Carlo Stonebanks wrote:
>> I think the ORDER BY is free to update the rows in any order it needs
>> to. The key is to put the sequence further down. How about this?
> Adam - thanks. Unless I hear otherwise I will assume that you mean the
> UPDATE is free to update the rows in any way it wants - irregardless of
> how whether the data return in the FROM clause is ordered.
Whoops - you're right, that's what I meant. Even though one of the
tables is sorted, when it joins the two tables, the planner might decide
it's cheaper to update impt_table by iteratively looking up values from
the sorted table that match the (unordered) rows from impt_table, or
by hashing the keys which I think also results in unsorted updates.
You can use "explain" to see the details.
In response to
pgsql-general by date
|Next:||From: Tom Lane||Date: 2009-04-28 21:51:55|
|Subject: Re: Restore Crashes Postgres |
|Previous:||From: Raymond O'Donnell||Date: 2009-04-28 20:19:07|
|Subject: Re: Welcome to the pgsql-bugs list!|