ADD: pytest tests and run in CI (#45)

* add pytest for tests

* add tests and run in ci

* add tifffile and imagecodecs to dev deps

* patch resolution unit for py 3.7

* fix for python 3.7 tifffile

* fix version_info check to be false for 3.7

* remove the --help cmds in tests

* add stubs for tests of converters

* add ci status badge

* add tests of geojson conversion

* use np.allclose instead of == for floats

.github/workflows/ci.yml

@@ -24,7 +24,5 @@ jobs:
           python -m pip install torch torchvision --extra-index-url https://download.pytorch.org/whl/cpu
           python -m pip install --editable .[dev] --find-links https://girder.github.io/large_image_wheels
       - name: Run tests
-        # TODO: add more tests...
         run: |
-          wsinfer --help
-          wsinfer run --help
+          python -m pytest --verbose tests/

README.md

@@ -6,6 +6,8 @@ Original H&E                        |  Heatmap of Tumor Probability
 
 🔥 🚀 Blazingly fast pipeline to run patch-based classification models on whole slide images.
 
+![Continuous Integration](https://github.com/kaczmarj/patch-classification-pipeline/actions/workflows/ci.yml/badge.svg)
+
 # Table of contents
 
 - [Available models](#available-models)

setup.cfg

@@ -47,7 +47,10 @@ install_requires =
 dev =
     black
     flake8
+    imagecodecs  # for tifffile
     mypy
+    pytest
+    tifffile
 
 [options.entry_points]
 console_scripts =

tests/test_all.py

@@ -0,0 +1,141 @@
+import json
+from pathlib import Path
+import sys
+
+from click.testing import CliRunner
+import numpy as np
+import pandas as pd
+import pytest
+import tifffile
+
+
+@pytest.fixture
+def tiff_image(tmp_path: Path) -> Path:
+    x = np.empty((4096, 4096, 3), dtype="uint8")
+    x[...] = [160, 32, 240]  # rgb for purple
+    path = Path(tmp_path / "images" / "purple.tif")
+    path.parent.mkdir(exist_ok=True)
+
+    if sys.version_info >= (3, 8):
+        tifffile.imwrite(
+            path,
+            data=x,
+            compression="zlib",
+            tile=(256, 256),
+            # 0.25 micrometers per pixel.
+            resolution=(40000, 40000),
+            resolutionunit=tifffile.RESUNIT.CENTIMETER,
+        )
+    else:
+        # Earlier versions of tifffile do not have resolutionunit kwarg.
+        tifffile.imwrite(
+            path,
+            data=x,
+            compression="zlib",
+            tile=(256, 256),
+            # 0.25 micrometers per pixel.
+            resolution=(40000, 40000, "CENTIMETER"),
+        )
+
+    return path
+
+
+def test_cli_list():
+    from wsinfer.cli.cli import cli
+
+    runner = CliRunner()
+    result = runner.invoke(cli, ["list"])
+    assert "resnet34" in result.output
+    assert "TCGA-BRCA-v1" in result.output
+    assert result.exit_code == 0
+
+
+def test_cli_run_and_convert(tiff_image: Path, tmp_path: Path):
+    from wsinfer.cli.cli import cli
+
+    runner = CliRunner()
+    results_dir = tmp_path / "inference"
+    result = runner.invoke(
+        cli,
+        [
+            "run",
+            "--wsi_dir",
+            tiff_image.parent,
+            "--model",
+            "resnet34",
+            "--weights",
+            "TCGA-BRCA-v1",
+            "--results_dir",
+            results_dir,
+        ],
+    )
+    assert result.exit_code == 0
+    assert (results_dir / "model-outputs").exists()
+    df = pd.read_csv(results_dir / "model-outputs" / "purple.csv")
+    assert df.columns.tolist() == [
+        "slide",
+        "minx",
+        "miny",
+        "width",
+        "height",
+        "prob_notumor",
+        "prob_tumor",
+    ]
+    assert (df.loc[:, "slide"] == str(tiff_image)).all()
+    assert (df.loc[:, "width"] == 350).all()
+    assert (df.loc[:, "height"] == 350).all()
+    assert (df.loc[:, "width"] == 350).all()
+    assert np.allclose(df.loc[:, "prob_notumor"], 0.9525967836380005)
+    assert np.allclose(df.loc[:, "prob_tumor"], 0.04740329459309578)
+
+    # Test conversion scripts.
+    geojson_dir = results_dir / "geojson"
+    result = runner.invoke(cli, ["togeojson", str(results_dir), str(geojson_dir)])
+    assert result.exit_code == 0
+    with open(geojson_dir / "purple.json") as f:
+        d = json.load(f)
+    assert len(d["features"]) == 144
+
+    for geojson_row in d["features"]:
+        assert geojson_row["type"] == "Feature"
+        assert geojson_row["id"] == "PathTileObject"
+        assert geojson_row["geometry"]["type"] == "Polygon"
+
+    # Check the probability values.
+    assert all(
+        np.allclose(dd["properties"]["measurements"][0]["value"], 0.9525967836380004)
+        for dd in d["features"]
+    )
+    assert all(
+        np.allclose(dd["properties"]["measurements"][1]["value"], 0.0474032945930957)
+        for dd in d["features"]
+    )
+
+    # Check the names.
+    assert all(
+        dd["properties"]["measurements"][0]["name"] == "prob_notumor"
+        for dd in d["features"]
+    )
+    assert all(
+        dd["properties"]["measurements"][1]["name"] == "prob_tumor"
+        for dd in d["features"]
+    )
+
+    # Check the coordinate values.
+    for df_row, geojson_row in zip(df.itertuples(), d["features"]):
+        maxx = df_row.minx + df_row.width
+        maxy = df_row.miny + df_row.height
+        df_coords = [
+            [maxx, df_row.miny],
+            [maxx, maxy],
+            [df_row.minx, maxy],
+            [df_row.minx, df_row.miny],
+            [maxx, df_row.miny],
+        ]
+        assert [df_coords] == geojson_row["geometry"]["coordinates"]
+
+
+@pytest.mark.xfail
+def test_convert_to_sbu():
+    # TODO: create a synthetic output and then convert it. Check that it is valid.
+    assert False