| 1 | 1 | A side effect of #792 is that we must be able to store non-latin1 characters in the database. Eg. µ (micro) and Ω (ohm). Since MySQL supports the using different character sets on different columns, we originally though that this would not be problematic since it was possible to force the UnitSymbols.symbol column to use UTF-8. In practice there is a problem since we must now set the characterEncoding=UTF-8 in the connection url (db.url setting in base.config). This works fine for the UnitSymbols.symbol column, but not for 'text'-type columns that are latin1 columns. So trying to insert a description containing, for example 'åäö', results in 'åäö' being stored in the database. Eg. it seems like the UTF-8-encoded string is being inserted without first converting it to latin1. This would probably be ok if there was no conversion when reading the value back. But in this case the the bad string is "converted" to UTF-8 encoding again, resulting in: åäö. My feeling is that this might be a MySQL bug since there is no problem for 'varchar' type columns, though I don't know if this is because there is or isn't a conversion for both directions. |
