How do I make an NSPredicate that filters local characters properly?

Question

My app is using a Realm database to store a number of items. With a SearchBar, I search (case insensitive) for an item that may already be stored in the database (and if not, it will be added). The item names often contain one of the swedish characters and my problem is that when searching for a swedish character (å, ä or ö), the filter passes "a" or "o" as well, i.e. a search for "ä" results in a/A, å/Å and ä/Ä.

extension ItemsVC: UISearchResultsUpdating {

    func updateSearchResults(for searchController: UISearchController) {

        if isFiltering() {
            let predicate = NSPredicate(format: "name CONTAINS[cd] %@ AND NONE owners.name == %@", searchController.searchBar.text!, (category?.name)!)
            items = realm.objects(Item.self).filter(predicate).sorted(byKeyPath: "name", ascending: true)

            tableView.reloadData()
        }
    }
}

The d in `CONTAINS[cd]` means diacritic insensitive. Does `CONTAINS[c]` give the results you want? — Willeke, Feb 11 '19 at 00:24
Unfortunately not. Removing the d also removes the case insensitivity for the special characters. — Stalle, Feb 11 '19 at 10:03
Absolutely. It appears like the special swedish characters are not considered upper lower case of the same one. — Stalle, Feb 11 '19 at 13:58

Dávid Pásztor · Accepted Answer · 2019-02-11T15:16:19.023

This seems like a bug in Realm's filter implementation. If you supply the same predicate to an NSArray containing the same Objects, those return the correct results.

class Person: Object {
    @objc dynamic var name:String = ""
}

let people = stringsWithAccents.map{Person(value: ["name":$0])}
try realm.write {
    realm.add(people)
}

let namePredicate = NSPredicate(format: "name CONTAINS[c] %@", searchString)
let foundPeople = realm.objects(Person.self).filter(namePredicate)
print(foundPeople) // Results<Person> <0x7fec74c13060> ( [0] Person { name = á; })

print((people as NSArray).filtered(using: namePredicate)) // [Person { name = á;},Personn { name = Á;}]

I've opened an issue on RealmCocoa's GitHub.

In the meantime as a workaround you can manually check for both the lowercased and uppercased version:

let customCaseInsentiveNamePredicate = NSPredicate(format: "name CONTAINS %@ OR name CONTAINS %@", searchString, searchString.uppercased())
print(realm.objects(Person.self).filter(customCaseInsentiveNamePredicate))

How do I make an NSPredicate that filters local characters properly?

1 Answers1