Rename experimental JSON tests.

[bdice]

Description

This PR renames the "experimental" JSON reader tests. These are now production grade and not experimental.

This task is tracked in #15537.

Checklist

• I am familiar with the Contributing Guidelines.
• New or existing tests cover these changes.
• The documentation is up to date with these changes.

[bdice]

@karthikeyann / @vuule: Could one of you help me understand this line?

cudf/cpp/tests/io/nested_json_test.cpp

Line 624 in 7b9e815

INSTANTIATE_TEST_SUITE_P(Experimental, JsonParserTest, testing::Bool());

This boolean parameter labeled "Experimental" appears to control whether we call device_parse_nested_json or host_parse_nested_json. Do we still need both of those code paths (with tests)? Should we rename this flag from Experimental to DeviceParse?

[vuule]

@karthikeyann / @vuule: Could one of you help me understand this line?

cudf/cpp/tests/io/nested_json_test.cpp

Line 624 in 7b9e815

INSTANTIATE_TEST_SUITE_P(Experimental, JsonParserTest, testing::Bool());

This boolean parameter labeled "Experimental" appears to control whether we call device_parse_nested_json or host_parse_nested_json. Do we still need both of those code paths (with tests)? Should we rename this flag from Experimental to DeviceParse?

IMO we can remove the host path, and thus this test. We now have a lot of tests of the new reader, I think we don't have to test this part of the implementation against the reference any more. If not, yes, renaming to DeviceParse sounds good to me
We do definitely need @karthikeyann's blessing to remove the test :)

[bdice]

Great. I renamed to IsFullGPU because that's how the variables are named in the tests. If @karthikeyann thinks we can remove the host parsing logic, I'll do that (and remove the tests) in a follow-up PR.

[bdice]

/merge

[karthikeyann]

Yes. We could remove the host parsing logic. It may not be in sync with device code with a lot of recent changes in JSON tokenizer, and tree algorithms. @elstehle shall we remove host parsing code and their tests?

[davidwendt]

Did you merge this with only one cpp review approval?

[bdice]

Did you merge this with only one cpp review approval?

Oops. @davidwendt I thought you were the second approver. My apologies.

[bdice]

@vuule and/or @karthikeyann feel free to take a look over this. I thought Vukasin had approved first and David was second, but I now see that I imagined the first ✅ where there was none. I hope we don’t need to revert but I’ll be happy to apply any follow-ups in another PR if needed.

[vuule]

How could you do this @bdice ?!
Well, no harm done, I would approve as-is. Just nobody tell Greg.