Multislice compSPI fft

[geoffwoollard]

Addresses part of https://github.com/compSPI/ioSPI/issues/45

[geoffwoollard]

@ninamiolane @fredericpoitevin what’s going on in these linting errors? they pass my hooks when I commit locally… do I run black . FILE.py or black . FILE.py --check or are there some special flags. It’s seems there’s some difference between my local black and git black.

Locally I get

(simSPI_d) dhcp-128-189-220-173:tests gw$ black . --check
All done! ✨ 🍰 ✨
11 files would be left unchanged.

[fredericpoitevin]

@geoffwoollard it looks like your branch was not up-to-date with current master when you started working on it, the "build" workflow is not used anymore. Try to merge master in your branch, resolve possible conflicts and push again, please. This should update this PR and trigger the new workflows. Thanks

[geoffwoollard]

@geoffwoollard it looks like your branch was not up-to-date with current master when you started working on it, the "build" workflow is not used anymore. Try to merge master in your branch, resolve possible conflicts and push again, please. This should update this PR and trigger the new workflows. Thanks

Oops, I thought I rebased, but this must have been in ioSPI or compSPI since they are all involved with removing the numpy fft code. OK I'll do that.

[geoffwoollard]

@fredericpoitevin I don't know about that. It looks like nothing changes with build

https://github.com/geoffwoollard/simSPI/compare/multislice_compspifft...compSPI:master

[geoffwoollard]

OK I merged in compSPI/simSPI:master and am still getting the same linting error.

Why is the linting and building not split up as it is in newer PRs?

Maybe I'll try to close this PR and open another?

[codecov[bot]]

Codecov Report

Merging #66 (9877fd7) into master (9c62f20) will increase coverage by 0.06%. The diff coverage is 100.00%.

@@            Coverage Diff             @@
##           master      #66      +/-   ##
==========================================
+ Coverage   97.80%   97.85%   +0.06%     
==========================================
  Files          10       10              
  Lines         362      372      +10     
==========================================
+ Hits          354      364      +10     
  Misses          8        8              

Impacted Files	Coverage Δ
simSPI/multislice.py	100.00% <100.00%> (ø)

---

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 9c62f20...9877fd7. Read the comment docs.

[fredericpoitevin]

Oh I saw build and assumed it was the workflow name but it was already the lint workflow, sorry. Well it does not hurt to check.

[fredericpoitevin]

If you want to make sure you test with the same version as GitHub Actions, check .github/workflows/lint.yml And see how things are pip installed. I built a simSPI-test conda environment locally using dev-requirements.txt - maybe consider using it to lint locally?

[geoffwoollard]

I turned off the precommit hook, since it's black was conflicting with my local black.

Now the linting is passing on github.

Not sure why that happened, and I need to resolve it before I turn on the precommit hook again.

In anycase this PR should be good to review.

There's some back and forth pytorch and numpy things happening. But that's because the compSPI fft functions need torch tensors that are shaped (1,1,n_pixels,n_pixels)

[geoffwoollard]

merging now since https://github.com/compSPI/ioSPI/pull/49 broke simSPI (since simSPI used fourier in ioSPI), and this PR will fixes that.