optimize and deduplicate tree about arborist HOT 2 CLOSED

Interesting question as to how far we want to go with this.

One simple approach would be:

• Get the list of nodes grouped by name.
• For every name with more than 1 instance, get the set of edges in
• Starting from the deepest in the tree, re-attempt to placeDep, with preferDedupe: true enabled.
• If it ends up in the same place again, re-attempt with the other nodes in the set.
• Repeat until every node in the set is unchanged.

I think it might be surprising if npm dedupe fetched new packages from the registry, so it should strictly be placing, not fetching.

A more dramatic and powerful approach would be to implement a variant of pubgrub, falling back to nesting when a conflict can be solved that way. It still won't be able to resolve all peer dep conflicts (eg, root module depends on X and Y which peer depend on incompatible versions of Z), but it would result in much more flattening in any non-pathological use case. Of course, doing that implies:

• Lots more fetching of packuments, (as opposed to just fetching the manifest which is the best match for a given spec)
• A dedupe that must reach out to the network.
• Figuring out how pubgrub would interact with git deps, remote deps, etc. In those cases, I think we'd have to fetch and unpack potentially a lot of stuff. (Imagine a package X where each version depends on a git dep Y with a different hash, or a remote dep Z with a different target url -- we'd have to fetch them all.)

I think as a separate command (or something like npm dedupe --extra-flat or something), it could work, but it would not be network efficient enough for the default install.

from arborist.

Better expressed in #66 and #67. If we wanna do something fancier, we can explore that in a future version.

from arborist.