feat(path)!: consolidated path libraries

[hacdias]

This is the long-awaited consolidation of the multiple path libraries we've had, which closes #198. For more information, see the changelog in the PR.

Kubo PR: ipfs/kubo#10063.

[codecov]

Codecov Report

Merging #334 (9252a2b) into main (b388690) will increase coverage by 0.03%.
The diff coverage is 71.54%.

@@            Coverage Diff             @@
##             main     #334      +/-   ##
==========================================
+ Coverage   65.69%   65.72%   +0.03%     
==========================================
  Files         209      207       -2     
  Lines       25196    25131      -65     
==========================================
- Hits        16552    16518      -34     
+ Misses       7177     7151      -26     
+ Partials     1467     1462       -5     

Files	Coverage Δ
gateway/assets/assets.go	69.62% <100.00%> (ø)
gateway/errors.go	86.48% <100.00%> (-0.36%)	⬇️
gateway/gateway.go	85.60% <ø> (+3.71%)	⬆️
gateway/handler_block.go	69.76% <100.00%> (ø)
gateway/handler_car.go	75.42% <100.00%> (ø)
gateway/handler_defaults.go	31.76% <100.00%> (ø)
gateway/handler_tar.go	79.36% <100.00%> (ø)
gateway/handler_unixfs_file.go	73.01% <100.00%> (ø)
namesys/base.go	88.60% <ø> (ø)
namesys/republisher/repub.go	57.84% <100.00%> (ø)
... and 19 more

... and 12 files with indirect coverage changes

[hacdias]

@ipfs/kubo-maintainers I think this is something important to get out of the door for many reasons (see #198). I just rebased this work I did in June on top of the recent changes. I can only imagine this having large repercussions thorough our other packages, especially Kubo. Therefore, before continuing, I would like to ask for some feedback regarding the Path interface.

One of the biggest changes is the removal of the IsValid() bool function from the path type. The idea is that if a path exists, then it must be valid. There's no point on passing around a path that will be invalid and always calling for IsValid(). Now the verification is done at creation time. This is what scares me the most when bubbling this to Kubo, as processes that would otherwise fail later in the process will now fail earlier. Or maybe not the case.

[Jorropo]

I shouldn't look at the implementation before the API so the code review is half done.

I don't understand what is the difference between ResolvedPath and ImmutablePath ?

[aschmahmann]

Thank you for starting this 🙏. The current state of the path libraries is .... too much

[hacdias]

@aschmahmann @lidel I want to ping you for a re-review. I've already addressed the feedback, tests are passing here and in Kubo. This PR is almost turning 4 months old and I really don't want that to happen.

[lidel]

Some feedback inline, mostly about making interface more clear, not tied to our namespaces (allowing people to use boxo with /foo without having to fork) and avoiding cognitive noise caused by legacy decisions in Kubo.

To avoid making waves this difficult month, I think we should minimize risk and ship this in Kubo 0.24, and we should wait with merge until 0.23 ships.

[hacdias]

To avoid making waves this difficult month, I think we should minimize risk and ship this in Kubo 0.24, and we should wait with merge until 0.23 ships.

It already has: https://github.com/ipfs/kubo/releases/tag/v0.23.0!

addressed

[hacdias]

I am going to go ahead and merge this. I think it has marinated for enough time and I've already addressed the feedbacks we had. Tests are passing here and in Kubo, and Kubo 0.23 has been released. So merging this at the beginning of the new release cycle (0.24) will allow us to test it until the final release of 0.24.