feat: Add support showing size of absent git-annex'ed files

[stv0g]

git-annex encodes the apparent size of files their symlinks.

gdu now is capable of extracting those sizes from the broken symlinks to calculate the total size of git-annex repositories.

Reference: https://git-annex.branchable.com/internals/key_format/

Note: real usage remains zero. gdu needs to be invoked via:

gdu --follow-symlinks --show-apparent-size --show-annexed-size

[stv0g]

I fully understand, if this PR is a bit out-of-scope for gdu. Maybe other git-annex users might find it useful :)

[codecov]

Codecov Report

Attention: Patch coverage is 79.16667% with 15 lines in your changes missing coverage. Please review.

Project coverage is 87.55%. Comparing base (a0fd828) to head (917a6ec).

Files with missing lines	Patch %	Lines
cmd/gdu/main.go	0.00%	4 Missing ⚠️
cmd/gdu/app/app.go	0.00%	2 Missing and 1 partial ⚠️
pkg/analyze/symlink.go	86.95%	2 Missing and 1 partial ⚠️
pkg/analyze/sequential.go	33.33%	2 Missing ⚠️
pkg/analyze/stored.go	0.00%	2 Missing ⚠️
internal/testanalyze/analyze.go	0.00%	1 Missing ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##           master     #404      +/-   ##
==========================================
- Coverage   87.64%   87.55%   -0.10%     
==========================================
  Files          44       46       +2     
  Lines        4329     4386      +57     
==========================================
+ Hits         3794     3840      +46     
- Misses        460      470      +10     
- Partials       75       76       +1     

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

[dundee]

I have fixed few small issues. Could you cover your logic with tests?

[stv0g]

Hi Daniel,

Yes, I can implement some test cases. The use of EvalSymlinks unfortunately, does not work for me, as it iteratively calls os.Readlink which break if any component of the link does not exist.

git-annex stores files in this hierarchy:

      ├─► .git                                                                                                                                                                                                                                                                                                                   
      │    └─► annex                                                                                                                                                                                                                                                                                                          
      │         └─► objects                                                                                                                                                                                                                                                                                              
      │              └─► 7d                                                                                                                                                                                                                                                                                        
      │                   └─► 86                                                                                                                                                                                                                                                                      
      │                        └─►SHA256E-s43632--7d865e959b2466918c9863afca942d0fb89d7c9ac0c99bafc3749504ded97730.jpg                                                                                                                                                                              
      │                             └─► SHA256E-s43632--7d865e959b2466918c9863afca942d0fb89d7c9ac0c99bafc3749504ded97730.jpg                                                                                                                                                                                        
      ├──► existing_file.jpg -> .git/annex/objects/7d/86/SHA256E-s43632--7d865e959b2466918c9863afca942d0fb89d7c9ac0c99bafc3749504ded97730.jpg                                                                                                                                                                          
      └──► missing_file.zip -> .git/annex/objects/08/d2/SHA256E-s744506832--08d2daced60b5eb6509044d5eefca82e7a6899350f49adc0083014229739515e.zip    (broken symlink)                                                                                                                                                                       

Due two-levels of hash directories, and a dedicated directory per link, causes EvalSymlinks to fail.

For the purpose of this PR, I am only interested in the -s<SIZE>-- fragment of the symlink.

[dundee]

I have updated the code to contain both EvalSymlinks (which is needed for proper handling of symlinks in subdirectories) and Readlink fot git-annex. Wdyt?

It seems to work for me fine.

$ go run github.com/dundee/gdu/v5/cmd/gdu -LAan ~/annex
  1.1   MiB /.git
@ 923.0 MiB file.mp4

 $ go run github.com/dundee/gdu/v5/cmd/gdu -an ~/annex
   1.1 MiB /.git
@  202 B file.mp4

Sorry for taking so long, I wanted to get hands on Git annex.

[stv0g]

Hey @dundee,

oh thats great :) Yes, I think this is the best solution.

Thanks for looking into it :) I did not expect, that you will look into git-annex :)

[stv0g]

I've also pushed a fix to make golangci-lint happy.

[dundee]

Great, thanks for your contribution!