NOTE: Doing this in SQL will be WAYYYYYYYYY faster and use far less memory. I would only try this PHP approach if you cannot modify the SQL for some reason.
Maybe try out... hit the db twice, first time only get non-overrides, then get the overrides in second pass -- coerce your arrays to be indexed by name and array_merge them. (Uses a fair chunk of memory given the number of arrays and copies - but it's easy to understand and keeps it simple.
$initial = get_non_overridden();
$override = get_overridden();
$init_indexed = index_by_name($initial);
$over_indexed = index_by_name($override);
$desired_result = array_merge($init_indexed, $over_indexed);
Assuming your database gives you a standard rowset (array of rows, where each row is a hash of fields->values). We want something that looks like this instead:
[
'Eclipse' => [
'name' => 'Eclipse',
'override' => '0',
'percentage' => '75%'
],
'Something' => [
'name' => 'Something',
'override' => '0',
'percentage' => '20%'
],
]
So index_by_name would be:
function index_by_name($rowset) {
$result = array();
foreach ($rowset as $row) {
$result[ $row['name'] ] = $row;
}
return $result;
}
There are ways to tweak your efficiency either in memory or run time, but that's the gist of what I was thinking.
array_merge then overwrites the initial ones with the overridden ones.
NOTE: this all assumes that there is only one row where Eclipse override is 1. If you have twenty Eclipse|0 and one Eclipse|1, this will work, if you have two Eclipse|1 you'd only see one of them... and there's no way to say which one.