The correct answer is that you should use case-folded keys as dictionary keys. This is not the same as converting them to upper or lower case and it won't destroy the O(1) average case search/insert complexity.
Unfortunately, Cocoa doesn't seem to have an appropriate NSString
method to case-fold a string, but Core Foundation has CFStringFold()
which you can use for that purpose. Let's write a short function to do the necessary work:
NSString *foldedString(NSString *s, NSLocale *locale)
{
CFMutableStringRef ret = CFStringCreateMutableCopy(kCFAllocatorDefault, 0,
(__bridge CFStringRef)s);
CFStringNormalize(ret, kCFStringNormalizationFormD);
CFStringFold(ret, kCFCompareCaseInsensitive, (__bridge CFLocaleRef)locale);
return (__bridge_transfer NSString *)ret;
}
Note that the locale argument is important. If you specify NULL
, you will get the current system locale. This will be fine in most cases, but Turkish users might be surprised that "I" matches "i" rather than "ı". You might therefore want to pass [NSLocale currentLocale]
, and if you're saving the results you might also want to save the locale identifier and create the locale from that.
So, when adding to the dictionary, you now need to do
[dict setObject:obj forKey:foldedString(myKey, locale)];
and to look up again
[dict objectForKey:foldedString(myKey, locale)];
One final observation is that you might wish to store the case-folded keys alongside the original values, then you don't have to fold them on every access to the dictionary.