...
>> Yes, its actually quite esay: you dump as you feel apropriate,
>> then create the database with the encoding you want,
>> restore w/o creating database and you are done.
>> Restore sets the client encoding to what it actually was
>> in the dump data (in your case latin-1) and the database
>> would be utf-8 - postgres automatically recodes. No need
>> for iconv and friends.
>>
>> Regards
>> Tino
>>
>
> First of all, thank you for your answer. However, I suspect I did not
> understand your answer, since the commands I used were:
>
> 1) pg_dump -Ft -b -f dump.sql.tar database
> 2) dropdb database
> 3) createdb -E UNICODE database
> 4) pg_restore -d database dump.sql.tar
>
> According to my experience, this produces a "double encoding". As you
> can see, I hand-created the database, with the proper encoding.
> However, when I reimported the database, the result was a latin1
> encoded in utf-8, rather than a pure utf-8.
>
> How my procedure was different with respect to yours?
That was the correct way. I wonder if you have recoding support
enabled? Did you build postgres yourself?
Latin-1 double encoded into utf-8 seems not like possible...
utf-8 barfs on most latin-1 characters, current 8.1 is very
picky about it. So maybe you can work with a small
test table to find out what's going wrong here.
(The changing of the client_enccoding setting in the backup is only
needed in the case when you had data in the wrong encoding
- like SQLAscii filled with latin-1 or something)
Regards
Tino