Character data test - character sets

[imanuelcostigan]

One of the parts of the data_character() tests is executing the following query: SELECT 'Кирилл' AS A.

By default, this will return ????? in SQL Server (even natively). What needs to be executed is SELECT N'Кирилл' AS A. This is because, by default, SQL Server uses a non-UNICODE character set so one needs to declare a string literal as UNICODE using the N' prefix.

As a result this test case fails.

[krlmlr]

Is the N prefix part of the standard, or a SQL server extension? Anyway, we might need a tweak here.

[imanuelcostigan]

i suspect the latter. A quick search suggests that maybe Oracle has a similar "feature"

[krlmlr]

From T-SQL docs:

Unicode strings have a format similar to character strings but are preceded by an N identifier (N stands for National Language in the SQL-92 standard).

MySQL and PostgreSQL accept this syntax (I'm not sure about the semantics), SQLite rejects it.

MySQL has "introducers", too, looks like you can ask SELECT _utf8'text'. However, things work out without these introducers here because the connection character set is changed to utf8 first thing after establishing the connection. I suspect this isn't possible with jTDS (see somewhat related SO question), but I might be wrong.