[REF] Replace MAPCA code with mapca library

[tsalo]

Closes #698.

Changes proposed in this pull request:

• Remove MA-PCA-related code from tedana.
• Use new mapca library (once it's released on PyPi).

[jbteves]

This is GREAT! Thanks so much for everyone who moved it into its own package, great job!

[notZaki]

The output might be different because mapca performs normalization along temporal-domain, while the implementation in tedana is along spatial-domain. At least, I think that's the case.

The license might also be worth discussing: mapca is currently licensed under GPLv2, and I think that merging it will require tedana to also change to GPL, and then any downstream dependencies such as fmriprep, would similarly have to change to GPL in order to use the next release of tedana.

[eurunuela]

Changes look good, thanks @tsalo ! Something is wrong with the 3 echo test though.

[emdupre]

The license might also be worth discussing: mapca is currently licensed under GPLv2, and I think that merging it will require tedana to also change to GPL, and then any downstream dependencies such as fmriprep, would similarly have to change to GPL in order to use the next release of tedana.

Changing the license is really a non-starter for both fMRIPrep and tedana, since it requires getting all existing contributors to the packages to re-consent to have their code release under the new license.

[tsalo]

The license might also be worth discussing: mapca is currently licensed under GPLv2, and I think that merging it will require tedana to also change to GPL, and then any downstream dependencies such as fmriprep, would similarly have to change to GPL in order to use the next release of tedana.

I don't believe licensing impacts dependencies (i.e., I can have any licensed code as a dependency, but when including the actual code in my library it does matter). If anything, it was just a problem that we had MAPCA code within tedana before.

[jbteves]

The good news is that tedana is already GPLv2.1, so we don't have to change anything.
Per my understanding of the licensing: fMRIPrep can include tedana provided that there is a notice that tedana code is licensed under GPL. Since we're packaged on PyPI with a license notice both in the repository and on PyPI, they should be fine. They would run into trouble only if they started distributing our code with a different license (which they don't). If you're curious, they use the BSD license (allows derivative works without assent).

[notZaki]

I don't believe licensing impacts dependencies (i.e., I can have any licensed code as a dependency, but when including the actual code in my library it does matter). If anything, it was just a problem that we had MAPCA code within tedana before.

True, the situation is kinda sticky with the current ma_pca code already in tedana.

The good news is that tedana is already GPLv2.1, so we don't have to change anything.

My understanding was that tedana uses L-GPL, which does not have the same restrictions as GPL [1] [2].

Based off this table/matrix it's ok to use a GPL library (mapca) inside L-GPL code (tedana) but the combination must be relicensed under the GPL (without the L).

[codecov]

Codecov Report

Merging #641 (eb0047f) into main (eec1a85) will decrease coverage by 0.60%.
The diff coverage is 100.00%.

@@            Coverage Diff             @@
##             main     #641      +/-   ##
==========================================
- Coverage   93.70%   93.09%   -0.61%     
==========================================
  Files          26       25       -1     
  Lines        2032     1782     -250     
==========================================
- Hits         1904     1659     -245     
+ Misses        128      123       -5     

Impacted Files	Coverage Δ
tedana/decomposition/__init__.py	100.00% <100.00%> (ø)
tedana/decomposition/pca.py	86.36% <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 eec1a85...eb0047f. Read the comment docs.

[eurunuela]

LGTM! Thanks @tsalo