fix(cli): outputs unique list of grants for each dest

[kimskimchi]

Summary

Checklist

• Wrote appropriate unit tests - CLI: setup tests #1129
• Considered security implications of the change
• Updated associated docs where necessary
• Updated associated configuration where necessary
• Change is backwards compatible if it needs to be (user can upgrade without manual steps?)
• Nothing sensitive logged
• Commit message conforms to Conventional Commit

Related Issues

Resolves #1071

[BruceMacD]

You can get similar functionality to this using a map of maps without adding a new dependency here. People use maps with a boolean value as the key to get set functionality in Go (yes, it is a bit ugly haha).

gs := make(map[string]map[string]bool)
for _, g := range grants {
	privileges, ok := gs[g.Resource]
	if !ok {
		privileges = make(map[string]bool)
		gs[g.Resource] = privileges
	}

	privileges[g.Privilege] = true
}

Then for the privileges you can iterate through the keys in the nested privileges maps.

[kimskimchi]

Yea, it was a bit ugly and thought it would be re-used in a bunch of places in the future. Is it better to avoid this dependency?

[BruceMacD]

In my opinion yeah, I think most Go devs would understand something like the example

[kimskimchi]

so I was digging into this for a bit actually (should I implement myself VS use the library), maybe for a bit too long 😅 , and these were the reasons for my choice:

1. Set is a common struct that gets used a lot (in a lot of other languages), so most people understand the library itself
2. This dependency is a highly used one (https://github.com/deckarep/golang-set), thus heavily tested in major companies (Docker, 1Password, etc)
3. It was much easier to read (especially because mine was a nested set)
4. efficient: I end up coding very little (eg. print, which is also super useful to call when debugging)
5. Implementing functions for sets myself would be more prone to error vs using a widely used/tested library that's 8 yrs old >_<
6. and... most importantly: I wasn't convinced yet of the counter-arguments (including [0]), as the pros of using a library seemed to outweigh the cons

[0] https://stackoverflow.com/questions/34018908/golang-why-dont-we-have-a-set-datastructure

(apologies for long reply but this was my research 😛)