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
29
class _LRUNode(object):
30
"""This maintains the linked-list which is the lru internals."""
32
__slots__ = ('prev', 'next_key', 'key', 'value')
34
def __init__(self, key, value):
36
self.next_key = _null_key
44
prev_key = self.prev.key
45
return '%s(%r n:%r p:%r)' % (self.__class__.__name__, self.key,
46
self.next_key, prev_key)
49
class LRUCache(object):
50
"""A class which manages a cache of entries, removing unused ones."""
52
def __init__(self, max_cache=100, after_cleanup_count=None):
54
# The "HEAD" of the lru linked list
55
self._most_recently_used = None
56
# The "TAIL" of the lru linked list
57
self._least_recently_used = None
58
self._update_max_cache(max_cache, after_cleanup_count)
60
def __contains__(self, key):
61
return key in self._cache
63
def __getitem__(self, key):
66
# Inlined from _record_access to decrease the overhead of __getitem__
67
# We also have more knowledge about structure if __getitem__ is
68
# succeeding, then we know that self._most_recently_used must not be
70
mru = self._most_recently_used
72
# Nothing to do, this node is already at the head of the queue
74
# Remove this node from the old location
76
next_key = node.next_key
77
# benchmarking shows that the lookup of _null_key in globals is faster
78
# than the attribute lookup for (node is self._least_recently_used)
79
if next_key is _null_key:
80
# 'node' is the _least_recently_used, because it doesn't have a
81
# 'next' item. So move the current lru to the previous node.
82
self._least_recently_used = node_prev
84
node_next = cache[next_key]
85
node_next.prev = node_prev
86
node_prev.next_key = next_key
87
# Insert this node at the front of the list
88
node.next_key = mru.key
90
self._most_recently_used = node
95
return len(self._cache)
97
def __setitem__(self, key, value):
98
"""Add a new value to the cache"""
100
raise ValueError('cannot use _null_key as a key')
101
if key in self._cache:
102
node = self._cache[key]
104
self._record_access(node)
106
node = _LRUNode(key, value)
107
self._cache[key] = node
108
self._record_access(node)
110
if len(self._cache) > self._max_cache:
111
# Trigger the cleanup
114
def cache_size(self):
115
"""Get the number of entries we will cache."""
116
return self._max_cache
118
def get(self, key, default=None):
119
node = self._cache.get(key, None)
122
self._record_access(node)
126
"""Get the list of keys currently cached.
128
Note that values returned here may not be available by the time you
129
request them later. This is simply meant as a peak into the current
132
:return: An unordered list of keys that are currently cached.
134
# GZ 2016-06-04: Maybe just make this return the view?
135
return list(self._cache.keys())
138
"""Get a new dict with the same key:value pairs as the cache"""
139
return dict((k, n.value) for k, n in self._cache.items())
142
"""Clear the cache until it shrinks to the requested size.
144
This does not completely wipe the cache, just makes sure it is under
145
the after_cleanup_count.
147
# Make sure the cache is shrunk to the correct size
148
while len(self._cache) > self._after_cleanup_count:
151
def _record_access(self, node):
152
"""Record that key was accessed."""
153
# Move 'node' to the front of the queue
154
if self._most_recently_used is None:
155
self._most_recently_used = node
156
self._least_recently_used = node
158
elif node is self._most_recently_used:
159
# Nothing to do, this node is already at the head of the queue
161
# We've taken care of the tail pointer, remove the node, and insert it
164
if node is self._least_recently_used:
165
self._least_recently_used = node.prev
166
if node.prev is not None:
167
node.prev.next_key = node.next_key
168
if node.next_key is not _null_key:
169
node_next = self._cache[node.next_key]
170
node_next.prev = node.prev
172
node.next_key = self._most_recently_used.key
173
self._most_recently_used.prev = node
174
self._most_recently_used = node
177
def _remove_node(self, node):
178
if node is self._least_recently_used:
179
self._least_recently_used = node.prev
180
self._cache.pop(node.key)
181
# If we have removed all entries, remove the head pointer as well
182
if self._least_recently_used is None:
183
self._most_recently_used = None
184
if node.prev is not None:
185
node.prev.next_key = node.next_key
186
if node.next_key is not _null_key:
187
node_next = self._cache[node.next_key]
188
node_next.prev = node.prev
189
# And remove this node's pointers
191
node.next_key = _null_key
193
def _remove_lru(self):
194
"""Remove one entry from the lru, and handle consequences.
