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."""
26
class _LRUNode(object):
27
"""This maintains the linked-list which is the lru internals."""
29
__slots__ = ('prev', 'next_key', 'key', 'value', 'cleanup', 'size')
31
def __init__(self, key, value, cleanup=None):
33
self.next_key = _null_key
36
self.cleanup = cleanup
37
# TODO: We could compute this 'on-the-fly' like we used to, and remove
38
# one pointer from this object, we just need to decide if it
39
# actually costs us much of anything in normal usage
46
prev_key = self.prev.key
47
return '%s(%r n:%r p:%r)' % (self.__class__.__name__, self.key,
48
self.next_key, prev_key)
50
def run_cleanup(self):
52
if self.cleanup is not None:
53
self.cleanup(self.key, self.value)
56
# Just make sure to break any refcycles, etc
60
class LRUCache(object):
61
"""A class which manages a cache of entries, removing unused ones."""
63
def __init__(self, max_cache=100, after_cleanup_count=None,
64
after_cleanup_size=symbol_versioning.DEPRECATED_PARAMETER):
65
if symbol_versioning.deprecated_passed(after_cleanup_size):
66
symbol_versioning.warn('LRUCache.__init__(after_cleanup_size) was'
67
' deprecated in 1.11. Use'
68
' after_cleanup_count instead.',
70
after_cleanup_count = after_cleanup_size
72
# The "HEAD" of the lru linked list
73
self._most_recently_used = None
74
# The "TAIL" of the lru linked list
75
self._least_recently_used = None
76
self._update_max_cache(max_cache, after_cleanup_count)
78
def __contains__(self, key):
79
return key in self._cache
81
def __getitem__(self, 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
# Remove this node from the old location
94
next_key = node.next_key
95
# benchmarking shows that the lookup of _null_key in globals is faster
96
# than the attribute lookup for (node is self._least_recently_used)
97
if next_key is _null_key:
98
# 'node' is the _least_recently_used, because it doesn't have a
99
# 'next' item. So move the current lru to the previous node.
100
self._least_recently_used = node_prev
102
node_next = cache[next_key]
103
node_next.prev = node_prev
104
node_prev.next_key = next_key
105
# Insert this node at the front of the list
106
node.next_key = mru.key
108
self._most_recently_used = node
113
return len(self._cache)
116
"""Walk the LRU list, only meant to be used in tests."""
117
node = self._most_recently_used
119
if node.prev is not None:
120
raise AssertionError('the _most_recently_used entry is not'
121
' supposed to have a previous entry'
123
while node is not None:
124
if node.next_key is _null_key:
125
if node is not self._least_recently_used:
126
raise AssertionError('only the last node should have'
127
' no next value: %s' % (node,))
130
node_next = self._cache[node.next_key]
131
if node_next.prev is not node:
132
raise AssertionError('inconsistency found, node.next.prev'
133
' != node: %s' % (node,))
134
if node.prev is None:
135
if node is not self._most_recently_used:
136
raise AssertionError('only the _most_recently_used should'
137
' not have a previous node: %s'
140
if node.prev.next_key != node.key:
141
raise AssertionError('inconsistency found, node.prev.next'
142
' != node: %s' % (node,))
146
def add(self, key, value, cleanup=None):
147
"""Add a new value to the cache.
149
Also, if the entry is ever removed from the cache, call
152
:param key: The key to store it under
153
:param value: The object to store
154
:param cleanup: None or a function taking (key, value) to indicate
155
'value' should be cleaned up.
158
raise ValueError('cannot use _null_key as a key')
159
if key in self._cache:
160
node = self._cache[key]
165
node.cleanup = cleanup
166
self._record_access(node)
168
node = _LRUNode(key, value, cleanup=cleanup)
169
self._cache[key] = node
170
self._record_access(node)
172
if len(self._cache) > self._max_cache:
173
# Trigger the cleanup
176
def cache_size(self):
177
"""Get the number of entries we will cache."""
178
return self._max_cache
180
def get(self, key, default=None):
181
node = self._cache.get(key, None)
184
self._record_access(node)
188
"""Get the list of keys currently cached.
190
Note that values returned here may not be available by the time you
191
request them later. This is simply meant as a peak into the current
194
:return: An unordered list of keys that are currently cached.
196
return self._cache.keys()
199
"""Get the key:value pairs as a dict."""
200
return dict((k, n.value) for k, n in self._cache.iteritems())
203
"""Clear the cache until it shrinks to the requested size.
205
This does not completely wipe the cache, just makes sure it is under
206
the after_cleanup_count.
208
# Make sure the cache is shrunk to the correct size
209
while len(self._cache) > self._after_cleanup_count:
212
def __setitem__(self, key, value):
213
"""Add a value to the cache, there will be no cleanup function."""
214
self.add(key, value, cleanup=None)
216
def _record_access(self, node):
217
"""Record that key was accessed."""
