Fix random choice of coordinates within true values of a binary mask #256
Conversation
Codecov Report
@@ Coverage Diff @@
## master #256 +/- ##
=========================================
Coverage 100.00% 100.00%
=========================================
Files 17 17
Lines 1203 1206 +3
Branches 114 114
=========================================
+ Hits 1203 1206 +3
Continue to review full report at Codecov.
|
| def random_choice_true_mask2d(binary_mask: np.ndarray) -> Tuple[int, int]: | ||
| """Return a random pair of indices where the ``binary_mask`` is True. | ||
| Parameters | ||
| ---------- | ||
| binary_mask : np.ndarray | ||
| Binary array. | ||
| Returns | ||
| ------- | ||
| Tuple[int, int] | ||
| Random pair of indices where the ``binary_mask`` is True. | ||
| """ | ||
| x = np.random.choice(np.where(binary_mask)[0]) | ||
| y = np.random.choice(np.where(binary_mask[x])[0]) | ||
|
|
||
| return x, y |
There was a problem hiding this comment.
I don't see any reason of a new util functions that is used basically in one single place. If in future we need this in more than 1 code portion so we can introduce it in the util module.
There was a problem hiding this comment.
alright, should it be a private static method of RandomTiler then?
There was a problem hiding this comment.
for 2 lines and then used in only one part in the randomtiler? i don't think so just copy and paste that 2 lines when you need those :)
There was a problem hiding this comment.
mmm not so easily testable in that case
There was a problem hiding this comment.
why? you could check the call args and assert that the y has the correct value you are expecting and that depends on the x.
scale_coordinates(
reference_coords=CoordinatePair(x_ul_lvl, y_ul_lvl, x_br_lvl, y_br_lvl),
reference_size=binary_mask.shape[::-1],
target_size=slide.dimensions,
)
during the test you can assert that the scale_coordinates has been called with the right x_ul_lvl and y_ul_lvl .. do you?
There was a problem hiding this comment.
in the test I want to check that the coordinates I sampled correspond to a True value in the mask, and I won't be checking that with your proposal...
| _scale_coordinates.assert_called_once_with( | ||
| reference_coords=CP(x_ul=0, y_ul=0, x_br=128, y_br=128), | ||
| reference_size=(500, 500), |
There was a problem hiding this comment.
i don't understand why this is not enough?
this
_random_choice_true_mask2d.return_value = (0, 0)
is basically asserting that x and y are 0 and so this line asserts the same thing i believe
There was a problem hiding this comment.
here I assert the fact that starting from (0,0) - which I know because I mocked random_choice_true_mask2d - I get to (128,128) via the scaling of the tile dimensions and then the sum.. but I'm not really testing if the coordinates UL are correct
There was a problem hiding this comment.
oh i see they're random that's the point. I don't know maybe the integration tests are enough to cover this .... but if you feel more safe this way, i can buy that. So... just let me know and I'll stamp the PR
There was a problem hiding this comment.
I would prefer to keep the unit test as it is, via the integration tests we have we never noticed (visually), probably due to a lucky seed
fixes #246