196
If there are no more references to the lru, then this entry should be
197
removed from the cache.
199
self._remove_node(self._least_recently_used)
202
"""Clear out all of the cache."""
203
# Clean up in LRU order
207
def resize(self, max_cache, after_cleanup_count=None):
208
"""Change the number of entries that will be cached."""
209
self._update_max_cache(max_cache,
210
after_cleanup_count=after_cleanup_count)
212
def _update_max_cache(self, max_cache, after_cleanup_count=None):
213
self._max_cache = max_cache
214
if after_cleanup_count is None:
215
self._after_cleanup_count = self._max_cache * 8 // 10
217
self._after_cleanup_count = min(after_cleanup_count,
222
class LRUSizeCache(LRUCache):
223
"""An LRUCache that removes things based on the size of the values.
225
This differs in that it doesn't care how many actual items there are,
226
it just restricts the cache to be cleaned up after so much data is stored.
228
The size of items added will be computed using compute_size(value), which
229
defaults to len() if not supplied.
232
def __init__(self, max_size=1024 * 1024, after_cleanup_size=None,
234
"""Create a new LRUSizeCache.
236
:param max_size: The max number of bytes to store before we start
237
clearing out entries.
238
:param after_cleanup_size: After cleaning up, shrink everything to this
240
:param compute_size: A function to compute the size of the values. We
241
use a function here, so that you can pass 'len' if you are just
242
using simple strings, or a more complex function if you are using
243
something like a list of strings, or even a custom object.
244
The function should take the form "compute_size(value) => integer".
245
If not supplied, it defaults to 'len()'
248
self._compute_size = compute_size
249
if compute_size is None:
250
self._compute_size = len
251
self._update_max_size(max_size, after_cleanup_size=after_cleanup_size)
252
LRUCache.__init__(self, max_cache=max(int(max_size // 512), 1))
254
def __setitem__(self, key, value):
255
"""Add a new value to the cache"""
257
raise ValueError('cannot use _null_key as a key')
258
node = self._cache.get(key, None)
259
value_len = self._compute_size(value)
260
if value_len >= self._after_cleanup_size:
261
# The new value is 'too big to fit', as it would fill up/overflow
262
# the cache all by itself
263
trace.mutter('Adding the key %r to an LRUSizeCache failed.'
264
' value %d is too big to fit in a the cache'
265
' with size %d %d', key, value_len,
266
self._after_cleanup_size, self._max_size)
268
# We won't be replacing the old node, so just remove it
269
self._remove_node(node)
272
node = _LRUNode(key, value)
273
self._cache[key] = node
275
self._value_size -= self._compute_size(node.value)
276
self._value_size += value_len
277
self._record_access(node)
279
if self._value_size > self._max_size:
284
"""Clear the cache until it shrinks to the requested size.
286
This does not completely wipe the cache, just makes sure it is under
287
the after_cleanup_size.
289
# Make sure the cache is shrunk to the correct size
290
while self._value_size > self._after_cleanup_size:
293
def _remove_node(self, node):
294
self._value_size -= self._compute_size(node.value)
295
LRUCache._remove_node(self, node)
297
def resize(self, max_size, after_cleanup_size=None):
298
"""Change the number of bytes that will be cached."""
299
self._update_max_size(max_size, after_cleanup_size=after_cleanup_size)
300
max_cache = max(int(max_size // 512), 1)
301
self._update_max_cache(max_cache)
303
def _update_max_size(self, max_size, after_cleanup_size=None):
304
self._max_size = max_size
305
if after_cleanup_size is None:
306
self._after_cleanup_size = self._max_size * 8 // 10
308
self._after_cleanup_size = min(after_cleanup_size, self._max_size)
added
removed
breezy/lru_cache.py
