SQL optimisation - sql

Question

I want to optimize some of the SQL and just need an opinion on whether I should do it or leave it as is and why I should do it. SQL queries are executed via PHP & Java, I will show an example in PHP which will give an idea of what Im doing.
Main concerns are:
-Maintainability.
-Ease of altering tables without messing with all the legacy code
-Speed of SQL (is it a concern???)
-Readability
Example of what I have right now:
I take a LONG array from a customer (cant make it smaller unfortunately) and update the existing values with the new values provided by a customer in the following way:
$i = 0;
foreach($values as $value)
{
$sql = "UPDATE $someTable SET someItem$i = '$value' WHERE username='$username'";
mysql_query($sql, $con);
$i+=1;
}
Its easy to see from the above example that if the array of values is long, than I execute a lot of SQL statements.
Should I instead do something like:
$i = 0;
$j = count($values);
$sql = "UPDATE $someTable SET ";
foreach($values as $value)
{
if($i < $j) //append values to the sql string up to the last item
{
$sql .= "someItem$i = '$value', ";
}
$i+=1;
}
$sql .= "someItem$i = '$value' WHERE username='$username'"; //add the last item and finish the statement
mysql_query($sql, $con); //execute query once
OR which way should it be done / should I bother making these changes? (there a lot of the type and they all have 100+ items)
Thanks in advance.

Answer 1

The only way you'll get a definitive answer is to run both of these methods and profile it to see how long they take. With that said, I'm confident that running one UPDATE statement with a hundred name value pairs will be faster than running 100 UPDATE statements.

Answer 2

Don't run 100 seperate UPDATE statements!
Use a MySQL wrapper class which, when given an array of name => value pairs will return an SQL UPDATE statement. Its really simple. I'm just looking for the one we use now...
We use something like this (registration required) but adapted a little more to suit our needs. Really basic but very very handy.
For instance, the Update method is just this
/**
* Generate SQL Update Query
* #param string $table Target table name
* #param array $data SQL Data (ColumnName => ColumnValue)
* #param string $cond SQL Condition
* #return string
**/
function update($table,$data,$cond='')
{
$sql = "UPDATE $table SET ";
if (is_string($data)) {
$sql .= $data;
} else {
foreach ($data as $k => $v) {
$sql .= "`" . $k . "`" . " = " . SQL::quote($v) . ",";
}
$sql = SQL::trim($sql , ',');
}
if ($cond != '') $sql .= " WHERE $cond";
$sql .= ";";
return $sql;
}

Answer 3

If you can't change the code, make sure it is enclosed in transaction (if the storage engine is InnoDB) so no non-unique indexes will be updated before commiting transaction (this will speed up the write) and the new row won't be flushed to disk.
If this is MyISAM table, use UPDATE LOW_PRIORTY or lock table before the loop and unlock after read.
Of course, I'm sure you have index on the username column, but just to mention it - you need such index.