1
# Copyright (C) 2006, 2008 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."""
25
class _LRUNode(object):
26
"""This maintains the linked-list which is the lru internals."""
28
__slots__ = ('prev', 'next', 'key', 'value', 'cleanup', 'size')
30
def __init__(self, key, value, cleanup=None):
35
self.cleanup = cleanup
36
# TODO: We could compute this 'on-the-fly' like we used to, and remove
37
# one pointer from this object, we just need to decide if it
38
# actually costs us much of anything in normal usage
50
return '%s(%r n:%r p:%r)' % (self.__class__.__name__, self.key,
53
def run_cleanup(self):
54
if self.cleanup is not None:
55
self.cleanup(self.key, self.value)
57
# Just make sure to break any refcycles, etc
61
class LRUCache(object):
62
"""A class which manages a cache of entries, removing unused ones."""
64
def __init__(self, max_cache=100, after_cleanup_count=None,
65
after_cleanup_size=symbol_versioning.DEPRECATED_PARAMETER):
66
if symbol_versioning.deprecated_passed(after_cleanup_size):
67
symbol_versioning.warn('LRUCache.__init__(after_cleanup_size) was'
68
' deprecated in 1.11. Use'
69
' after_cleanup_count instead.',
71
after_cleanup_count = after_cleanup_size
73
# The "HEAD" of the lru linked list
74
self._most_recently_used = None
75
# The "TAIL" of the lru linked list
76
self._last_recently_used = None
77
self._update_max_cache(max_cache, after_cleanup_count)
79
def __contains__(self, key):
80
return key in self._cache
82
def __getitem__(self, key):
83
node = self._cache[key]
84
# Inlined from _record_access to decrease the overhead of __getitem__
85
# We also have more knowledge about structure if __getitem__ is
86
# succeeding, then we know that self._most_recently_used must not be
88
mru = self._most_recently_used
90
# Nothing to do, this node is already at the head of the queue
92
elif node is self._last_recently_used:
93
self._last_recently_used = node.prev
94
# Remove this node from the old location
97
if node_prev is not None:
98
node_prev.next = node_next
99
if node_next is not None:
100
node_next.prev = node_prev
101
# Insert this node to the front of the list
104
self._most_recently_used = node
109
return len(self._cache)
112
"""Walk the LRU list, only meant to be used in tests."""
113
node = self._most_recently_used
115
if node.prev is not None:
116
raise AssertionError('the _most_recently_used entry is not'
117
' supposed to have a previous entry'
119
while node is not None:
120
if node.next is None:
121
if node is not self._last_recently_used:
122
raise AssertionError('only the last node should have'
123
' no next value: %s' % (node,))
125
if node.next.prev is not node:
126
raise AssertionError('inconsistency found, node.next.prev'
127
' != node: %s' % (node,))
128
if node.prev is None:
129
if node is not self._most_recently_used:
130
raise AssertionError('only the _most_recently_used should'
131
' not have a previous node: %s'
134
if node.prev.next is not node:
135
raise AssertionError('inconsistency found, node.prev.next'
136
' != node: %s' % (node,))
140
def add(self, key, value, cleanup=None):
141
"""Add a new value to the cache.
143
Also, if the entry is ever removed from the cache, call
146
:param key: The key to store it under
147
:param value: The object to store
148
:param cleanup: None or a function taking (key, value) to indicate
149
'value' should be cleaned up.
151
if key in self._cache:
152
node = self._cache[key]
155
node.cleanup = cleanup
157
node = _LRUNode(key, value, cleanup=cleanup)
158
self._cache[key] = node
159
self._record_access(node)
161
if len(self._cache) > self._max_cache:
162
# Trigger the cleanup
165
def cache_size(self):
166
"""Get the number of entries we will cache."""
167
return self._max_cache
169
def get(self, key, default=None):
170
node = self._cache.get(key, None)
173
self._record_access(node)
177
"""Get the list of keys currently cached.
179
Note that values returned here may not be available by the time you
180
request them later. This is simply meant as a peak into the current
183
:return: An unordered list of keys that are currently cached.
185
return self._cache.keys()
188
"""Get the key:value pairs as a dict."""
189
return dict((k, n.value) for k, n in self._cache.iteritems())
192
"""Clear the cache until it shrinks to the requested size.
194
This does not completely wipe the cache, just makes sure it is under
195
the after_cleanup_count.
197
# Make sure the cache is shrunk to the correct size
198
while len(self._cache) > self._after_cleanup_count:
201
def __setitem__(self, key, value):
202
"""Add a value to the cache, there will be no cleanup function."""
203
self.add(key, value, cleanup=None)
205
def _record_access(self, node):
206
"""Record that key was accessed."""
