Go back to using --omitNull instead of --omitEmpty

[Gabriella439]

Fixes #86

The motivation of this is to more accurately model the Kubernetes
API semantics by not auto-omitting empty fields. This is because
a field set to Some ([] : List T) is not necessarily the same
as None (List T).

This makes the typical case a bit more verbose (more Somes), but
the difference to the ./examples shows that it's not too bad.

[arianvp]

Lgtm

[amarrella]

LGTM, thanks for this!

[PierreR]

The script that builds the example in dhall-kubernetes still uses --omitEmpty
https://github.com/dhall-lang/dhall-kubernetes/blob/master/scripts/build-examples.py#L49

@Gabriel439 Is it still correct or is it something that should be changed ?

[Gabriella439]

@PierreR: That script is dead code and is not used. I just forgot to delete it when I replaced that logic with Nix. I have a pull request up to remove the now-defunct scripts:

#118

[PierreR]

@Gabriel439 Thx

Looks like the nix script also uses the --omit-empty flag: https://github.com/dhall-lang/dhall-kubernetes/blob/master/nix/nixpkgs.nix#L80 ?

As an aside question, what's the best option for dhall-kubernetes between dhall-yaml and dhall-json}/bin/dhall-to-yaml ?

[Gabriella439]

@PierreR: I think both should be fine if all you use is the command-line tool. The main advantage of the dhall-yaml package is that it is based on a pure Haskell implementation of YAML instead of binding to a C library for YAML support, but it is a GPL-licensed. Functionally, they both should behave the same as far as the command-line interface goes.

I also have another pull request up to also remove the --omit-empty from the Nix code,
too: #119