Should we be using charset 'utf8mb4'?

[pacso]

As per this article, the MySQL charset 'utf8' only supports up to 3 bytes per character, when UTF-8 actually allows up to 4 bytes per character.

Shouldn't we be using 'utf8mb4' instead of 'utf8' (which is just an alias for 'utf8mb3') in order to fully support the UTF-8 character set?

Am happy to submit a PR, but wanted to discuss first.

[brianmario]

The only place I can think of where we specify utf8 in this gem is in the constructor where we default to that (instead of utf8mb4). We could change the default there, but I think utf8mb4 is only supported in MySQL 5.5+. But at that point in the code, we don't know what version of the server we're going to connect to so I'm not sure how best to handle that...

@sodabrew you have any ideas? I guess we could require a newer MySQL server version but I dunno... There may still be plenty of folks out there on <=5.1. I'm inclined to keep this the way it is but maybe add some documentation suggesting the use of utf8mb4 if you're using 5.5+?

[pacso]

Documentation might be the right way to go. You’re right that there are going to be plenty of people using older versions of MySQL who could be affected, but in all likelihood, if they’re not updating their DB version they probably aren’t updating their gems either...

In any case, documenting it rather than distributing a breaking change is definitely preferable. I do think it’s important that people are aware of this though.

[sodabrew]

I actually removed the MySQL 5.1 integration tests already because I could reliably keep the Travis images working, and updated the README accordingly: https://github.com/brianmario/mysql2#compatibility

Changing the default value surely counts as a breaking change, but one years-overdue to happen.

[eikes]

I was surprised to find out that my rails app wasn't able to handle emojis due to this problem. utf8 should not be used at all since it is not really UTF-8, utf8mb4 should definitely be the default.

[jeremy]

Possible to let MySQL autodetect charset (using the "auto" charset aka MYSQL_AUTODETECT_CHARSET_NAME) then map it to the appropriate Ruby string encoding? Then client charset needn't be declared up front; use what the db says.