMDB-BINARY')
     assert len(dataset) == 1000
     assert dataset.num_features == 0
     assert dataset.num_classes == 2
@@ -20,5 +14,3 @@ def test_imdb_binary():
     assert data.edge_index.size() == (2, 146)
     assert data.y.size() == (1, )
     assert data.num_nodes == 20
-
-    shutil.rmtree(root)

test/datasets/test_karate.py

@@ -1,13 +1,9 @@
-from torch_geometric.datasets import KarateClub
-
-
-def test_karate():
-    dataset = KarateClub()
-
+def test_karate(get_dataset):
+    dataset = get_dataset(name='KarateClub')
+    assert str(dataset) == 'KarateClub()'
     assert len(dataset) == 1
     assert dataset.num_features == 34
     assert dataset.num_classes == 4
-    assert dataset.__repr__() == 'KarateClub()'
 
     assert len(dataset[0]) == 4
     assert dataset[0].edge_index.size() == (2, 156)

test/datasets/test_mutag.py

@@ -1,24 +1,14 @@
-import os.path as osp
-import random
-import shutil
-import sys
-
-from torch_geometric.datasets import TUDataset
-
-
-def test_mutag():
-    root = osp.join('/', 'tmp', str(random.randrange(sys.maxsize)))
-    dataset = TUDataset(root, 'MUTAG')
-
+def test_mutag(get_dataset):
+    dataset = get_dataset(name='MUTAG')
     assert len(dataset) == 188
     assert dataset.num_features == 7
     assert dataset.num_classes == 2
-    assert dataset.__repr__() == 'MUTAG(188)'
+    assert str(dataset) == 'MUTAG(188)'
 
     assert len(dataset[0]) == 4
     assert dataset[0].edge_attr.size(1) == 4
 
-    dataset = TUDataset(root, 'MUTAG', use_node_attr=True)
-    assert dataset.num_features == 7
 
-    shutil.rmtree(root)
+def test_mutag_with_node_attr(get_dataset):
+    dataset = get_dataset(name='MUTAG', use_node_attr=True)
+    assert dataset.num_features == 7