class _BoundedCacheSet:
# implements LRU. I could've implemented this using
# a set and then removed a random item from the set whenever the set
# got too large. Hmmm...
def __init__(self, max_items):
assert max_items > 1
self._max_items = max_items
self._data = IndexedMultiMap() # recent accesses.
def add(self, key): # O(log n)
i = self._data.find_key_by_value(key)
t = time()
if i.at_end():
self._data[t] = key
else:
self._data.update_key(i,t)
while len(self._data) > self._max_items:
j = self._data.begin()
assert not j.at_end()
self._data.erase(j)
def __contains__(self, key):
i = self._data.find_key_by_value(key)
return not i.at_end()
def remove(self, key):
i = self._data.find_key_by_value(key)
if i.at_end():
raise KeyError()
self._data.erase(i)
def __str__(self):
return str(self._data)
def __init__(self,
default_ttl=None,
expire_interval=60,
touch_on_access=False,
max_items=None,
recent_items=RECENT_SIZE):
"""
@param default_ttl: time to live when using __setitem__ rather than set.
@param expire_interval: time between removals of expired items in seconds. Otherwise,
expired items are removed lazily.
@param touch_on_access: refresh item expire time by ttl when item is accessed.
@param max_items: maximum size of cache. (see replacement algorithm above)
"""
self._exp = IndexedMultiMap(
) # expiration times. Multiple items can have the same expiration
# times, but there can only be one instance of any one key
# in the CacheMap.
self._data = {}
self._ttl = default_ttl
self._touch = touch_on_access
self._max_items = max_items
self._expire_interval = expire_interval
if max_items is not None:
self._recent = _BoundedCacheSet(int(min(recent_items, max_items)))
else:
self._recent = None
reactor.callLater(self._expire_interval, self._expire)
class _BoundedCacheSet:
# implements LRU. I could've implemented this using
# a set and then removed a random item from the set whenever the set
# got too large. Hmmm...
def __init__(self, max_items):
assert max_items > 1
self._max_items = max_items
self._data = IndexedMultiMap() # recent accesses.
def add(self, key): # O(log n)
i = self._data.find_key_by_value(key)
t = time()
if i.at_end():
self._data[t] = key
else:
self._data.update_key(i, t)
while len(self._data) > self._max_items:
j = self._data.begin()
assert not j.at_end()
self._data.erase(j)
def __contains__(self, key):
i = self._data.find_key_by_value(key)
return not i.at_end()
def remove(self, key):
i = self._data.find_key_by_value(key)
if i.at_end():
raise KeyError()
self._data.erase(i)
def __str__(self):
return str(self._data)
def __init__(self, default_ttl = None, expire_interval = 60, touch_on_access = False,
max_items = None, recent_items = RECENT_SIZE ):
"""
@param default_ttl: time to live when using __setitem__ rather than set.
@param expire_interval: time between removals of expired items in seconds. Otherwise,
expired items are removed lazily.
@param touch_on_access: refresh item expire time by ttl when item is accessed.
@param max_items: maximum size of cache. (see replacement algorithm above)
"""
self._exp = IndexedMultiMap() # expiration times. Multiple items can have the same expiration
# times, but there can only be one instance of any one key
# in the CacheMap.
self._data = {}
self._ttl = default_ttl
self._touch = touch_on_access
self._max_items = max_items
self._expire_interval = expire_interval
if max_items is not None:
self._recent = _BoundedCacheSet(int(min(recent_items,max_items)))
else:
self._recent = None
reactor.callLater(self._expire_interval, self._expire)
class CacheMap:
"""this cache class allows caching with arbitrary expiration times. This is different
from BTL.Cache which assumes all items placed in the cache remain valid for the same
duration.
Like a BTL.Map, there can only be one instance of any given key in
the cache at a time. Subsequent inserts (__setitem__ or set)
with the same key, will update the ttl and value for that
key.
Unlike a BTL.Map, CacheMap does not perform in-order iteration based on key, and
key lookups (__getitem__) take average O(1) time.
The map also has the option to have bounded size in which case it imposes the
following replacement algorithm: remove the oldest entries first unless
those entries are in the recent access set. Here 'old' refers to duration
in the cache. Recent set has bounded size.
"""
# BTL.Cache places the cache entries in a queue. We instead maintain an
# IndexedMultiMap ordered based on expiration times. The index allows nodes in the
# map to be looked up in O(1) time based on value.
def __init__(self, default_ttl = None, expire_interval = 60, touch_on_access = False,
max_items = None, recent_items = RECENT_SIZE ):
"""
@param default_ttl: time to live when using __setitem__ rather than set.
