| Class | TaskJuggler::DataCache |
| In: |
lib/taskjuggler/DataCache.rb
|
| Parent: | Object |
This class provides a global data cache that can be used to store and retrieve values indexed by a key. The cache is size limited. When maximum capacity is reached, a certain percentage of the least requested values is dropped from the cache. The primary purpose of this global cache is to store values that are expensive to compute but may be need on several occasions during the program execution.
# File lib/taskjuggler/DataCache.rb, line 57 def initialize resize flush # Counter for the number of writes to the cache. @stores = 0 # Counter for the number of found values. @hits = 0 # Counter for the number of not found values. @misses = 0 # Counter for hash collisions @collisions = 0 end
Ruby 1.8 has a buggy hash key generation algorithm that leads to many hash collisions. We completely disable caching on 1.8.
# File lib/taskjuggler/DataCache.rb, line 90 def cached(*args) yield end
args is a set of arguments that unambigously identify the data entry. It‘s converted into a hash to store or recover a previously stored entry. If we have a value for the key, return the value. Otherwise call the block to compute the value, store it and return it.
# File lib/taskjuggler/DataCache.rb, line 100 def cached(*args) key = args.hash if @entries.has_key?(key) e = @entries[key] if e.unhashedKey != args # Two different args produce the same hash key. This should be a # very rare event! @collisions += 1 yield else @hits += 1 e.value end else @misses += 1 store(yield, args, key) end end
For now, we use this randomly determined size.
# File lib/taskjuggler/DataCache.rb, line 71 def resize(size = 100000) @highWaterMark = size # Flushing out the least used entries is fairly expensive. So we only # want to do this once in a while. The lowWaterMark determines how much # of the entries will survive the flush. @lowWaterMark = size * 0.9 end