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)
55
# cleanup might raise an exception, but we want to make sure
56
# to break 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._least_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):
85
# Inlined from _record_access to decrease the overhead of __getitem__
86
# We also have more knowledge about structure if __getitem__ is
87
# succeeding, then we know that self._most_recently_used must not be
89
mru = self._most_recently_used
91
# Nothing to do, this node is already at the head of the queue
93
# Remove this node from the old location
95
next_key = node.next_key
96
# benchmarking shows that the lookup of _null_key in globals is faster
97
# than the attribute lookup for (node is self._least_recently_used)
98
if next_key is _null_key:
99
# 'node' is the _least_recently_used, because it doesn't have a
100
# 'next' item. So move the current lru to the previous node.
101
self._least_recently_used = node_prev
103
node_next = cache[next_key]
104
node_next.prev = node_prev
105
node_prev.next_key = next_key
106
# Insert this node at the front of the list
107
node.next_key = mru.key
109
self._most_recently_used = node
114
return len(self._cache)
117
"""Walk the LRU list, only meant to be used in tests."""
118
node = self._most_recently_used
120
if node.prev is not None:
121
raise AssertionError('the _most_recently_used entry is not'
122
' supposed to have a previous entry'
124
while node is not None:
125
if node.next_key is _null_key:
126
if node is not self._least_recently_used:
127
raise AssertionError('only the last node should have'
128
' no next value: %s' % (node,))
131
node_next = self._cache[node.next_key]
132
if node_next.prev is not node:
133
raise AssertionError('inconsistency found, node.next.prev'
134
' != node: %s' % (node,))
135
if node.prev is None:
136
if node is not self._most_recently_used:
137
raise AssertionError('only the _most_recently_used should'
138
' not have a previous node: %s'
141
if node.prev.next_key != node.key:
142
raise AssertionError('inconsistency found, node.prev.next'
143
' != node: %s' % (node,))
147
def add(self, key, value, cleanup=None):
148
"""Add a new value to the cache.
150
Also, if the entry is ever removed from the cache, call
153
:param key: The key to store it under
154
:param value: The object to store
155
:param cleanup: None or a function taking (key, value) to indicate
156
'value' should be cleaned up.
159
raise ValueError('cannot use _null_key as a key')
160
if key in self._cache:
161
node = self._cache[key]
165
# Maintain the LRU properties, even if cleanup raises an
168
node.cleanup = cleanup
169
self._record_access(node)
171
node = _LRUNode(key, value, cleanup=cleanup)
172
self._cache[key] = node
173
self._record_access(node)
175
if len(self._cache) > self._max_cache:
176
# Trigger the cleanup
179
def cache_size(self):
180
"""Get the number of entries we will cache."""
181
return self._max_cache
183
def get(self, key, default=None):
184
node = self._cache.get(key, None)
187
self._record_access(node)
191
"""Get the list of keys currently cached.
193
Note that values returned here may not be available by the time you
194
request them later. This is simply meant as a peak into the current
197
:return: An unordered list of keys that are currently cached.
199
return self._cache.keys()
202
"""Get the key:value pairs as a dict."""
203
return dict((k, n.value) for k, n in self._cache.iteritems())
206
"""Clear the cache until it shrinks to the requested size.
208
This does not completely wipe the cache, just makes sure it is under
209
the after_cleanup_count.
211
# Make sure the cache is shrunk to the correct size
212
while len(self._cache) > self._after_cleanup_count:
215
def __setitem__(self, key, value):
216
"""Add a value to the cache, there will be no cleanup function."""
217
self.add(key, value, cleanup=None)
219
def _record_access(self, node):
220
"""Record that key was accessed."""
221
# Move 'node' to the front of the queue
222
if self._most_recently_used is None:
223
self._most_recently_used = node
224
self._least_recently_used = node
226
elif node is self._most_recently_used:
227
# Nothing to do, this node is already at the head of the queue
229
# We've taken care of the tail pointer, remove the node, and insert it
232
if node is self._least_recently_used:
233
self._least_recently_used = node.prev
234
if node.prev is not None:
235
node.prev.next_key = node.next_key
236
if node.next_key is not _null_key:
237
node_next = self._cache[node.next_key]
238
node_next.prev = node.prev
240
node.next_key = self._most_recently_used.key
241
self._most_recently_used.prev = node
242
self._most_recently_used = node
245
def _remove_node(self, node):
246
if node is self._least_recently_used:
247
self._least_recently_used = node.prev
248
self._cache.pop(node.key)
249
# If we have removed all entries, remove the head pointer as well
250
if self._least_recently_used is None:
251
self._most_recently_used = None
255
# cleanup might raise an exception, but we want to make sure to
256
# maintain the linked list
257
if node.prev is not None:
258
node.prev.next_key = node.next_key
259
if node.next_key is not _null_key:
260
node_next = self._cache[node.next_key]
261
node_next.prev = node.prev
262
# And remove this node's pointers
264
node.next_key = _null_key
266
def _remove_lru(self):
267
"""Remove one entry from the lru, and handle consequences.
