Tatsuo Ishii <t-ishii@sra.co.jp> writes:
> I have added new function called "convert" similar to SQL99's convert.
> Convert converts encoding according to parameters. For example, if you
> have a table named "unicode" in an Unicode database,
> SELECT convert(text_field, 'LATIN1') FROM unicode;
> will return text in ISO-8859-1 representation.
I don't understand how this works. If you have a multibyte-enabled
backend, won't backend libpq try to convert all outgoing text to
whatever PGCLIENTENCODING says? How can it know that one particular
column of a result set is not in the regular encoding of this database,
but something else? Seems like libpq is going to mess up the results
by applying an inappropriate multibyte conversion.
regards, tom lane