@param expire_interval: time between removals of expired items in seconds. Otherwise,
expired items are removed lazily.
@param touch_on_access: refresh item expire time by ttl when item is accessed.
@param max_items: maximum size of cache. (see replacement algorithm above)
"""
self._exp = IndexedMultiMap() # expiration times. Multiple items can have the same expiration
# times, but there can only be one instance of any one key
# in the CacheMap.
self._data = {}
self._ttl = default_ttl
self._touch = touch_on_access
self._max_items = max_items
self._expire_interval = expire_interval
if max_items is not None:
self._recent = _BoundedCacheSet(int(min(recent_items,max_items)))
else:
self._recent = None
reactor.callLater(self._expire_interval, self._expire)
def __getitem__(self, key): # O(1) if not touch and not newly expired, else O(log n)
"""Raises KeyError if the key is not in the cache.
This can happen if the entry was deleted or expired."""
ttl,v = self._data[key] # ttl is duration, not absolute time.
i = self._exp.find_key_by_value(key) # O(1). Key in exp is time. 'key' variable
# is exp's value. :-)
if i.at_end():
raise KeyError()
t = time()
if i.key() < t: # expired.
del self[key] # O(log n)
raise KeyError()
if self._recent:
self._recent.add(key)
if self._touch:
self._exp.update_key(i,t+ttl) # O(1) if no reordering else O(log n)
return v
def __setitem__(self, key, value): # O(log n). actually O(log n + RECENT_SIZE)
assert self._ttl > 0, "no default TTL defined. Perhaps the caller should call set " \
"rather than __setitem__."
t = time()
if self._data.has_key(key):
ttl,_ = self._data[key]
else:
ttl = self._ttl
self.set(key,value,ttl)
if self._recent:
self._recent.add(key)
# perform cache replacement if necessary.
if self._max_items is not None and len(self._data) > self._max_items:
to_remove = []
for t,k in self._exp.iteritems():
# worst case is O(RECENT_SIZE), but it is highly unlikely
# that all members of the recent access set are the oldest
# in the cache.
if k not in self._recent:
to_remove.append(k)
if len(to_remove) >= len(self._data) - self._max_items:
break
for k in to_remove:
del self[k]
def set(self, key, value, ttl):
"""Set using non-default TTL. ttl is a duration, not an absolute
time."""
t = time()
self._data[key] = (ttl, value)
i = self._exp.find_key_by_value(key)
if i.at_end():
self._exp[t+ttl] = key
else:
assert i.value() == key
self._exp.update_key(i,t+ttl)
def __delitem__(self, key): # O(log n)
del self._data[key]
i = self._exp.find_key_by_value(key)
if not i.at_end(): # No KeyError is generated if item is not in
# Cache because it could've been expired.
self._exp.erase(i)
def __len__(self):
"""Returns number of entries in the cache. Includes any
expired entries that haven't been removed yet.
Takes O(1) time."""
return len(self._data)
def num_unexpired(self):
"""Returns number of unexpired entries in the cache.
Any expired entries are removed before computing the length.
Takes worst case O(n) time where n = the number of expired
entries in the cache when this is called."""
self._expire2()
return len(self._data)
def has_key(self, key):
return self._data.has_key(key)
def __contains__(self, key):
return self._data.has_key(key)
def keys(self):
return self._data.keys()
def _expire(self):
self._expire2()
reactor.callLater(self._expire_interval, self._expire)
def _expire2(self):
t = time()
#try:
while True:
i = self._exp.begin()
if i.at_end():
break
if i.key() < t:
key = i.value()
self._exp.erase(i)
del self._data[key]
else:
break
assert len(self._data) == len(self._exp)
def __init__(self, max_items):
assert max_items > 1
self._max_items = max_items
self._data = IndexedMultiMap() # recent accesses.
class CacheMap:
"""this cache class allows caching with arbitrary expiration times. This is different
from BTL.Cache which assumes all items placed in the cache remain valid for the same
duration.
Like a BTL.Map, there can only be one instance of any given key in
the cache at a time. Subsequent inserts (__setitem__ or set)
with the same key, will update the ttl and value for that
key.
Unlike a BTL.Map, CacheMap does not perform in-order iteration based on key, and
key lookups (__getitem__) take average O(1) time.
