Quench nt test userwarn

[ghost]

Fed up with seeing Serializer Userwarnings in the test output. Is the latin-1 format testing anything specific? Can it be simply swapped out for utf-8 or am I missing something? (I'm not strong on the parse/serializer side of things).

[aucampia]

Can it be simply swapped out for utf-8 or am I missing something? (I'm not strong on the parse/serializer side of things).

If we are not specifically testing character sets support and handing of passing invalid char sets then I think valid charsets (e.g. utf-8 in this case ) is more appropriate than charsets that cause warnings (e.g. ascii/1latin-1` in this case). So I think this change makes a lot of sense.

There should be tests to check the handling of passing of invalid charsets but they should be in a separate place and not mixed in with tests for something different, and they could (should?) even go as far as expecting the warnings. And I also think we should rather error out in some of these cases, especially if the internal handling is to just use utf-8 - I guess there are different approaches here but I think erroring out when an unsupported charset is supplied will result in less surprises for users than warning would. Though this is something we can consider as a seperate issue and discuss potentially for future major versions as this will break compatibility and warrant a major version bump.

I made a bunch of round trip tests that I still have to break off from #1418 that cover character sets, and tests for handling of invalid charsets could maybe go into something like that.

[nicholascar]

In all the 1.1 versions of the W3C Specifications we see UTF-8 mandated, e.g.:

• N-Triples: https://www.w3.org/TR/n-triples/#h2_n-triples-changes
• Turtle: https://www.w3.org/TR/turtle/#h3_sec-mime

So I think we can always use UTF-8. I wondered if there were any good reasons this wasn't already the case but presumably it's just for the historical reasons of different character support in Python < 3 and perhaps in W3C Specs < 1.1.