use utf8mb4 encoding as a default

[pangwa]

utf8mb4 supports a wider range characters than utf8(e.g. emoj). It would be safe to use utf8mb4 instead of utf8.

[naushadh]

I'd love to move to full UTF-8 range support as well (via utf8mb4); however, upstream commentary #679 suggests this change might be b/w incompatible. The user has no way to "go back" to utf8 if they want to.

Edit: realized something, see below.

[naushadh]

@pangwa Can't setMySQLConnectInfoCharset be used to override the default encoding already?

[pangwa]

ah, it looks like setMySQLConnectInfoCharset should work. Thanks. I'm closing this issue.

[alexkazik]

There is another issue with utf8/utf8mb4: the create table instruction also uses uft8, and there is no switch for that.
A solution is to specify the sql-type for all textual columns by hand
since it does not help when the connection supports 4-byte-utf8 but the database only 3-byte-utf8.

[naushadh]

@alexkazik you're right. re-opening the issue. Just stumbled upon the upstream issue: yesodweb#813

[naushadh]

Could MySQLConnectInfo's charset property be re-used to ensure the migrations are invoked with the right encoding?

[alexkazik]

I like the idea, i'd say use utf8mb4 when it's also set in the ci, and urf8(mb3) otherwise - for compatibility.
I'll make a patch for persistent-mysql first and if it's accepted there then also for persistent-mysql-haskell.
Sounds good?

[naushadh]

👍