Re: WIP patch: add (PRE|POST)PROCESSOR options to COPY

Tom Lane <tgl@sss.pgh.pa.us>

From: Tom Lane <tgl@sss.pgh.pa.us>
To: Alvaro Herrera <alvherre@2ndquadrant.com>
Cc: Simon Riggs <simon@2ndquadrant.com>, Fujii Masao <masao.fujii@gmail.com>, Etsuro Fujita <fujita.etsuro@lab.ntt.co.jp>, Craig Ringer <ringerc@ringerc.id.au>, pgsql-hackers@postgresql.org
Date: 2012-11-14T16:20:59Z
Lists: pgsql-hackers
Alvaro Herrera <alvherre@2ndquadrant.com> writes:
> Simon Riggs escribi:
>> On 14 November 2012 15:09, Fujii Masao <masao.fujii@gmail.com> wrote:
>>> Here, progname for COPY IN is the user-supplied program that takes filename as
>>> its argument and that writes on standard output.

>> I think we should be using FDWs/SRFs here, not inventing new
>> syntax/architectures for executing external code, so -1 from me.

> Hmm, but then you are forced to write C code, whereas the "external
> program" proposal could have you writing a only shell script instead.

I disagree with Simon's objection also, because neither reading from
nor writing to an external program is likely to fit the model of
reading/updating a table very well.  For instance, there's no good
reason to suppose that reading twice will give the same results.  So
force-fitting this usage into the FDW model is not going to work well.

Nor do I really see the argument why a "pipe_fdw" module is cleaner
than a "COPY TO/FROM pipe" feature.

			regards, tom lane