[mih]

This can be a basis for implementing patch generation for diff.

This PR is the companion to datalad/datalad#6797 (which has been abandoned).

TODO

• Prevent adding duplicate entries to .gitmodules datalad#7088 is added a slowing workaround to GitRepo.__save_add_submodules() due to the lack of type-annotation ability in core. In order to preserve the speed when handling large numbers of subdatasets, we better replace GitRepo.__save_add_submodules() in here too.

[codecov]

Codecov Report

Merging #91 (520f64c) into main (35af205) will decrease coverage by 4.54%.
The diff coverage is 95.29%.

@@            Coverage Diff             @@
##             main      #91      +/-   ##
==========================================
- Coverage   93.43%   88.89%   -4.55%     
==========================================
  Files          30       32       +2     
  Lines        2210     2828     +618     
==========================================
+ Hits         2065     2514     +449     
- Misses        145      314     +169     

Impacted Files	Coverage Δ
datalad_next/patches/__init__.py	100.00% <ø> (ø)
datalad_next/patches/gitrepo.py	93.36% <93.36%> (ø)
datalad_next/patches/tests/test_gitrepo.py	100.00% <100.00%> (ø)
datalad_next/_version.py	45.39% <0.00%> (-54.61%)	⬇️
datalad_next/gitremote/datalad_annex.py	89.50% <0.00%> (+0.27%)	⬆️

---

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 35af205...520f64c. Read the comment docs.

[mih]

I am closing this PR after a year of nothingness. It can now join the original datalad/datalad#6797 in the afterlife.

The underlying issues remain, of course. Working on #361 I found different approaches to do this. Rather than an expensive patch (the approach here), I think a "change iterator" around git diff-index (and friends) would be a simpler approach.

The implementation that I am now abandoning here is already a half-step in that direction (git diff-files), and became the workhorse in gooey-status-light. The substantial speed differences compared to status in -core make it evident that going further in this direction is useful.