Douglas,
> 1. Copy to non-existent table - create it from the data type using most generic datatypes
Yes, that would be nice. Types chosen would pretty much have to be
TEXT, NUMERIC, and TIMESTAMP for everything though; anything else is too
variable.
> 2. provide column mapping function - from what I understand that the column list on a copy from is the target
columnsto load. I assume there must be a header record and the same columns in the
So, like:
COPY table1 ( cola, colb, colc ) FROM 'somecsv.csv' ( 'name', 'place',
'date') WITH CSV HEADER
... which would copy 'name' to cola, 'place' to colb, 'date' to colc,
and ignore any other columns present in the file?
--Josh Berkus