1
# Copyright (C) 2006, 2008, 2009 Canonical Ltd
3
# This program is free software; you can redistribute it and/or modify
4
# it under the terms of the GNU General Public License as published by
5
# the Free Software Foundation; either version 2 of the License, or
6
# (at your option) any later version.
8
# This program is distributed in the hope that it will be useful,
9
# but WITHOUT ANY WARRANTY; without even the implied warranty of
10
# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
11
# GNU General Public License for more details.
13
# You should have received a copy of the GNU General Public License
14
# along with this program; if not, write to the Free Software
15
# Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
17
"""A simple least-recently-used (LRU) cache."""
19
from __future__ import absolute_import, division
33
class _LRUNode(object):
34
"""This maintains the linked-list which is the lru internals."""
36
__slots__ = ('prev', 'next_key', 'key', 'value')
38
def __init__(self, key, value):
40
self.next_key = _null_key
48
prev_key = self.prev.key
49
return '%s(%r n:%r p:%r)' % (self.__class__.__name__, self.key,
50
self.next_key, prev_key)
53
class LRUCache(object):
54
"""A class which manages a cache of entries, removing unused ones."""
56
def __init__(self, max_cache=100, after_cleanup_count=None):
58
# The "HEAD" of the lru linked list
59
self._most_recently_used = None
60
# The "TAIL" of the lru linked list
61
self._least_recently_used = None
62
self._update_max_cache(max_cache, after_cleanup_count)
64
def __contains__(self, key):
65
return key in self._cache
67
def __getitem__(self, key):
70
# Inlined from _record_access to decrease the overhead of __getitem__
71
# We also have more knowledge about structure if __getitem__ is
72
# succeeding, then we know that self._most_recently_used must not be
74
mru = self._most_recently_used
76
# Nothing to do, this node is already at the head of the queue
78
# Remove this node from the old location
80
next_key = node.next_key
81
# benchmarking shows that the lookup of _null_key in globals is faster
82
# than the attribute lookup for (node is self._least_recently_used)
83
if next_key is _null_key:
84
# 'node' is the _least_recently_used, because it doesn't have a
85
# 'next' item. So move the current lru to the previous node.
86
self._least_recently_used = node_prev
88
node_next = cache[next_key]
89
node_next.prev = node_prev
90
node_prev.next_key = next_key
91
# Insert this node at the front of the list
92
node.next_key = mru.key
94
self._most_recently_used = node
99
return len(self._cache)
101
def __setitem__(self, key, value):
102
"""Add a new value to the cache"""
104
raise ValueError('cannot use _null_key as a key')
105
if key in self._cache:
106
node = self._cache[key]
108
self._record_access(node)
110
node = _LRUNode(key, value)
111
self._cache[key] = node
112
self._record_access(node)
114
if len(self._cache) > self._max_cache:
115
# Trigger the cleanup
118
def cache_size(self):
119
"""Get the number of entries we will cache."""
120
return self._max_cache
122
def get(self, key, default=None):
123
node = self._cache.get(key, None)
126
self._record_access(node)
130
"""Get the list of keys currently cached.
132
Note that values returned here may not be available by the time you
133
request them later. This is simply meant as a peak into the current
136
:return: An unordered list of keys that are currently cached.
138
# GZ 2016-06-04: Maybe just make this return the view?
139
return list(viewkeys(self._cache))
142
"""Get a new dict with the same key:value pairs as the cache"""
143
return dict((k, n.value) for k, n in viewitems(self._cache))
146
"""Clear the cache until it shrinks to the requested size.
148
This does not completely wipe the cache, just makes sure it is under
149
the after_cleanup_count.
151
# Make sure the cache is shrunk to the correct size
152
while len(self._cache) > self._after_cleanup_count:
155
def _record_access(self, node):
156
"""Record that key was accessed."""
