Allow passing fs.FS when calling functions

[knqyf263]

Issue

Close #563

ToDo

• FSContext
• FSConfig
• FSKey
• Override fs.FS in WASI
• Add an example
• Unit test
• Example test

[codefromthecrypt]

taking a look now, thanks!

[codefromthecrypt]

Thanks for getting this started, as it helps me know where you are going. I nudged towards a specific direction, so that we can avoid exporting more api-visible symbols and get the same result.

[codefromthecrypt]

ping me when you are ready for the next round. I'll keep a look out

[knqyf263]

@codefromthecrypt I think I finished reflecting your comments. There are a few concerns though.

1. I've added a test for WithFS to PathOpen only.
   • I'm not sure if it is enough or you may prefer another approach
2. FSConfig is still embedded into ModuleConfig
   • Do you think it should be extracted?

[codefromthecrypt]

I wanted to reply quickly to you about some confusion I caused. I apologize for wasting some of your time. In general this looks good.

The main test needed would be to ensure that the context FS overrides vs adds to the one you expect. So one neat way to prove that is to have a test with the base FS having two files and the overlay having only one with a different content.

Meanwhile once done experimenting, maybe roll back the changes to the examples/wasi as we don't put experiemental things in that dir.

Finally, thanks again, I think after the adjustments here, this is getting to polish stage!

[codefromthecrypt]

@mathetake I think this is on the right track. I'm going to be mostly away this evening until tomorrow. Can you make sure this gets past the finish line. Cheers!

[mathetake]

I think this is on the right track :D

[mathetake]

I'm good when you two feel good to go!

[knqyf263]

Meanwhile once done experimenting, maybe roll back the changes to the examples/wasi as we don't put experiemental things in that dir.

Reverted. 56714b8

[knqyf263]

The main test needed would be to ensure that the context FS overrides vs adds to the one you expect. So one neat way to prove that is to have a test with the base FS having two files and the overlay having only one with a different content.

I wanted to do that, but I didn't find a good place to add the test. Also, it appeared to require somewhat larger changes. I probably missed something.

BTW, please feel free to add any commits. Sometimes it's faster to write code than to explain it in writing.

[codefromthecrypt]

@knqyf263 thanks for all the work here. I'll be at the conference all day, but I can review and add commits to address any things, possibly today but more likely by tomorrow. In any case thanks for putting this together as you made it possible to land during a busy time 👍

[codefromthecrypt]

ps I started fiddling with this, but might not push commits until tomorrow. Main thing I noticed was tension in the packages and made internal/fs instead. There are some tests in sys_test.go that should move into there also. I'll do it!

[codefromthecrypt]

I landed the other PRs that would drift this, and ran out of time to finish this up (still at conference and should meet people :) ). I'll finish this soon

[codefromthecrypt]

I force pushed a squash of the preceding work because the merges were getting hairy. I'll add a commit later with the changes I mentioned above

[codefromthecrypt]

@mathetake I finished re-org, can you take a look and add a commit as needed? otherwise merge it.

[mathetake]

@knqyf263 thank you so much for working on this! 🏅 👍

[knqyf263]

@codefromthecrypt Thanks for your help!
I saw you added internal/fs package. It looks good!