Skip text->binary conversion of unnecessary columns in contrib/file_fdw.
When reading from a text- or CSV-format file in file_fdw, the datatype
input routines can consume a significant fraction of the runtime.
Often, the query does not need all the columns, so we can get a useful
speed boost by skipping I/O conversion for unnecessary columns.
To support this, add a "convert_selectively" option to the core COPY code.
This is undocumented and not accessible from SQL (for now, anyway).
Etsuro Fujita, reviewed by KaiGai Kohei
Branch
------
master
Details
-------
http://git.postgresql.org/pg/commitdiff/a36088bcfae85eeeb55e85c3f06c61cb2f0621c6
Modified Files
--------------
contrib/file_fdw/file_fdw.c | 147 ++++++++++++++++++++++++++++++++++++++++++-
src/backend/commands/copy.c | 53 +++++++++++++++
2 files changed, 197 insertions(+), 3 deletions(-)