Constant for common selectors

[willscott]

There is currently no way to easily get an ipld.Node version of a selector representing "give me the whole dag".
This is a very common request, and that it involves something like

	ssb := builder.NewSelectorSpecBuilder(basicnode.Prototype__Any{})
	ss := ssb.ExploreRecursive(selector.RecursionLimitNone(), ssb.ExploreUnion(
		ssb.Matcher(),
		ssb.ExploreAll(ssb.ExploreRecursiveEdge()),
	))
	ss.Node()

but there isn't documentation or examples of that anywhere leads to a lot of confusion.

[ribasushi]

@willscott @mvdan fwiw markets does put it behind a const-like function: https://pkg.go.dev/github.com/filecoin-project/go-fil-markets/shared#AllSelector

[ribasushi]

@willscott ... and it is different from your example... confusing 🤔

[willscott]

the difference is if you want to 'match' all nodes, or just 'cover' all nodes

see https://github.com/ipld/go-ipld-prime/blob/master/traversal/selector/matcher.go#L9-L18

[rvagg]

What's the practical difference between these two operations? matching vs covering? I don't grok the difference in how you'd consume these things as distinct operations. It seems to me that the AllSelector version is doing a dag walk, while the ExploreUnion + Matcher is ... addressing the entire dag? What are the usage patterns that make this distinction helpful?

[willscott]

I think the proof of "i did the walk to search for _x_" is the covering, and the actually returned things are the matching subset.

…

On Thu, Jun 3, 2021 at 6:41 PM Rod Vagg ***@***.***> wrote: What's the practical difference between these two operations? matching vs covering? I don't grok the difference in how you'd consume these things as distinct operations. It seems to me that the AllSelector version is doing a dag walk, while the ExploreUnion + Matcher is ... addressing the entire dag? What are the usage patterns that make this distinction helpful? — You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub <#171 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AADIRJZKMSCWX232GLDOLZTTRAVLZANCNFSM45FEMBQA> .

[ribasushi]

@willscott just checking in case I missed a thread elsewhere: has here been any movement on this?

[willscott]

@ribasushi there is: it's in #199

[mvdan]

I'd love it if we could get this over the finish line :) Another person is struggling to use go-car with selector nodes: ipld/go-ipld-adl-hamt#24 (comment)

[warpfork]

#199 did introduce some of these, and is now merged, so I think this can be marked closed as a success.