Add `opam tree` subcommand to display a dependency tree

[g2p]

opam query only shows flat results.

opam tree, similar to cargo tree in Rust, would display a dependency tree. Here's how it looks like in Rust:

cargo tree
postgres v0.10.2 (file:///Volumes/git/rust/rust-postgres)
├── bufstream v0.1.1
├── byteorder v0.4.2
├── hex v0.1.0
├── log v0.3.4
│   └── libc v0.2.4
├── net2 v0.2.20
│   ├── cfg-if v0.1.0
│   ├── kernel32-sys v0.2.1
│   │   └── winapi v0.2.5
│   ├── libc v0.2.4 (*)
│   ├── winapi v0.2.5 (*)
│   └── ws2_32-sys v0.2.1
│       └── winapi v0.2.5 (*)
└── phf v0.7.9
    └── phf_shared v0.7.9

Note the Unicode-art.

Items marked with (*) are later instances of common subtrees; the subtrees are skipped instead of being printed another time.

Useful options: --duplicate

This displays a forest of inverted trees, starting at every duplicate dependency, towards the root. This is useful for examining how dependency versions get constrained.

[cdaringe]

i often have install deps that fail, and i want this feature to see who is calling for it. perhaps the parent node has bad ranges, needs to be pinned, etc.

[cdaringe]

npm ls / yarn why are additional prior art that emit reduced graphs with the requested package as the leaf, and print up the dep tree to the root. super helpful

[vphantom]

Is there any way to get that kind of information, albeit not as nicely for now? I experimented with opam lock and I was surprised by some of the dependencies of my little project. I tried to track some down on https://opam.ocaml.org/packages/ but for some of them I'm stumped. (Like cppo, I'd expect as a build dependency but it's a normal dependency. No idea who needs it.)

(To clarify: I know that opam-query exists but it is incompatible with OCaml 4.08+ because of a dependency on an old version of containers. So I'm looking for other ways to get the information.)

[cannorin]

I'm working on it and got this so far:

I'll open a PR once I complete docs and add adequate tests.

[avsm]

Your opam why output looks absolutely awesome. I want this in opam! :-)

[dbuenzli]

@cannorin if you have some time to spend on this I'd suggest to look at this paper which seems to have interesting techniques for ASCII DAG visualisations.

[cannorin]

@dbuenzli I've just read the paper. Looks interesting for sure, but it would need some additional work to allow annotating edges, which would be needed to show the constraints (e.g. (>= 4.0.0)). Also, we would need to try to make the graph as less wide as possible (since we are working with CLI), but the paper only mentions about it as a future work.