Tom Lane-2 wrote
> COPY is not smart at all. It just looks at the column types of the
> target table and assumes that the incoming data is of those types.
> (More precisely, it applies the input conversion function of each
> column's data type, after having separated and de-escaped the text
> according to datatype-independent format rules.)
>
> In that case, adjusting the source data is the way to go. Or you could
> look at using an external ETL tool to do that for you. We've resisted
> putting much transformational smarts into COPY because the main goal
> for it is to be as fast and reliable as possible.
>
> regards, tom lane
I see, it's that input conversion function that I would have needed to
change.
I understand and agree, we depend on COPY FROM in many contexts to upload
huge batches of data orders of magnitude faster than INSERT. We've also used
it recently as a generic table interface for smaller configuration tables,
but you're right: speed and reliability should be the primary focus of COPY
FROM.
Thanks for all the quick responses, you guys.
--
View this message in context:
http://postgresql.1045698.n5.nabble.com/Implicit-casts-to-array-types-tp5736582p5736610.html
Sent from the PostgreSQL - general mailing list archive at Nabble.com.