What the naming conventions actually say is that "Methods and class properties should use lowerCamel naming." (my emphasis).
The properties added to the object returned by a query are not class properties. This is confirmed by the PHP documentation which defines "properties" (or "class member variables") as "defined by using one of the keywords public, protected, or private, followed by a normal variable declaration". There is no such declaration that applies here.
The object you are looking at is of type StdClass. There isn't much about the concept behind StdClass in the PHP documentation (unless I missed it), but the concept of a class-less object with dynamic fields is sometimes known as "anonymous object". You can create one in PHP by doing:
$obj = (object)(array('field' => 'value'));
echo $obj->field; // Outputs 'value'
In the case of the object returned by the query, it doesn't even have methods - it could just be an array (which it would be if you fetched your data using DatabaseStatementInterface::fetchAssoc). So we can think of anonymous objects as containers, rather than instantiations of classes to which the coding style should apply.
Since this object is a container of fields from the database, I think it makes sense for the database convention to apply. It is an object wrapper around a database row.
As a Drupal programmer, if I get a result that is essentially a database row, I will actually expect it to be in underscore notation. While it may feel wrong, it is what people are used to expect. Every time I use [node_load][4]
, the result I get is an object in underscore notation. Even when you use the entity meta data wrappers, which essentially are there to provide an object interface around Drupal entities, you still use the underscore notation. Changing this approach would potentially confuse other Drupal programmers.
I understand that if you are writing, say, an API which returns your own data, you may want to abstract the fact the result is a database row and present it as an object like any other (as suggested by your param vs. result example). One option, though one that has some overheads, would be to create a class to hold such result, and casting the database result object to it. This would at least make it clear that what you are returning is a proper object, not just a container around a database row. You could do something like this:
Class MyApiResult {
function __construct($row) {
$vars = get_object_vars($row);
foreach ($vars as $name => $value) {
$cc_name = to_camel_case($field);
$this->{$cc_name} = $value;
}
}
}
Then your API would return:
return new MyApiResult($row);
However this has some overheads, it depends whether those will be acceptable in your particular case.