157
# Move 'node' to the front of the queue
158
if self._most_recently_used is None:
159
self._most_recently_used = node
160
self._least_recently_used = node
162
elif node is self._most_recently_used:
163
# Nothing to do, this node is already at the head of the queue
165
# We've taken care of the tail pointer, remove the node, and insert it
168
if node is self._least_recently_used:
169
self._least_recently_used = node.prev
170
if node.prev is not None:
171
node.prev.next_key = node.next_key
172
if node.next_key is not _null_key:
173
node_next = self._cache[node.next_key]
174
node_next.prev = node.prev
176
node.next_key = self._most_recently_used.key
177
self._most_recently_used.prev = node
178
self._most_recently_used = node
181
def _remove_node(self, node):
182
if node is self._least_recently_used:
183
self._least_recently_used = node.prev
184
self._cache.pop(node.key)
185
# If we have removed all entries, remove the head pointer as well
186
if self._least_recently_used is None:
187
self._most_recently_used = None
188
if node.prev is not None:
189
node.prev.next_key = node.next_key
190
if node.next_key is not _null_key:
191
node_next = self._cache[node.next_key]
192
node_next.prev = node.prev
193
# And remove this node's pointers
195
node.next_key = _null_key
197
def _remove_lru(self):
198
"""Remove one entry from the lru, and handle consequences.
200
If there are no more references to the lru, then this entry should be
201
removed from the cache.
203
self._remove_node(self._least_recently_used)
206
"""Clear out all of the cache."""
207
# Clean up in LRU order
211
def resize(self, max_cache, after_cleanup_count=None):
212
"""Change the number of entries that will be cached."""
213
self._update_max_cache(max_cache,
214
after_cleanup_count=after_cleanup_count)
216
def _update_max_cache(self, max_cache, after_cleanup_count=None):
217
self._max_cache = max_cache
218
if after_cleanup_count is None:
219
self._after_cleanup_count = self._max_cache * 8 // 10
221
self._after_cleanup_count = min(after_cleanup_count,
226
class LRUSizeCache(LRUCache):
227
"""An LRUCache that removes things based on the size of the values.
229
This differs in that it doesn't care how many actual items there are,
230
it just restricts the cache to be cleaned up after so much data is stored.
232
The size of items added will be computed using compute_size(value), which
233
defaults to len() if not supplied.
236
def __init__(self, max_size=1024 * 1024, after_cleanup_size=None,
238
"""Create a new LRUSizeCache.
240
:param max_size: The max number of bytes to store before we start
241
clearing out entries.
242
:param after_cleanup_size: After cleaning up, shrink everything to this
244
:param compute_size: A function to compute the size of the values. We
245
use a function here, so that you can pass 'len' if you are just
246
using simple strings, or a more complex function if you are using
247
something like a list of strings, or even a custom object.
248
The function should take the form "compute_size(value) => integer".
249
If not supplied, it defaults to 'len()'
252
self._compute_size = compute_size
253
if compute_size is None:
254
self._compute_size = len
255
self._update_max_size(max_size, after_cleanup_size=after_cleanup_size)
256
LRUCache.__init__(self, max_cache=max(int(max_size // 512), 1))
258
def __setitem__(self, key, value):
259
"""Add a new value to the cache"""
261
raise ValueError('cannot use _null_key as a key')
262
node = self._cache.get(key, None)
263
value_len = self._compute_size(value)
264
if value_len >= self._after_cleanup_size:
265
# The new value is 'too big to fit', as it would fill up/overflow
266
# the cache all by itself
267
trace.mutter('Adding the key %r to an LRUSizeCache failed.'
268
' value %d is too big to fit in a the cache'
269
' with size %d %d', key, value_len,
270
self._after_cleanup_size, self._max_size)
272
# We won't be replacing the old node, so just remove it
273
self._remove_node(node)
276
node = _LRUNode(key, value)
277
self._cache[key] = node
279
self._value_size -= self._compute_size(node.value)
280
self._value_size += value_len
281
self._record_access(node)
283
if self._value_size > self._max_size:
288
"""Clear the cache until it shrinks to the requested size.
290
This does not completely wipe the cache, just makes sure it is under
291
the after_cleanup_size.
293
# Make sure the cache is shrunk to the correct size
294
while self._value_size > self._after_cleanup_size:
297
def _remove_node(self, node):
298
self._value_size -= self._compute_size(node.value)
299
LRUCache._remove_node(self, node)
301
def resize(self, max_size, after_cleanup_size=None):
302
"""Change the number of bytes that will be cached."""
303
self._update_max_size(max_size, after_cleanup_size=after_cleanup_size)
304
max_cache = max(int(max_size // 512), 1)
305
self._update_max_cache(max_cache)
307
def _update_max_size(self, max_size, after_cleanup_size=None):
308
self._max_size = max_size
309
if after_cleanup_size is None:
310
self._after_cleanup_size = self._max_size * 8 // 10
312
self._after_cleanup_size = min(after_cleanup_size, self._max_size)
added
removed
breezy/lru_cache.py
