Closes #720

As I said, this is on hold until we start working on 2.2. In the mean time:

1. Please add tests (including for list and mixing of lists and dicts).
2. Please add a news fragment
3. Please update the documentation.

This pull request introduces 1 alert when merging 49b9d46 into 6c732c1 - view on LGTM.com

new alerts:

• 1 for Unused local variable

@omry @Jasha10 I have a gap at the docs session, I've added a segment under docs/source/usage.rst for sphinx and seems like I'm missing something.
Would like to get a hand here 😀

From the doctest output:

File "usage.rst", line 791, in default
Failed example:
    OmegaConf.missing_keys({
        "foo": {"bar": "???"},
        "missing": "???",
         "list": ["a", None, "???"]
    })
Expected:
    {'list.2', 'foo.bar', 'missing'}
Got:
    {'missing', 'foo.bar', 'list.2'}

The elements of a set are not guaranteed to be printed in any particular order...
Here is something that should make the tests pass, though it is sacrifices clarity for the reader:

>>> sorted(OmegaConf.missing_keys({
...     "foo": {"bar": "???"},
...     "missing": "???",
...     "list": ["a", None, "???"]
... }))
['foo.bar', 'list.2', 'missing']

From the doctest output:

File "usage.rst", line 791, in default
Failed example:
    OmegaConf.missing_keys({
        "foo": {"bar": "???"},
        "missing": "???",
         "list": ["a", None, "???"]
    })
Expected:
    {'list.2', 'foo.bar', 'missing'}
Got:
    {'missing', 'foo.bar', 'list.2'}

The elements of a set are not guaranteed to be printed in any particular order...
Here is something that should make the tests pass, though it is sacrifices clarity for the reader:

>>> sorted(OmegaConf.missing_keys({
...     "foo": {"bar": "???"},
...     "missing": "???",
...     "list": ["a", None, "???"]
... }))
['foo.bar', 'list.2', 'missing']

Yeah it does sacrifice readability, I wonder if there is a way around it to make it verify the output rather than the output's representation.

Yeah it does sacrifice readability, I wonder if there is a way around it to make it verify the output rather than the output's representation.

assert ret == {'list.2', 'foo.bar', 'missing'}

Please a news fragment file. This is similar to how it's done for Hydra.
Context here.

I've added news/720.feature. Let me know if something is wrong/missing.

I think we can now revisit this PR since work on OmegaConf 2.2 has started.

Made a merge commit to resolve conflicts with master branch.

Merged.
Thanks again @Ohad31415!