269
If there are no more references to the lru, then this entry should be
270
removed from the cache.
272
self._remove_node(self._least_recently_used)
275
"""Clear out all of the cache."""
276
# Clean up in LRU order
280
def resize(self, max_cache, after_cleanup_count=None):
281
"""Change the number of entries that will be cached."""
282
self._update_max_cache(max_cache,
283
after_cleanup_count=after_cleanup_count)
285
def _update_max_cache(self, max_cache, after_cleanup_count=None):
286
self._max_cache = max_cache
287
if after_cleanup_count is None:
288
self._after_cleanup_count = self._max_cache * 8 / 10
290
self._after_cleanup_count = min(after_cleanup_count,
295
class LRUSizeCache(LRUCache):
296
"""An LRUCache that removes things based on the size of the values.
298
This differs in that it doesn't care how many actual items there are,
299
it just restricts the cache to be cleaned up after so much data is stored.
301
The size of items added will be computed using compute_size(value), which
302
defaults to len() if not supplied.
305
def __init__(self, max_size=1024*1024, after_cleanup_size=None,
307
"""Create a new LRUSizeCache.
309
:param max_size: The max number of bytes to store before we start
310
clearing out entries.
311
:param after_cleanup_size: After cleaning up, shrink everything to this
313
:param compute_size: A function to compute the size of the values. We
314
use a function here, so that you can pass 'len' if you are just
315
using simple strings, or a more complex function if you are using
316
something like a list of strings, or even a custom object.
317
The function should take the form "compute_size(value) => integer".
318
If not supplied, it defaults to 'len()'
321
self._compute_size = compute_size
322
if compute_size is None:
323
self._compute_size = len
324
self._update_max_size(max_size, after_cleanup_size=after_cleanup_size)
325
LRUCache.__init__(self, max_cache=max(int(max_size/512), 1))
327
def add(self, key, value, cleanup=None):
328
"""Add a new value to the cache.
330
Also, if the entry is ever removed from the cache, call
333
:param key: The key to store it under
334
:param value: The object to store
335
:param cleanup: None or a function taking (key, value) to indicate
336
'value' should be cleaned up.
339
raise ValueError('cannot use _null_key as a key')
340
node = self._cache.get(key, None)
341
value_len = self._compute_size(value)
342
if value_len >= self._after_cleanup_size:
343
# The new value is 'too big to fit', as it would fill up/overflow
344
# the cache all by itself
345
trace.mutter('Adding the key %r to an LRUSizeCache failed.'
346
' value %d is too big to fit in a the cache'
347
' with size %d %d', key, value_len,
348
self._after_cleanup_size, self._max_size)
350
# We won't be replacing the old node, so just remove it
351
self._remove_node(node)
352
if cleanup is not None:
356
node = _LRUNode(key, value, cleanup=cleanup)
357
self._cache[key] = node
359
self._value_size -= node.size
360
node.size = value_len
361
self._value_size += value_len
362
self._record_access(node)
364
if self._value_size > self._max_size:
369
"""Clear the cache until it shrinks to the requested size.
371
This does not completely wipe the cache, just makes sure it is under
372
the after_cleanup_size.
374
# Make sure the cache is shrunk to the correct size
375
while self._value_size > self._after_cleanup_size:
378
def _remove_node(self, node):
379
self._value_size -= node.size
380
LRUCache._remove_node(self, node)
382
def resize(self, max_size, after_cleanup_size=None):
383
"""Change the number of bytes that will be cached."""
384
self._update_max_size(max_size, after_cleanup_size=after_cleanup_size)
385
max_cache = max(int(max_size/512), 1)
386
self._update_max_cache(max_cache)
388
def _update_max_size(self, max_size, after_cleanup_size=None):
389
self._max_size = max_size
390
if after_cleanup_size is None:
391
self._after_cleanup_size = self._max_size * 8 / 10
393
self._after_cleanup_size = min(after_cleanup_size, self._max_size)
added
removed
bzrlib/lru_cache.py
