test_annotate more stable

Could we have a more stable test_annotate?

If you deal with M-pixels obviously the sum of the errors is high.

Basically it would be much more stable to compare pixels per pixels.

is_ok = numpy.alltrue(numpy.allclose(compare, copy, atol=1))

What do you think?

_____________________________ test_annotate[data1] _____________________________ data = {'beam': {'center': [2140082.632109303, -4231399.999999999], 'position': [2140199.9999999995, -4231399.999999999]}, 'd...': {'center': [2045848.0499999993, -4166239.9399999995], 'pixelsize': [-1497.6499999999976, -1551.4300000000046]}, ...} @pytest.mark.parametrize("data", (DATA, DATA2)) def test_annotate(data): orig_path = f"{DIR_PATH}/assets/downstream_{data['downstream']}.png" compare_path = f"{DIR_PATH}/assets/downstream_{data['downstream']}annotated.png" copy_path = f"/tmp/downstream{data['downstream']}_test.png"

    with Image.open(orig_path) as orig:
        copy = orig.copy()
        copy.save(copy_path)

    ann = AnnotateImage(copy_path)
    ann.annotate(
        data["image"]["center"],
        data["beam"]["position"],
        data["image"]["pixelsize"],
        1e-9,
        data["subsample"],
    )

    with Image.open(compare_path) as compare:
        with Image.open(copy_path) as copy:
            img_diff = ImageChops.difference(compare, copy)
            diff = np.array(img_diff)

          assert diff.sum() == approx(0)

E assert 17520 == 0 ± 1.0e-12 E comparison failed E Obtained: 17520 E Expected: 0 ± 1.0e-12