Dvc: status: implement support for import-ed files

[Baranowski]

See https://github.com/iterative/dvc/pull/2889#discussion_r357950906 for context

[efiop]

Context: we do have status implemented for the DependencyREPO, but it is not tested and also doesn't really run because the stage is locked.

[plaintextview]

Overview

+1 for me as well; when my Data Scientists clone a project repo which uses a data-registry repo, running dvc status on the project repo returns the following:

$ dvc status
WARNING: DVC-file 'data.dvc' is locked. Its dependencies are not going to be shown in the status output.
data.dvc:
    changed outs:
        not in cache:    data

Repro

My data-registry repo:

.
├── .dvc
│   ├── .gitignore
│   └── config
├── README.md
├── projectA
│   ├── .gitignore
│   └── data.dvc
└── projectB
    ├── .gitignore
    └── data.dvc

Update projectA to use data managed by data-registry repo:

$ git clone [email protected]:path/to/projectA.git
$ cd projectA
$ dvc init
$ dvc import [email protected]:<the-data-repo.git> projectA/data
$ git add data.dvc
$ git commit -m "Using DVC!!"
$ git push

Another user clones the (updated) projectA repo, and checks dvc:

$ git clone [email protected]:path/to/projectA.git
$ cd projectA
$ dvc status
WARNING: DVC-file 'data.dvc' is locked. Its dependencies are not going to be shown in the status output.
data.dvc:
    changed outs:
        not in cache:    data

It's less than optimal UX, and I know I will get emails over this ; )

[pared]

repro script:

mkdir remote_repo repo

set -e
set -x

maindir=$(pwd)

pushd remote_repo
git init -q && dvc init -q

echo data >> data
dvc add data

git add -A
git commit -am "add data"

popd

pushd repo
git init -q && dvc init -q
dvc import $maindir/remote_repo data

dvc status

[EDIT]
at the last status we will get following output:

+ dvc status
WARNING: DVC-file 'data.dvc' is locked. Its dependencies are not going to
 be shown in the status output.
Data and pipelines are up to date.

Ideally, we should get information if import-ed stage is up-to-date with current state of the revision it has been import-ed from. @dmpetrov correct me if I am wrong, but I guess that if no revision has been provided, it should be compared to current master, right?

[efiop]

Need to research how git does it, maybe it has some caching mechanism inside. It would be nice to do a similar thing to make dvc status instant even for imported files.

[dmpetrov]

@pared thank you for the repro script! It would be even more helpful to add important outputs into it (and probably some comments with problem/expected outputs) :)

[pared]

@dmpetrov sorry for that. I provided some context, please check it out. :)

[fabiosantoscode]

I'm picking this one up!