218
# Move 'node' to the front of the queue
219
if self._most_recently_used is None:
220
self._most_recently_used = node
221
self._least_recently_used = node
223
elif node is self._most_recently_used:
224
# Nothing to do, this node is already at the head of the queue
226
# We've taken care of the tail pointer, remove the node, and insert it
229
if node is self._least_recently_used:
230
self._least_recently_used = node.prev
231
if node.prev is not None:
232
node.prev.next_key = node.next_key
233
if node.next_key is not _null_key:
234
node_next = self._cache[node.next_key]
235
node_next.prev = node.prev
237
node.next_key = self._most_recently_used.key
238
self._most_recently_used.prev = node
239
self._most_recently_used = node
242
def _remove_node(self, node):
243
if node is self._least_recently_used:
244
self._least_recently_used = node.prev
245
self._cache.pop(node.key)
246
# If we have removed all entries, remove the head pointer as well
247
if self._least_recently_used is None:
248
self._most_recently_used = None
252
# Now remove this node from the linked list
253
if node.prev is not None:
254
node.prev.next_key = node.next_key
255
if node.next_key is not _null_key:
256
node_next = self._cache[node.next_key]
257
node_next.prev = node.prev
258
# And remove this node's pointers
260
node.next_key = _null_key
262
def _remove_lru(self):
263
"""Remove one entry from the lru, and handle consequences.
265
If there are no more references to the lru, then this entry should be
266
removed from the cache.
268
self._remove_node(self._least_recently_used)
271
"""Clear out all of the cache."""
272
# Clean up in LRU order
276
def resize(self, max_cache, after_cleanup_count=None):
277
"""Change the number of entries that will be cached."""
278
self._update_max_cache(max_cache,
279
after_cleanup_count=after_cleanup_count)
281
def _update_max_cache(self, max_cache, after_cleanup_count=None):
282
self._max_cache = max_cache
283
if after_cleanup_count is None:
284
self._after_cleanup_count = self._max_cache * 8 / 10
286
self._after_cleanup_count = min(after_cleanup_count,
291
class LRUSizeCache(LRUCache):
292
"""An LRUCache that removes things based on the size of the values.
294
This differs in that it doesn't care how many actual items there are,
295
it just restricts the cache to be cleaned up after so much data is stored.
297
The size of items added will be computed using compute_size(value), which
298
defaults to len() if not supplied.
301
def __init__(self, max_size=1024*1024, after_cleanup_size=None,
303
"""Create a new LRUSizeCache.
305
:param max_size: The max number of bytes to store before we start
306
clearing out entries.
307
:param after_cleanup_size: After cleaning up, shrink everything to this
309
:param compute_size: A function to compute the size of the values. We
310
use a function here, so that you can pass 'len' if you are just
311
using simple strings, or a more complex function if you are using
312
something like a list of strings, or even a custom object.
313
The function should take the form "compute_size(value) => integer".
314
If not supplied, it defaults to 'len()'
317
self._compute_size = compute_size
318
if compute_size is None:
319
self._compute_size = len
320
self._update_max_size(max_size, after_cleanup_size=after_cleanup_size)
321
LRUCache.__init__(self, max_cache=max(int(max_size/512), 1))
323
def add(self, key, value, cleanup=None):
324
"""Add a new value to the cache.
326
Also, if the entry is ever removed from the cache, call
329
:param key: The key to store it under
330
:param value: The object to store
331
:param cleanup: None or a function taking (key, value) to indicate
332
'value' should be cleaned up.
335
raise ValueError('cannot use _null_key as a key')
336
node = self._cache.get(key, None)
337
value_len = self._compute_size(value)
338
if value_len >= self._after_cleanup_size:
339
# The new value is 'too big to fit', as it would fill up/overflow
340
# the cache all by itself
341
trace.mutter('Adding the key %r to an LRUSizeCache failed.'
342
' value %d is too big to fit in a the cache'
343
' with size %d %d', key, value_len,
344
self._after_cleanup_size, self._max_size)
346
# We won't be replacing the old node, so just remove it
347
self._remove_node(node)
348
if cleanup is not None:
352
node = _LRUNode(key, value, cleanup=cleanup)
353
self._cache[key] = node
355
self._value_size -= node.size
356
node.size = value_len
357
self._value_size += value_len
358
self._record_access(node)
360
if self._value_size > self._max_size:
365
"""Clear the cache until it shrinks to the requested size.
367
This does not completely wipe the cache, just makes sure it is under
368
the after_cleanup_size.
370
# Make sure the cache is shrunk to the correct size
371
while self._value_size > self._after_cleanup_size:
374
def _remove_node(self, node):
375
self._value_size -= node.size
376
LRUCache._remove_node(self, node)
378
def resize(self, max_size, after_cleanup_size=None):
379
"""Change the number of bytes that will be cached."""
380
self._update_max_size(max_size, after_cleanup_size=after_cleanup_size)
381
max_cache = max(int(max_size/512), 1)
382
self._update_max_cache(max_cache)
384
def _update_max_size(self, max_size, after_cleanup_size=None):
385
self._max_size = max_size
386
if after_cleanup_size is None:
387
self._after_cleanup_size = self._max_size * 8 / 10
389
self._after_cleanup_size = min(after_cleanup_size, self._max_size)
added
removed
bzrlib/lru_cache.py
