Обсуждение: Client Encoding and Latin characters

My database is encoded UTF8. I recently was uploading (via COPY) some
census data which included place names with ñ, é, ü, and other such
characters. The upload choked on the Latin characters. Following the
docs, I was able to fix this with:

SET CLIENT_ENCODING TO 'LATIN1';
COPY table FROM 'filename';

After which I

SET CLIENT_ENCODING TO 'UTF8';

I typically use COPY FROM to bulk load data. My question is, is there
any disadvantage to setting the default client_encoding as LATIN1? I
expect to never be dealing with Asian languages, or most of the other
LATINx languages. If I ever try to COPY FROM data incompatible with
LATIN1, the command will just choke, and I can pick an appropriate
encoding and try again, right?

Thanks,
--Lee

--
Lee Hachadoorian
PhD Student, Geography
Program in Earth & Environmental Sciences
CUNY Graduate Center

Lee Hachadoorian <lee.hachadoorian@gmail.com> writes:
> My database is encoded UTF8. I recently was uploading (via COPY) some
> census data which included place names with �, �, �, and other such
> characters. The upload choked on the Latin characters. Following the
> docs, I was able to fix this with:

> SET CLIENT_ENCODING TO 'LATIN1';
> COPY table FROM 'filename';

> After which I

> SET CLIENT_ENCODING TO 'UTF8';

> I typically use COPY FROM to bulk load data. My question is, is there
> any disadvantage to setting the default client_encoding as LATIN1? I
> expect to never be dealing with Asian languages, or most of the other
> LATINx languages. If I ever try to COPY FROM data incompatible with
> LATIN1, the command will just choke, and I can pick an appropriate
> encoding and try again, right?

Uh, no.  You can pretty much assume that LATIN1 will take any random
byte string; likewise for any other single-byte encoding.  UTF8 as a
default is a bit safer because it's significantly more likely that it
will be able to detect non-UTF8 input.

            regards, tom lane

> Uh, no.  You can pretty much assume that LATIN1 will take any random
> byte string; likewise for any other single-byte encoding.  UTF8 as a
> default is a bit safer because it's significantly more likely that it
> will be able to detect non-UTF8 input.
>
>                        regards, tom lane
>

So, IIUC, the general approach is:

*Leave the default client_encoding = server_encoding (in this case UTF8)
*Rely on the client to change client_encoding on a session basis only

Thanks,
--Lee

--
Lee Hachadoorian
PhD Student, Geography
Program in Earth & Environmental Sciences
CUNY Graduate Center