Tom Lane wrote:
> OK, reasonable arguments were made why not to allow multi-character
> separators. Should we then match the server and insist on a single-byte
> separator? It's a bit inconsistent if psql can be made to emit "csv"
> files that COPY can't read, especially when it's otherwise a subset
> of what COPY allows.
I seem to be the only one advocating to accept an Unicode
character here, and I won't insist on it.
There's still a minor annoyance to solve if we want to claim full
compatibility with COPY CSV: backslash followed by dot must be
quoted to avoid being interpreted as an end of data indicator.
A proposed fix is attached. print_csv_vertical() is left unchanged
because it's not possible currently to end up with \. alone
on a line with the expanded display: we'd need to allow
first for an empty field separator, I believe.
Best regards,
--
Daniel Vérité
PostgreSQL-powered mailer: http://www.manitou-mail.org
Twitter: @DanielVerite