The map also has the option to have bounded size in which case it imposes the
following replacement algorithm: remove the oldest entries first unless
those entries are in the recent access set. Here 'old' refers to duration
in the cache. Recent set has bounded size.
"""
# BTL.Cache places the cache entries in a queue. We instead maintain an
# IndexedMultiMap ordered based on expiration times. The index allows nodes in the
# map to be looked up in O(1) time based on value.
def __init__(self,
default_ttl=None,
expire_interval=60,
touch_on_access=False,
max_items=None,
recent_items=RECENT_SIZE):
"""
@param default_ttl: time to live when using __setitem__ rather than set.
@param expire_interval: time between removals of expired items in seconds. Otherwise,
expired items are removed lazily.
@param touch_on_access: refresh item expire time by ttl when item is accessed.
@param max_items: maximum size of cache. (see replacement algorithm above)
"""
self._exp = IndexedMultiMap(
) # expiration times. Multiple items can have the same expiration
# times, but there can only be one instance of any one key
# in the CacheMap.
self._data = {}
self._ttl = default_ttl
self._touch = touch_on_access
self._max_items = max_items
self._expire_interval = expire_interval
if max_items is not None:
self._recent = _BoundedCacheSet(int(min(recent_items, max_items)))
else:
self._recent = None
reactor.callLater(self._expire_interval, self._expire)
def __getitem__(
self,
key): # O(1) if not touch and not newly expired, else O(log n)
"""Raises KeyError if the key is not in the cache.
This can happen if the entry was deleted or expired."""
ttl, v = self._data[key] # ttl is duration, not absolute time.
i = self._exp.find_key_by_value(
key) # O(1). Key in exp is time. 'key' variable
# is exp's value. :-)
if i.at_end():
raise KeyError()
t = time()
if i.key() < t: # expired.
del self[key] # O(log n)
raise KeyError()
if self._recent:
self._recent.add(key)
if self._touch:
self._exp.update_key(i, t +
ttl) # O(1) if no reordering else O(log n)
return v
def __setitem__(self, key,
value): # O(log n). actually O(log n + RECENT_SIZE)
assert self._ttl > 0, "no default TTL defined. Perhaps the caller should call set " \
"rather than __setitem__."
t = time()
if self._data.has_key(key):
ttl, _ = self._data[key]
else:
ttl = self._ttl
self.set(key, value, ttl)
if self._recent:
self._recent.add(key)
# perform cache replacement if necessary.
if self._max_items is not None and len(self._data) > self._max_items:
to_remove = []
for t, k in self._exp.iteritems():
# worst case is O(RECENT_SIZE), but it is highly unlikely
# that all members of the recent access set are the oldest
# in the cache.
if k not in self._recent:
to_remove.append(k)
if len(to_remove) >= len(self._data) - self._max_items:
break
for k in to_remove:
del self[k]
def set(self, key, value, ttl):
"""Set using non-default TTL. ttl is a duration, not an absolute
time."""
t = time()
self._data[key] = (ttl, value)
i = self._exp.find_key_by_value(key)
if i.at_end():
self._exp[t + ttl] = key
else:
assert i.value() == key
self._exp.update_key(i, t + ttl)
def __delitem__(self, key): # O(log n)
del self._data[key]
i = self._exp.find_key_by_value(key)
if not i.at_end(): # No KeyError is generated if item is not in
# Cache because it could've been expired.
self._exp.erase(i)
def __len__(self):
"""Returns number of entries in the cache. Includes any
expired entries that haven't been removed yet.
Takes O(1) time."""
return len(self._data)
def num_unexpired(self):
"""Returns number of unexpired entries in the cache.
Any expired entries are removed before computing the length.
Takes worst case O(n) time where n = the number of expired
entries in the cache when this is called."""
self._expire2()
return len(self._data)
def has_key(self, key):
return self._data.has_key(key)
def __contains__(self, key):
return self._data.has_key(key)
def keys(self):
return self._data.keys()
def _expire(self):
self._expire2()
reactor.callLater(self._expire_interval, self._expire)
def _expire2(self):
t = time()
#try:
while True:
i = self._exp.begin()
if i.at_end():
break
if i.key() < t:
key = i.value()
self._exp.erase(i)
del self._data[key]
else:
break
assert len(self._data) == len(self._exp)
def __init__(self, max_items):
assert max_items > 1
self._max_items = max_items
self._data = IndexedMultiMap() # recent accesses.
