Tolerate lowercase kubenetes kinds

[f4tq]

• Istio stresses the type system by using unexpected kind definitions as explained in issue 360.

[f4tq]

The tests that are failing also fail on a clean master.

Looks like it has to do with certs i.e test_config.rb:98.

[cben]

Thanks!
right, #359 is about the tests. I'd like to try fix them before merging any other changes.

[cben]

This ternary assumes we'll see either kind: CamelCase or kind: lowercase, right?
What if we get kind: "lowerCamelCase"?
Also, even if using a lowercase kind as-is works, it feels pretty accidental given how we look for capital letters in prefix...

The purpose of this weird prefix dance is automatically adapting to the plural ending, which is sometimes s and sometimes es (and sometimes nothing e.g. Endpoints=endpoints but we special-case those above).

Except it doesn't extract just the s / es — m[1] ends up with the whole lowercase end.
To take a fuller example:

kind = "OAuthAccessToken"
name = "oauthaccesstokens"
prefix = kind[0..kind.rindex(/[A-Z]/)]      #  "OAuthAccessT"
m = name.match(/^#{prefix.downcase}(.*)$/)  # #<MatchData "oauthaccesstokens" 1:"okens">
# Using new underscore algorithm from #355
underscore_entity(kind)                  # "o_auth_access_token"
underscore_entity(prefix)                # "o_auth_access_t"
underscore_entity(prefix) + m[1]         # "o_auth_access_tokens"

Can we rewrite the whole thing to work in a clearer way?

cc @eatwithforks

[f4tq]

All the ternary does is check if there are any upper case letters. If there are, do the original action. If there aren't, then take kind as is.

This way you don't get a negative index which was causing an uncaught exception (and crash).

There are 2 tests. This fix allows any kind, irregardless of casing, to make it through.

[cben]

Fair enough. This is a strict improvement. I'll merge this and see if I can rewrite it clearer separately...

[cben]

close-cycting to re-run tests after #364 test fix.