On 01/05/11 11:53, Greg Stark wrote: <blockquote cite="mid:BANLkTikP66KizPDXsvqCtibYLCfXxS7yBw@mail.gmail.com"
type="cite"><prewrap="">On Sat, Apr 30, 2011 at 9:21 PM, Tom Lane <a class="moz-txt-link-rfc2396E"
href="mailto:tgl@sss.pgh.pa.us"><tgl@sss.pgh.pa.us></a>wrote:
</pre><blockquote type="cite"><pre wrap="">- it would require a query in which
every relation is linked to every other relation by a join clause.
But that *can* happen (remember that clauses generated by transitive
equality do count).
</pre></blockquote><pre wrap="">
It sounds like you're describing precisely a "star schema" join which
isn't an uncommon design pattern at all.
</pre></blockquote><font size="-1"><font face="Helvetica"><br /> Nice example here:<br /><br /><a
class="moz-txt-link-freetext"
href="http://archives.postgresql.org/pgsql-bugs/2011-04/msg00100.php">http://archives.postgresql.org/pgsql-bugs/2011-04/msg00100.php</a><br
/><br/> Strictly only a 'star-like' query as the foreign key references go the opposite way from a true star. However
itillustrates the planner memory growth well (1.1G on 32-bit 1.7G on 64-bit systems).<br /><br /> A point I didn't
mentionis that the memory use is quite dependent on the choice of "word" values for the "AND keyword = 'word'" clause -
thetext example had 6 all the same. Setting them all different (even after adjusting the data so the there *was* a
numberof matching rows to find) resulted in significantly less memory consumed (I can dig up some examples if it might
beinteresting).<br /><br /> Cheers<br /><br /> Mark<br /></font></font>
