It has been good to see so many results in the leaderboard. Currently Uno looks unbeatable with his current lead :)
Over the last week I’ve been working on finishing the leaderboard feature so now as you scroll down the leaderboard it will load more results past the original 11 entry limit.
To support this I created a utility that will insert fake users and results into the leaderboard that I can use for testing the leaderboard once it has gotten larger.
Unfortunately this enabled me to discover that the frame rate begins to drop as the leaderboard surpasses 300 entries. This is because the default Unity scroller I am currently using doesn’t hide entries when there not currently being displayed meaning it constantly has to manage 300 entries. When it gets to a 1000 it's very stuttery. To fix this I believe I will have to make a new scroller from scratch that hides undisplayed entries to minimise the processing performed by Unity to display the leaderboard.
There is also currently an issue where once you scroll to the bottom of the leaderboard it will begin creating new entries to be populated with values. However because it is at the bottom of the leaderboard there are no new value to populate the entry with. So these empty entries are deleted. Sometimes the Unity scroller thinks it's passed the end of the leaderboard meaning it continues to create empty entries until stopped by the user.
Most of the development went into the server which can now load arbitrary chunks of data from either a cache or the leaderboard database depending on whether the data is already in the cache. This currently has a problem that if I make lots of small requests when someone is scrolling slowly, the cache will end up containing many small blocks of cached leaderboard results. This can make it slow to return a larger range result as it needs to examine and build the result out of many small cache blocks. There are future optimisations that I can make to this to improve performance. This will be either ensuring that range requests happen with a predictable size and alignment, this will mean just cache results and not assemble results out of smaller cache blocks. Alternatively, possibly getting smaller cache blocks that are next to each other to merge, so there are less cache blocks to deal with when servicing a request.