Merge pull request #63 from SystemsGenetics/integration-tests

Add integration tests for StarchArea and BlushColor pipelines

Author: AdenAthar

tests/test_Analyses/test_integration.py

@@ -0,0 +1,224 @@
+"""
+Integration tests for analysis pipelines.
+
+Unlike unit tests that call _calculateStarch / _calculateBlush directly with
+in-memory numpy arrays, these tests exercise the full _processImage path:
+write a real PNG to disk → create RGBImage from that path → call _preRun +
+_processImage → verify the result has the expected structure and values.
+
+This catches bugs in file I/O, Image/ImageIO wiring, and Value assembly that
+unit tests miss (e.g. the FileDirValue validation-order bug documented in
+BUGFIXES.md would have surfaced here).
+"""
+
+import os
+
+import cv2
+import numpy as np
+import pytest
+
+from Granny.Analyses.BlushColor import BlushColor
+from Granny.Analyses.StarchArea import StarchArea, StarchScales
+from Granny.Models.Images.RGBImage import RGBImage
+
+
+# ---------------------------------------------------------------------------
+# Helpers
+# ---------------------------------------------------------------------------
+
+def _write_synthetic_png(path: str, bgr_value: tuple, size: int = 80) -> None:
+    """Write a solid-colour BGR image to *path* as a PNG."""
+    img = np.full((size, size, 3), bgr_value, dtype=np.uint8)
+    cv2.imwrite(path, img)
+
+
+def _write_mixed_starch_png(path: str, dark_fraction: float = 0.5, size: int = 100) -> None:
+    """
+    Write a two-tone image with a controlled dark/bright split.
+
+    Uses values 20 (dark) and 200 (bright) so normalization is never degenerate.
+    After normalizing to [0, 255], the dark pixels land near 0 and the bright
+    ones near 255, giving predictable starch classification at threshold=140.
+    """
+    img = np.zeros((size, size, 3), dtype=np.uint8)
+    split = int(size * dark_fraction)
+    img[:, :split] = 20     # dark → starch after normalisation
+    img[:, split:] = 200    # bright → no starch
+    cv2.imwrite(path, img)
+
+
+# ---------------------------------------------------------------------------
+# StarchArea integration
+# ---------------------------------------------------------------------------
+
+class TestStarchAreaIntegration:
+    def test_processimage_returns_rgb_image(self, tmp_path):
+        """_processImage must return an Image with the underlying array set."""
+        img_path = str(tmp_path / "apple_fruit_01.png")
+        _write_synthetic_png(img_path, bgr_value=(30, 30, 30))
+
+        analysis = StarchArea()
+        analysis._preRun()
+        result = analysis._processImage(RGBImage(img_path))
+
+        assert result is not None
+        assert result.getImage() is not None
+
+    def test_processimage_result_has_rating_value(self, tmp_path):
+        """The result Image must expose a 'rating' FloatValue after processing."""
+        img_path = str(tmp_path / "apple_fruit_01.png")
+        _write_synthetic_png(img_path, bgr_value=(80, 80, 80))
+
+        analysis = StarchArea()
+        analysis._preRun()
+        result = analysis._processImage(RGBImage(img_path))
+
+        assert "rating" in result.getMetaData(), "result is missing 'rating' key"
+        rating = result.getValue("rating").getValue()
+        assert 0.0 <= rating <= 1.0, f"rating {rating} is out of [0, 1] range"
+
+    def test_processimage_result_has_all_scale_indices(self, tmp_path):
+        """The result must include a starch index for every variety in StarchScales."""
+        img_path = str(tmp_path / "apple_fruit_01.png")
+        _write_mixed_starch_png(img_path)
+
+        analysis = StarchArea()
+        analysis._preRun()
+        result = analysis._processImage(RGBImage(img_path))
+
+        varieties = {k for k in vars(StarchScales) if not k.startswith("_")}
+        metadata_keys = set(result.getMetaData().keys())
+        missing = varieties - metadata_keys
+        assert not missing, f"Missing starch scale indices: {missing}"
+
+    def test_processimage_higher_threshold_gives_higher_rating(self, tmp_path):
+        """Higher threshold should classify more pixels as starch on the same image."""
+        img_path = str(tmp_path / "apple_fruit_01.png")
+        _write_mixed_starch_png(img_path, dark_fraction=0.5)
+
+        low_analysis = StarchArea()
+        low_analysis.starch_threshold.setValue(80)
+        low_analysis._preRun()
+        low_result = low_analysis._processImage(RGBImage(img_path))
+
+        high_analysis = StarchArea()
+        high_analysis.starch_threshold.setValue(200)
+        high_analysis._preRun()
+        high_result = high_analysis._processImage(RGBImage(img_path))
+
+        low_rating = low_result.getValue("rating").getValue()
+        high_rating = high_result.getValue("rating").getValue()
