Re: DBI/mysql gathering all rows on execute

[gmax]

There are two possible causes, perhaps both of them at the same time:

• If there is no index in any of link1.link_from, url1.urlid, link1.link_to, and url2.urlid, then the join takes much longer than it should. If this is the case, add an appropriate index, using an ALTER TABLE statement.
• Your query is returning an huge amount of rows, which are clogging both your server and client memory. The reason is that MySQL default mode of fetching is All-At-Once (mysql_store_result). You can affect this behavior by setting the appropriate mode identifier after preparing the statement:

  my $sth = $dbh->prepare($sql);
  $sth->{"mysql_use_result"} = 1;
  

  Be aware that this mode requires that you either fetch all the records or use a "finish" method before issuing another request to the server.
  See MySQL manual for an explanation of this behavior (it's related to the C API) and DBD::mysql for more details on this modifier in the DBD interface.
  With "mysql_use_result", as opposed to "mysql_store_result", the server won't send any record to the client until you use a "fetch" method. What happens with the default mode is that the server sends all the records at once, and the "fetch" method is just browsing a local copy of the data.

[sgifford]

Thanks, it was $sth->{"mysql_use_result"} = 1; that I needed!

The right indexes are all in place (and EXPLAIN SELECT agrees), there are just a huge number of rows in the result set.

Now here's the other weird thing, which doesn't matter for my real project but is still unusual. When I make this change, the test program I posted above works better, but now it hangs for a long time during the finish. That surprises me; I would expect it to cancel the query and return immediately. Anybody know what's going on?