Re: portability of "designated initializers"

Alvaro Herrera escreveu:

> I've already modified your patch a bit ... please send your updated
> patch so I can merge it into mine. However, my changes were also
> relatively minor. Since Tom wants it to be entirely rewritten then
> maybe merging minor fixes to it is a waste of time ...
>
Since Alvaro is involved in finishing another patches, I improved the first
patch based on discussion. What did I do?

(i) change from relOptXXX arrays to relOpts as Tom pointed out in [1];
(ii) fix some memory alloc/free problems;
(iii) add some code to uniform the output of boolean parameters (output will
be true/false);
(iv) change the *_base_thresh names to *_threshold. I know the names are not
appropriated but postgresql.conf use similar ones;
(v) rollback the RelationGetXXX changes as Tom pointed out in [2];
(vi) fillfactor is stored as a separate variable in RelationData;
(vi) include documentation.

I did some small benchmarks to figure out if storing the parameters as a bytea
(rd_options) has any performance penalty; i don't see any significant
difference but my laptop is not a good parameter. Comments?

PS> I don't include regression tests but if you think it's important to
exercise that code path, I will elaborate some tests.

[1] http://archives.postgresql.org/pgsql-hackers/2008-11/msg01494.php
[2] http://archives.postgresql.org/pgsql-hackers/2008-11/msg01500.php

--
Euler Taveira de Oliveira
http://www.timbira.com/