Benchmarked with what code?

With this code (please test it for any database that has large number of records and columns):

function microtime_float()
{
    list($usec, $sec) = explode(" ", microtime());
    return ((float)$usec + (float)$sec);
}

echo "<font color='green'>1. mysql_fetch_assoc() + array</b></font>";

$link = mysql_connect(HOST,USER,PASS);
mysql_select_db(DB,$link);

$result = mysql_query("SELECT ... FROM ... ORDER BY '...' DESC LIMIT 300");

mysql_close($link);

$start = microtime_float();

while($tab1 = mysql_fetch_assoc($result))
{
}

$stop = microtime_float();

echo "<BR><BR>".($stop - $start)." [s].";

echo "<BR><BR><font color='green'>2. mysql_fetch_row + array</b></font>";

$start = microtime_float();

while($tab2 = mysql_fetch_row($result))
{
}

$stop = microtime_float();

echo "<BR><BR>".($stop - $start)." [s].";

echo "<BR><BR><font color='green'>3.  mysql_fetch_row + list()</b></font>";

$start = microtime_float();

while(list([variables for all 42 columns]) = mysql_fetch_row($result))
{
}

$stop = microtime_float();

echo "<BR><BR>".($stop - $start)." [s].";

echo "<BR><BR><font color='green'>4. mysql_fetch_row() + foreach()</b></font>";

while($info = mysql_fetch_row($result))
{
	foreach($info as $field);
}

$stop = microtime_float();


echo "<BR><BR>".($stop - $start)." [s].";

echo "<BR><BR><font color='green'>5. mysql_fetch_object()</b></font>";

while($object = mysql_fetch_row($result))
{
}

$stop = microtime_float();

echo "<BR><BR>".($stop - $start)." [s].";

Obviusly there should be:

while($object = mysql_fetch_object($result))
{
}

instead of:

while($object = mysql_fetch_row($result))
{
}

but it is not so important as we mainly compare mysql_fetch_assoc() with mysql_fetch_row().

I don't think the docs should be changed.  This is worst-case analysis, and  performance is only measurably different because the reporter is extracting too many rows.  In the typical case I would never recommend someone grab greater than 100 in one go, and in that case the round trip cost is much more visible as being a large part of the cost.

If there's any docs correction to be made - it should be that 90K rows at a time is excessive, and not a best practice.

Hi,

I've replicated your test case by using the World.sql file provided by MySQL.
http://dev.mysql.com/doc/world-setup/en/world-setup.html

There's an issue with the way your tests are run, in that whichever one runs 
first always runs slowest.

So I'll get:
fetch row: 0.0063321590423584
fetch asc: 4.0531158447266E-6

Then:

fetch asc: 0.0053088665008545
fetch row: 3.0994415283203E-6

When I reverse the order. This is likely the result of some processing occurring 
when the data is pulled out of the result set.



paul