remove imresize

[NileGraddis]

addresses #660 #659

just a bugfix so master isn't broken :)

[njmei]

Wouldn't from skimage.transform import resize as imresize be easier and more straightforward?

[nicain]

test_lsn_image_to_screen covers lsn_image_to_screen (which your change depends on), but this test is not run on Appveyor/Travis. Can you check test_stimulus_info.py manually, or run a CI branch build?

[NileGraddis]

@njmei It's not quite a drop-in. The comparison is:

target_size = tuple( int(pixels_per_patch * dimsize) for dimsize in img.shape[::-1] )
img_full_res = np.array(Image.fromarray(img).resize(target_size, 0))

vs.

target_size = tuple( int(pixels_per_patch * dimsize) for dimsize in img.shape )
img_full_res = resize(img, target_size, order=0)

The second option is a bit pithier but also uses a different backend (imresize was a thin PIL wrapper). warp also appears to be a bit of a WIP:
https://github.com/scikit-image/scikit-image/blob/bcefaf5d2c2535c440610bb72b2794e1d249951d/skimage/transform/_warps.py#L817

Anyways I think the tradeoff is:

• improved readability (using resize)
• low chance of regressions / unexpected behavior (using PIL)

I don't have a strong opinion. Care to vote?

[NileGraddis]

@nicain I ran these tests manually. They pass

[njmei]

Oh interesting, I wonder why the scipy documentation suggested the skimage resize then...
https://docs.scipy.org/doc/scipy-1.1.0/reference/generated/scipy.misc.imresize.html#scipy.misc.imresize

I'll defer to your judgement!

[nicain]

OK, Looks good to me. @NileGraddis your call on Image vs skimage.resize

[nicain]

@NileGraddis Since this is part of the 0.16.3 release, can you update the changelog, what's new" section of index.rst, and Rebase the release candidate branch after merging (as per https://github.com/AllenInstitute/AllenSDK/blob/master/RELEASE_INSTRUCTIONS.md)?

Just did this, so don't worry about it.

[codecov-io]

Codecov Report

Merging #662 into master will decrease coverage by <.01%.
The diff coverage is 20%.

@@            Coverage Diff             @@
##           master     #662      +/-   ##
==========================================
- Coverage   44.94%   44.94%   -0.01%     
==========================================
  Files          92       92              
  Lines       11880    11880              
==========================================
- Hits         5340     5339       -1     
- Misses       6540     6541       +1

Impacted Files	Coverage Δ
allensdk/core/reference_space.py	93% <ø> (-0.07%)	⬇️
allensdk/brain_observatory/stimulus_info.py	78.58% <20%> (-0.18%)	⬇️

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update 966dbf7...4b524ff. Read the comment docs.

[codecov-io]

Codecov Report

Merging #662 into master will decrease coverage by <.01%.
The diff coverage is 20%.

@@            Coverage Diff             @@
##           master     #662      +/-   ##
==========================================
- Coverage   44.94%   44.94%   -0.01%     
==========================================
  Files          92       92              
  Lines       11880    11880              
==========================================
- Hits         5340     5339       -1     
- Misses       6540     6541       +1

Impacted Files	Coverage Δ
allensdk/core/reference_space.py	93% <ø> (-0.07%)	⬇️
allensdk/brain_observatory/stimulus_info.py	78.58% <20%> (-0.18%)	⬇️

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update 966dbf7...4b524ff. Read the comment docs.