207
# Move 'node' to the front of the queue
208
if self._most_recently_used is None:
209
self._most_recently_used = node
210
self._last_recently_used = node
212
elif node is self._most_recently_used:
213
# Nothing to do, this node is already at the head of the queue
215
elif node is self._last_recently_used:
216
self._last_recently_used = node.prev
217
# We've taken care of the tail pointer, remove the node, and insert it
220
if node.prev is not None:
221
node.prev.next = node.next
222
if node.next is not None:
223
node.next.prev = node.prev
225
node.next = self._most_recently_used
226
self._most_recently_used = node
227
node.next.prev = node
230
def _remove_node(self, node):
231
if node is self._last_recently_used:
232
self._last_recently_used = node.prev
233
self._cache.pop(node.key)
234
# If we have removed all entries, remove the head pointer as well
235
if self._last_recently_used is None:
236
self._most_recently_used = None
238
# Now remove this node from the linked list
239
if node.prev is not None:
240
node.prev.next = node.next
241
if node.next is not None:
242
node.next.prev = node.prev
243
# And remove this node's pointers
247
def _remove_lru(self):
248
"""Remove one entry from the lru, and handle consequences.
250
If there are no more references to the lru, then this entry should be
251
removed from the cache.
253
self._remove_node(self._last_recently_used)
256
"""Clear out all of the cache."""
257
# Clean up in LRU order
261
def resize(self, max_cache, after_cleanup_count=None):
262
"""Change the number of entries that will be cached."""
263
self._update_max_cache(max_cache,
264
after_cleanup_count=after_cleanup_count)
266
def _update_max_cache(self, max_cache, after_cleanup_count=None):
267
self._max_cache = max_cache
268
if after_cleanup_count is None:
269
self._after_cleanup_count = self._max_cache * 8 / 10
271
self._after_cleanup_count = min(after_cleanup_count,
276
class LRUSizeCache(LRUCache):
277
"""An LRUCache that removes things based on the size of the values.
279
This differs in that it doesn't care how many actual items there are,
280
it just restricts the cache to be cleaned up after so much data is stored.
282
The size of items added will be computed using compute_size(value), which
283
defaults to len() if not supplied.
286
def __init__(self, max_size=1024*1024, after_cleanup_size=None,
288
"""Create a new LRUSizeCache.
290
:param max_size: The max number of bytes to store before we start
291
clearing out entries.
292
:param after_cleanup_size: After cleaning up, shrink everything to this
294
:param compute_size: A function to compute the size of the values. We
295
use a function here, so that you can pass 'len' if you are just
296
using simple strings, or a more complex function if you are using
297
something like a list of strings, or even a custom object.
298
The function should take the form "compute_size(value) => integer".
299
If not supplied, it defaults to 'len()'
302
self._compute_size = compute_size
303
if compute_size is None:
304
self._compute_size = len
305
self._update_max_size(max_size, after_cleanup_size=after_cleanup_size)
306
LRUCache.__init__(self, max_cache=max(int(max_size/512), 1))
308
def add(self, key, value, cleanup=None):
309
"""Add a new value to the cache.
311
Also, if the entry is ever removed from the cache, call
314
:param key: The key to store it under
315
:param value: The object to store
316
:param cleanup: None or a function taking (key, value) to indicate
317
'value' should be cleaned up.
319
node = self._cache.get(key, None)
320
value_len = self._compute_size(value)
321
if value_len >= self._after_cleanup_size:
322
# The new value is 'too big to fit', as it would fill up/overflow
323
# the cache all by itself
324
trace.mutter('Adding the key %r to an LRUSizeCache failed.'
325
' value %d is too big to fit in a the cache'
326
' with size %d %d', key, value_len,
327
self._after_cleanup_size, self._max_size)
329
# We won't be replacing the old node, so just remove it
330
self._remove_node(node)
331
if cleanup is not None:
335
node = _LRUNode(key, value, cleanup=cleanup)
336
self._cache[key] = node
338
self._value_size -= node.size
339
node.size = value_len
340
self._value_size += value_len
341
self._record_access(node)
343
if self._value_size > self._max_size:
348
"""Clear the cache until it shrinks to the requested size.
350
This does not completely wipe the cache, just makes sure it is under
351
the after_cleanup_size.
353
# Make sure the cache is shrunk to the correct size
354
while self._value_size > self._after_cleanup_size:
357
def _remove_node(self, node):
358
self._value_size -= node.size
359
LRUCache._remove_node(self, node)
361
def resize(self, max_size, after_cleanup_size=None):
362
"""Change the number of bytes that will be cached."""
363
self._update_max_size(max_size, after_cleanup_size=after_cleanup_size)
364
max_cache = max(int(max_size/512), 1)
365
self._update_max_cache(max_cache)
367
def _update_max_size(self, max_size, after_cleanup_size=None):
368
self._max_size = max_size
369
if after_cleanup_size is None:
370
self._after_cleanup_size = self._max_size * 8 / 10
372
self._after_cleanup_size = min(after_cleanup_size, self._max_size)
added
removed
bzrlib/lru_cache.py
