diff: --hide-missing doesn't seem to work

[kalebo]

Bug Report

Description

The documentation seems to suggest that I can list only dvc files that have been explicitly checked out, ignoring any that don't exist in the local cache. This is would be very useful in my particular workflow as I am primarily doing partial checkouts of specific datasets in DVC. E.g., fetching only a specific directory of dvc files.

However, the issue is that --hide-missing doesn't seem to work. This was confirmed by Gao on the discord server.

Reproduce

• dvc add and push several files.
• git commit and push to a remote
• clone the git repo on a new machine or on the same machine after clearing the local cache.
• dvc pull a single specific file
• run dvc diff and note that all the unpulled files are marked as deleted
• run dvc diff --hide-missing and note that the diff is identical to the previous command

Expected

The expected behavior is that the is no output with --hide-missing as the unpulled files are hidden and the one explicitly pulled file has not been modified.

Environment information

DVC version: 2.9.4 (pip)
---------------------------------
Platform: Python 3.9.2 on Linux-5.10.0-10-amd64-x86_64-with-glibc2.31
Supports:
        webhdfs (fsspec = 2022.1.0),
        http (aiohttp = 3.8.1, aiohttp-retry = 2.4.6),
        https (aiohttp = 3.8.1, aiohttp-retry = 2.4.6),
        s3 (s3fs = 2022.1.0, boto3 = 1.21.7)
Cache types: hardlink, symlink
Cache directory: ext4 on /dev/mapper/vg0-scratch
Caches: local
Remotes: s3
Workspace directory: ext4 on /dev/mapper/vg0-scratch
Repo: dvc, git

[dberenbaum]

This looks like it was introduced in #7353 and is coming from changes to the if statement in:

dvc/dvc/repo/diff.py

Lines 187 to 199 in c81d841

	def _filter_missing(repo_fs, paths):
	for path in paths:
	try:
	info = repo_fs.info(path)
	dvc_info = info.get("dvc_info")
	if (
	dvc_info
	and info["type"] == "directory"
	and not dvc_info["meta"].obj
	):
	yield path
	except FileNotFoundError:
	pass

There are a couple different issues here:

1. Under the logic introduced in dvcfs: detach from pipeline outputs #7353, it's only possible for directories to be considered missing.
2. Even directories appear not to be working as expected:

$ git clone [email protected]:iterative/example-get-started.git
$ cd example-get-started
$ dvc diff --hide-missing
Deleted:
    data/data.xml
    data/features/
    data/prepared/
    model.pkl

files summary: 2 deleted

@efiop Any thoughts on this?