On 21-12-2014 13:39, Geoff Worboys geoff@...
[firebird-support] wrote:

> I wrote:
>> This:
>
>> SELECT Upper(_win1252 'ƒ') FROM RDB$DATABASE
>
>> (Where 'ƒ' is decimal 131 / 0x83)
>
>> Gives me error 335544321
>> arithmetic exception, numeric overflow, or string truncation
>> Cannot transliterate character between character sets.
>
> I've just answered my own question.
>
> Internal to Firebird the string is converted to Unicode, so
> WIN1252 0x83 becomes U+0192. This is then converted to
> uppercase which becomes U+0191. And then it tries to convert
> back to WIN1252 and there is no WIN1252 mapping for that
> unicode character (U+0191).
>
> Bummer.
>
> I can understand the logic in Firebird, but I don't relish
> trying to explain to my customers: this will all work as long
> as they stay away from 'ƒ'. (Such an explanation is likely
> to have them using words starting with Ƒ, but I will have
> to tell them, they can't use that in this database because we
> don't have a mapping for it. ;-)
>
> As a short term solution I'm going to translate U+0191 back to
> WIN1252 0x83 in my own Firebird build. Long term I hope to
> move to Unicode.
>
> Thanks for listening, you've been a wonderful audience. :-)

I think this is a bug and it should be reported as such. If there is no
uppercase equivalent of a character in the current character set, then
UPPER() should use the character itself, not throw a transliteration error.

However, have you considered using a case insensitive collation instead
of using UPPER() (assuming the actual problem is with comparisons, not
with displaying uppercase characters)?

Mark
--
Mark Rotteveel