>>>>> "Matthias" == Matthias Apitz <guru@unixarea.de> writes:
Matthias> ... WITH DELIMITER '|', FORMAT CSV ;
I think you misread the docs; the new-style syntax would be
COPY ... WITH (DELIMITER '|', FORMAT CSV);
where the parens are not optional. The old-style syntax with no parens
after WITH is what you were using before.
Matthias> 2. The CSV export is done by some Sybase tool which escapes
Matthias> the delimiter as '\|', i.e. putting a backslash before the
Matthias> delimiter. I found no way that COPY understands this
Matthias> excaping. Any ideas?
That sounds like the file is not actually a CSV - why do you think it
is?
PG accepts two formats (actually 3 if you count binary format which is
rarely used):
TEXT: records are delimited by newlines, columns are delimited by a
delimiter character (default tab), there are no quotation marks, any
newline, backslash, or delimiter in the data must be escaped as a
backslash-sequence (e.g. \n or \012)
CSV: columns may be quoted (in which case delimiters and newlines inside
them are ignored), records are delimited by newlines _outside_ quoted
fields, there are no backslash-sequences or escapes outside of quoted
fields, quote characters inside quoted fields are doubled (though
there's an option to change this).
PG follows the CSV spec at https://www.ietf.org/rfc/rfc4180.txt fairly
closely.
--
Andrew (irc:RhodiumToad)