+        assert high_rating >= low_rating, (
+            f"Higher threshold should give higher starch ratio: low={low_rating}, high={high_rating}"
+        )
+
+    def test_processimage_result_image_same_shape_as_input(self, tmp_path):
+        """The annotated output image must have the same spatial dimensions as the input."""
+        img_path = str(tmp_path / "apple_fruit_01.png")
+        size = 64
+        _write_synthetic_png(img_path, bgr_value=(100, 100, 100), size=size)
+
+        analysis = StarchArea()
+        analysis._preRun()
+        result = analysis._processImage(RGBImage(img_path))
+
+        h, w, _ = result.getImage().shape
+        assert (h, w) == (size, size)
+
+    def test_processimage_preserves_image_name(self, tmp_path):
+        """getImageName() on the result should match the input filename."""
+        img_path = str(tmp_path / "apple_fruit_01.png")
+        _write_synthetic_png(img_path, bgr_value=(128, 128, 128))
+
+        analysis = StarchArea()
+        analysis._preRun()
+        result = analysis._processImage(RGBImage(img_path))
+
+        assert result.getImageName() == "apple_fruit_01.png"
+
+
+# ---------------------------------------------------------------------------
+# BlushColor integration
+# ---------------------------------------------------------------------------
+
+class TestBlushColorIntegration:
+    def _yellow_pear_image(self, path: str, size: int = 80) -> None:
+        """Write a yellow pear image: high B channel (fruit) with some A (blush)."""
+        # In LAB (OpenCV 0–255 encoding): L~128, A~148 (neutral), B~200 (yellow)
+        # We create a BGR image that converts to a known LAB range.
+        # A yellow-ish image: R≈200, G≈180, B≈50 in BGR → yellow pear background
+        img = np.full((size, size, 3), (50, 180, 200), dtype=np.uint8)
+        cv2.imwrite(path, img)
+
+    def _blush_pear_image(self, path: str, size: int = 80) -> None:
+        """Write an image where ~half the pixels will register as blush (high A channel)."""
+        img = np.zeros((size, size, 3), dtype=np.uint8)
+        # Left half: reddish (will have high A in LAB) — BGR (0, 80, 200)
+        img[:, :size // 2] = (0, 80, 200)
+        # Right half: yellow (low A, high B) — BGR (50, 180, 200)
+        img[:, size // 2:] = (50, 180, 200)
+        cv2.imwrite(path, img)
+
+    def test_processimage_returns_image(self, tmp_path):
+        img_path = str(tmp_path / "pear_fruit_01.png")
+        self._yellow_pear_image(img_path)
+
+        analysis = BlushColor()
+        analysis._preRun()
+        result = analysis._processImage(RGBImage(img_path))
+
+        assert result is not None
+        assert result.getImage() is not None
+
+    def test_processimage_result_has_rating_value(self, tmp_path):
+        img_path = str(tmp_path / "pear_fruit_01.png")
+        self._yellow_pear_image(img_path)
+
+        analysis = BlushColor()
+        analysis._preRun()
+        result = analysis._processImage(RGBImage(img_path))
+
+        assert "rating" in result.getMetaData()
+        rating = result.getValue("rating").getValue()
+        assert 0.0 <= rating <= 1.0, f"rating {rating} out of range"
+
+    def test_processimage_result_image_same_shape(self, tmp_path):
+        size = 60
+        img_path = str(tmp_path / "pear_fruit_01.png")
+        img = np.full((size, size, 3), (50, 180, 200), dtype=np.uint8)
+        cv2.imwrite(img_path, img)
+
+        analysis = BlushColor()
+        analysis._preRun()
+        result = analysis._processImage(RGBImage(img_path))
+
+        h, w, _ = result.getImage().shape
+        assert (h, w) == (size, size)
+
+    def test_processimage_preserves_image_name(self, tmp_path):
+        img_path = str(tmp_path / "pear_fruit_01.png")
+        self._yellow_pear_image(img_path)
+
+        analysis = BlushColor()
+        analysis._preRun()
+        result = analysis._processImage(RGBImage(img_path))
+
+        assert result.getImageName() == "pear_fruit_01.png"
+
+    def test_processimage_higher_blush_image_gives_higher_rating(self, tmp_path):
+        """An image with more reddish pixels should produce a higher blush rating."""
+        low_path = str(tmp_path / "low_blush_fruit_01.png")
+        high_path = str(tmp_path / "high_blush_fruit_01.png")
+        self._yellow_pear_image(low_path)
+        self._blush_pear_image(high_path)
+
+        analysis = BlushColor()
+        analysis._preRun()
+        low_result = analysis._processImage(RGBImage(low_path))
+        high_result = analysis._processImage(RGBImage(high_path))
+
+        low_rating = low_result.getValue("rating").getValue()
+        high_rating = high_result.getValue("rating").getValue()
+        assert high_rating >= low_rating, (
+            f"Expected blush image to score higher: low={low_rating}, high={high_rating}"
+        )