Re: delta relations in AFTER triggers

On Sun, Dec 4, 2016 at 11:35 PM, Haribabu Kommi
<kommi.haribabu@gmail.com> wrote:

> Moved to next CF with "waiting on author" status.

Patch v8 attempts to address the issues explicitly raised in
Thomas Munro's review.  An opaque query environment is created
that, for now, only passes through ephemeral named relations, of
which the only initial implementation is named tuplestores; but
the techniques are intended to support both other types of
ephemeral named relations and environmental properties (things
affecting parsing, planning, and execution that are not coming from
the system catalog) besides ENRs.  There is no clue in the access
to the QE whether something is, for example, stored in a list or a
hash table.  That's on purpose, so that the addition of other
properties or changes to their implementation doesn't affect the
calling code.

There were a few changes Thomas included in the version he posted,
without really delving into an explanation for those changes.  Some
or all of them are likely to be worthwhile, but I would rather
incorporate them based on explicit discussion, so this version
doesn't do much other than generalize the interface a little,
change some names, and add more regression tests for the new
feature.  (The examples I worked up for the rough proof of concept
of enforcement of RI through set logic rather than row-at-a-time
navigation were the basis for the new tests, so the idea won't get
totally lost.)  Thomas, please discuss each suggested change (e.g.,
the inclusion of the query environment in the parameter list of a
few more functions).

Changed to "Needs review" status.

--
Kevin Grittner
EDB: http://www.enterprisedb.com
The Enterprise PostgreSQL Company