Create an abstract Any for JSON/YAML

[j8r]

Merges duplications between YAML::Any and JSON::Any to a common Any abstract struct, which is inherited.

External shards implementing configurations formats will also be able to inherit from it.

Partially related to #5155

Note: this PR is a background work to implement this POC https://gist.github.com/j8r/2820157756ce55182c85667cb0456800 in a future PR - the ability to deal with dynamic keys:

js = JSON.parse %({"a": {"b": "c"}, "e": [{"o": "p"}, 12, 2]})

p js[["a", "r"]]?  #=> nil
puts js[["e", 0, "o"]] = JSON::Any.new "test"
puts js        #=> {"a" => {"b" => "c"}, "e" => [{"o" => "test"}, 12_i64, 2_i64]}
js.delete ["e",0, "o"]
puts js        #=> {"a" => {"b" => "c"}, "e" => [{}, 12_i64, 2_i64]}

[asterite]

@j8r Sorry, but these PRs should be discussed before implemented.

I personally agree with one of Go's proverb:

A little copying is better than a little dependency

Just because the code is similar, it shouldn't involve a dependency.

So a strong 👎 from me here.

[j8r]

Note that the API and implementations between the Anys aren't the same:
no YAML::Any#as_bool, YAML::Any#as_f32 and JSON::Any#as_bytes

In JSON::Any, there are as_i if @raw.is_a?(Int) and in YAML::Any @raw.as?(Int64).try &.to_i

[ysbaddaden]

I think we need more analysis and experiments on (de)serialization before conducting such changes. For example I'd love some attempts at a Serde like solution (from Rust).

[j8r]

I'll create a shard because for now dealing with dynamic mappings isn't possible in the stdlib

[ysbaddaden]

Yes, please experiment and return back your results, both successes and shortcomings :)

[j8r]

For information: a repository is created, https://github.com/j8r/dynany