Json string and boolean

[kasiabulat]

I added JsonString and JsonBoolean implementation.

addresses: #39922
cc: @joperezr @bartonjs @ericstj @ahsonkhan @terrajobst @JamesNK

[stephentoub]

The description here says "addresses: #39922", but this only adds a few of the types, and the PR doesn't actually allow for these to be used in any meaningful way. Should we wait to merge this until it's actually possible to construct a JsonDocument from these? Otherwise we're sort of adding dead code / public surface area.

[ericstj]

Should we wait to merge this until it's actually possible to construct a JsonDocument from these? Otherwise we're sort of adding dead code / public surface area.

This represents individually usable / testable chunks of the overall feature. We're OK with checking them in those chunks as we've committed to finishing the whole feature. If for some reason we cannot finish we'll pull out these chunks, but I don't forsee that happening.

[stephentoub]

usable / testable chunks

testable, yes. usable, no. But ok; it's fine if we're committed to either finishing it or backing this out.

[JamesNK]

Something to figure out is what happens when you call ToString on these types. Do you return the JSON representation or the .NET representation?

JsonBoolean b = new JsonBoolean(true);
b.ToString(); // "True" or "true"

That question becomes important around JsonString. Should the result from ToString be escaped and surrounded by double quotes, or should it be the .NET string? Originally Json.NET returned the quoted/escaped JSON representation and people didn't like it. Now ToString on all the JToken types return the JSON representation except JValue with a string, which just returns the internal string.

[benaadams]

@benaadams - you added that attribute to SortVersion here, along with a bunch of other places (dotnet/coreclr#21765), was it necessary?

It was so null checks (== null) would return to fast null checks rather than making a slower call to the overload

[kasiabulat]

Why is ReadonlySpan behaving so differently on those platforms? What would be the best way to solve it? (I can just exclude the test for Windows NETFX_x86_Release))

[ahsonkhan]

Test failure unrelated: https://github.com/dotnet/corefx/issues/24355
https://helix.dot.net/api/2019-06-17/jobs/9cef0714-71ec-4769-9a2a-0a8015257630/workitems/System.Net.NameResolution.Functional.Tests/console