Performance enhancement to views_date_format_sql_views_data_alter()

Tagged with commerce sprint because this is a dependency of commerce_reports.

@zany my thought on removing the references was that it won't have to set the reference in the loop, though I hardly think that would make or break the patch, the big fix is isset() from is_array() checks.

Quick show of how that will not produce errors:
https://3v4l.org/J2pcc

It would only produce errors I think if field was ever an object... maybe

Thank you @zany for replying so quick. Any chance the patch wasn't pushed?

Ah i see, thanks. Normally the bot picks up the commit and shows it here. Caught me off guard.

Any chance you could open a dev branch & release for this module?

Let's undo the reference change I made:) I was curious about the performance impact about that too from your question.

// Seed data.
foreach (range(0, 1000) as $x) {
  $row = array();
  foreach (range(0, 100) as $y) {
    $row[$y]['field'] = 1;
  }
  $data[uniqid()] = $row;
}
// Copy.
$data2 = $data;

$START = microtime(TRUE);
foreach($data as $module => &$table) {
  foreach($table as $id => &$field) {
    $field['field'] = 1;
  }
}
$END = microtime(TRUE) - $START;
echo "With reference took $END seconds\n";

$start = microtime(TRUE);
foreach ($data2 as $module => $table) {
  foreach ($table as $id => $field) {
    $data[$module][$id]['field'] = 1;
  }
}
$END = microtime(TRUE) - $START;
echo "With keys took $END seconds\n";

echo "Same array? \n";
var_dump($data === $data2);

Results:

With reference took 0.18254899978638 seconds
With keys took 0.47187304496765 seconds

With reference took 0.18951106071472 seconds
With keys took 0.33031606674194 seconds

With reference took 0.18168807029724 seconds
With keys took 0.31012606620789 seconds

I go in bouts of performance testing:-)

Want me to propose an interdiff or new patch for master or can you take care of that @zany?