data/tests/core_tests/test_tiles.py

"""
Copyright 2021, Dana-Farber Cancer Institute and Weill Cornell Medicine
License: GNU GPL 2.0
"""

import numpy as np
import pytest

from pathml.core import HESlide, Tile, types


@pytest.fixture
def emptytiles():
    slidedata = HESlide("tests/testdata/small_HE.svs")
    return slidedata.tiles


@pytest.fixture
def tiles():
    """
    dict of adjacent tiles
    """
    tiles = []
    for i in range(2):
        for j in range(2):
            # create tile
            shape = (224, 224, 3)
            coords = (224 * i, 224 * j)
            name = f"{i}_{j}"
            masks = {str(k): np.random.randint(2, size=shape) for k in range(2)}
            labs = {
                "test_string_label": "testlabel",
                "test_array_label": np.array([2, 3, 4]),
                "test_int_label": 3,
                "test_float_label": 3.0,
            }
            image = np.random.random_sample(shape)
            tile = Tile(
                image=image,
                name=name,
                coords=coords,
                slide_type=types.HE,
                masks=masks,
                labels=labs,
            )
            # add to dict
            tiles.append(tile)
    return tiles


@pytest.fixture
def tilesnonconsecutive():
    """
    dict of nonconsecutive tiles
    """
    tiles = []
    for i in range(2):
        for j in range(2):
            # create tile
            shape = (224, 224, 3)
            coords = (224 * 2 * (i + 1), 224 * 2 * (j + 2))
            name = f"{i}_{j}"
            masks = {str(k): np.random.randint(2, size=shape) for k in range(2)}
            labs = {
                "test_string_label": "testlabel",
                "test_array_label": np.array([2, 3, 4]),
                "test_int_label": 3,
                "test_float_label": 3.0,
            }
            image = np.random.randint(low=0, high=255, size=shape, dtype=np.uint8)
            tile = Tile(
                image=image,
                name=name,
                coords=coords,
                slide_type=types.HE,
                masks=masks,
                labels=labs,
            )
            # add to dict
            tiles.append(tile)
    return tiles


@pytest.mark.parametrize(
    "incorrect_input", ["string", True, 5, [5, 4, 3], {"dict": "testing"}]
)
def test_init(tiles, tilesnonconsecutive, incorrect_input):
    # init from dict
    slidedata = HESlide("tests/testdata/small_HE.svs", tiles=tiles)
    assert (slidedata.tiles[0].image == tiles[0].image.astype(np.float16)).all()
    # init len
    assert len(slidedata.tiles) == 4
    # incorrect input
    with pytest.raises(AssertionError):
        # fix obj
        slidedata = HESlide("tests/testdata/small_HE.svs", tiles=incorrect_input)
    # nonconsecutive tiles
    slidedata = HESlide("tests/testdata/small_HE.svs", tiles=tilesnonconsecutive)
    np.testing.assert_array_equal(
        slidedata.tiles[(896, 1344)].image, tilesnonconsecutive[3].image
    )


def test_repr(tiles):
    slidedata = HESlide("tests/testdata/small_HE.svs", tiles=tiles)
    assert repr(slidedata.tiles)


@pytest.mark.parametrize(
    "incorrect_input", ["string", None, True, 5, [5, 4, 3], {"dict": "testing"}]
)
@pytest.mark.parametrize(
    "incorrect_input2", [None, True, [5, 4, 3], {"dict": "testing"}]
)
def test_add_get(emptytiles, tileHE, incorrect_input, incorrect_input2):
    # add single tile
    tiles = emptytiles
    tiles.add(tileHE)
    # get by coords and by index
    assert (tiles[(1, 3)].image == tileHE.image).all()
    assert (tiles[0].image == tileHE.image).all()
    assert tiles[(1, 3)].name == tileHE.name
    assert tiles[(1, 3)].coords == tileHE.coords
    for label in tiles[(1, 3)].labels:
        if isinstance(tiles[(1, 3)].labels[label], np.ndarray):
            np.testing.assert_array_equal(
                tiles[(1, 3)].labels[label], tileHE.labels[label]
            )
        else:
            assert tiles[(1, 3)].labels[label] == tileHE.labels[label]
    assert tiles[(1, 3)].slide_type == tileHE.slide_type
    # get masks
    for mask in tiles.h5manager.h5["masks"].keys():
        # masks by coords and by index
        assert (tiles[(1, 3)].masks[mask] == tileHE.masks[mask]).all()
        assert (tiles[0].masks[mask] == tileHE.masks[mask]).all()
    # incorrect input
    with pytest.raises(ValueError or KeyError):
        tiles.add(incorrect_input)
    with pytest.raises(KeyError or IndexError):
        tiles[incorrect_input2]
    # wrong shape
    im = np.arange(np.product((225, 224, 3))).reshape((225, 224, 3))
    wrongshapetile = Tile(image=im, coords=(4, 5), name="wrong")
    with pytest.raises(ValueError):
        tiles.add(wrongshapetile)


@pytest.mark.parametrize(
    "incorrect_input", ["string", None, True, 5, [5, 4, 3], {"dict": "testing"}]
)
def test_remove(emptytiles, tileHE, incorrect_input):
    tiles = emptytiles
    tiles.add(tileHE)
    tiles.remove((1, 3))
    with pytest.raises(Exception):
        tiles[(1, 3)]
    with pytest.raises(KeyError):
        tiles.remove((1, 3))
    # incorrect input
    with pytest.raises(KeyError):
        tiles.remove(incorrect_input)