Vb/pdgone

[vineetbansal]

Stripped off pandas from the rest of the codebase.

[codecov]

Codecov Report

Merging #911 (5c089b7) into vb/src_ascopy (e98d09d) will increase coverage by 0.18%.
The diff coverage is 92.82%.

@@                Coverage Diff                @@
##           vb/src_ascopy     #911      +/-   ##
=================================================
+ Coverage          88.18%   88.37%   +0.18%     
=================================================
  Files                119      119              
  Lines              10118    10448     +330     
=================================================
+ Hits                8923     9233     +310     
- Misses              1195     1215      +20     

Impacted Files	Coverage Δ
src/aspire/commands/cov3d.py	0.00% <0.00%> (ø)
src/aspire/commands/denoise.py	0.00% <0.00%> (ø)
src/aspire/commands/preprocess.py	0.00% <0.00%> (ø)
src/aspire/source/image.py	91.83% <95.23%> (+0.86%)	⬆️
src/aspire/storage/starfile.py	90.00% <95.45%> (+2.74%)	⬆️
src/aspire/commands/extract_particles.py	86.79% <100.00%> (ø)
src/aspire/ctf/ctf_estimator.py	98.00% <100.00%> (-0.02%)	⬇️
src/aspire/denoising/class_avg.py	74.41% <100.00%> (+0.19%)	⬆️
src/aspire/denoising/denoised_src.py	92.59% <100.00%> (+0.28%)	⬆️
src/aspire/source/coordinates.py	92.30% <100.00%> (+0.09%)	⬆️
... and 4 more

📣 We’re building smart automated test selection to slash your CI/CD build times. Learn more

[vineetbansal]

A lot of the acrobatics happening in this branch have to do with the fact that:

• some code is accessing the _metadata attribute directly and is thus tied to its implementation.
• It's a bit tricky to get the nth row of each value of a dict of lists, merge two such dicts etc, hence some code with dict comprehension, zip etc.
• The relion interop code was written to differentiate between scalar values and vectors on the basis of the incoming object being a DataFrame. Now that they're all dicts, one has to peek at the values.

Probably a good project for someone down the line would be to have a Metadata class to contain some of the ugliness here. On the other hand, the relion code is dealing with general-purpose 2D data and is not relying on it coming from metadata of an ImageSource, so gains might be limited.

I've submitted this as a separate PR so it can be merged after vb/src_ascopy is merged, or before - depending on how much disruption is acceptable, and the resources available to review and test these more thoroughly.

[garrettwrong]

Thanks @vineetbansal , tried running this against a known bad file (10028 shiny particles), and ASPIRE still seems to indicate it is bad, so I'm happy with that. I think this is probably ready for Joakim to also sign off on.

There were a few small things, but I think they are just questions and minor cosmetic followups we can do. Thanks!

[garrettwrong]

Looks like the redundant PR merge conflicted themselves.

[garrettwrong]

Docs build failed because EMDB ftp was offline yesterday ( and so we could not download our datafile). Its online now, restarting jobs.

[janden]

Much cleaner, although I confess there are some list comprehension acrobatics here that I have trouble following (in the aspire.source.* modules). Perhaps some can be simplified (or commented) but this is not too urgent to hold up a merge.

[janden]

Looks good.