Bug/fix slide bounds in TiffFile

[jonasteuwen]

The TiffFile had the axis swapped (height, width) instead of width, height. This fixes #187.

Also took the opportunity to change a few stylistic things.

[jonasteuwen]

I did, you can look at the test coverage On Nov 20, 2023, at 1:06 PM, Shannon Doyle ***@***.***> wrote: @sinberlin2 commented on this pull request. In dlup/backends/tifffile_backend.py:

@@ -47,9 +47,9 @@ def __parse_tifffile(self) -> None:

for idx, page in enumerate(self._image.pages): # Remove channel dimension and swap rows and columns if len(page.shape) == 3: - self._shapes.append((page.shape[1], page.shape[2])) + self._shapes.append((page.shape[1], page.shape[0])) Did you test this? In the documentation of tifffile it says that a Tiffpage has Shape ([depth,] length, width) https:/cgohlke/tifffile/blob/master/tifffile/tifffile.py#L7729 —Reply to this email directly, view it on GitHub, or unsubscribe.You are receiving this because you authored the thread.Message ID: ***@***.***>

[jonasteuwen]

I think this is needed because the array swaps it with respect to the image (x,y) versus (rows, cols)On Nov 20, 2023, at 1:45 PM, Shannon Doyle ***@***.***> wrote: @sinberlin2 commented on this pull request. In tests/backends/test_backends.py:

+ original_array = np.asarray(test_image)

+ properties = tiff_slide.properties + tile_size = (properties["tifffile.level[0].TileWidth"], properties["tifffile.level[0].TileLength"]) + num_levels = int(np.ceil(np.log2(np.asarray(size[::-1]) / np.asarray(tile_size))).min()) + 1 + + # Let's try to read outside of the slide levels. This should give a RuntimeError + with pytest.raises(RuntimeError): + tiff_slide.read_region((0, 0), num_levels + 1, (1, 1)) + + # We need to check a few regions to make sure the backend is working correctly + # 1. Check the whole image + # 2. Check the top right corner + # 3. Check the bottom left corner + + # Let's create a list of origins and sizes. + _size = size[::-1] why does the size order need to be reversed? Can we not just specify the size when we create the tiffile. It is also not really clear in what order the original size specification is. I would assume it is [w,h]. Then it only needs to be reversed when writing the tiffile. —Reply to this email directly, view it on GitHub, or unsubscribe.You are receiving this because you authored the thread.Message ID: ***@***.***>

[jonasteuwen]

To make it consistent with _image.pyOn Nov 20, 2023, at 1:51 PM, Shannon Doyle ***@***.***> wrote: @sinberlin2 commented on this pull request. In dlup/annotations.py:

@@ -757,8 +757,8 @@ def read_region(

Parameters ---------- - coordinates: np.ndarray or tuple - region_size : np.ndarray or tuple + location: np.ndarray or tuple why the change to location? —Reply to this email directly, view it on GitHub, or unsubscribe.You are receiving this because you authored the thread.Message ID: ***@***.***>

[jonasteuwen]

It is possible, I just want to avoid that accidentally I am reading a region which is by accident correct (like the top left corner will typically be)On Nov 20, 2023, at 1:35 PM, Shannon Doyle ***@***.***> wrote: @sinberlin2 commented on this pull request. In tests/backends/test_backends.py:

+class TestBackends:

+ def read_region_and_properties_asserts(self, tiff_slide, openslide_slide, mode, size, mpp, pyramid, test_image): + original_array = np.asarray(test_image) + properties = tiff_slide.properties + tile_size = (properties["tifffile.level[0].TileWidth"], properties["tifffile.level[0].TileLength"]) + num_levels = int(np.ceil(np.log2(np.asarray(size[::-1]) / np.asarray(tile_size))).min()) + 1 + + # Let's try to read outside of the slide levels. This should give a RuntimeError + with pytest.raises(RuntimeError): + tiff_slide.read_region((0, 0), num_levels + 1, (1, 1)) + + # We need to check a few regions to make sure the backend is working correctly + # 1. Check the whole image + # 2. Check the top right corner + # 3. Check the bottom left corner + Why is this the top right and bottom left corner? It seems to me that the code corresponds to 2. the top left corner and 3. the middle —Reply to this email directly, view it on GitHub, or unsubscribe.You are receiving this because you authored the thread.Message ID: ***@***.***>

[sinberlin2]

The tests work. I added minor comments about naming.