From: http://aspn.activeState.com/aspn/cookbook/python/recipe/302997
Title: Thread-Safe Caching Object with File and http ustementations Submitter: Nicolas Lehuen (Other Recipes) Last Updated: 2006/02/08 Version No: 1.8 Category: Algorithms
Not rated yet
Description: Implementation of an abstract, thread-safe cache with minimal locking Four concrete implementations:.. A validating file cache, a validating HTTP cache, an experimental Python module cache and a function cache Plus, an abstract cache with weak references to its values . Source: Text source # - * - CODING: ISO-8859-1 - * -
From OS IMPORT STAT
From time import time, mktime
From RFC822 Import Parsedate
From Calendar Import Timegm
IMPORT URLLIB2
IMPORT RE
Import weakref
Import New
TRY:
From Threading Import Lock
Except Importerror:
From Dummy_threading Import Lock
NOT_INITIALIZED = Object ()
Class Entry (Object):
"" "A cache entry, mostly an internal object." "" "
DEF __INIT __ (Self, Key):
Object .__ init __ (self)
Self._key = key
Self._value = not_initialized
Self._lock = LOCK ()
Class Cache (Object):
"" "" "" "" "" ""
DEF __INIT __ (Self, Max_Size = 0):
"" "Builds a cache with a limit of max_size entries.
IF this limit is exceeded, The Least Recessly Used Entry is Discarded.
If max_size == 0, The Cache IS Unbounded (No Lru Rule Is Applied).
"" "
Object .__ init __ (self)
SELF._MAXSIZE = Max_size
SELF._DICT = {}
Self._lock = LOCK ()
# Hader of the access list
IF self._maxsize:
Self._head = entry (none)
Self._head._previous = self._head
Self._Head._Next = SELF._HEAD
DEF __SETITEM __ (Self, Name, Value):
"" "Populates the cache with a given name and value." "" "" "" "" ""
Entry = SELF._GET_ENTRY (Key)
Entry._lock.acquire ()
TRY:
SELF._PACK (Entry, Value)
SELF.COMMIT ()
Finally:
Entry._lock.release ()
DEF __GETITEM __ (Self, Name):
"" "Gets a value from the cache, builds it if need.
"" "
Return self._checkItem (name) [2]
DEF __DELITEM __ (Self, Name):
Self._lock.acquire ()
TRY:
Key = Self.Key (Name)
Del self._dict [key]
Finally:
Self._lock.release ()
DEF _GET_ENTRY (Self, Key):
Self._lock.acquire ()
TRY:
Entry = self._dict.get (key)
IF NOT ENTRY:
Entry = entry (key)
SELF._DICT [key] = Entry
IF self._maxsize:
Entry._next = entry._previous = none
SELF._ACCESS (Entry)
Self._checklru ()
Elif self._maxsize:
SELF._ACCESS (Entry)
Return Entry
Finally:
Self._lock.release ()
Def_CheckItem (Self, Name):
"" "Gets a value from the cache, builds it if need.
Returns a tuple is_new, key, value, entry.
IF is_new is true, the result had to be rebuilt.
"" "
Key = Self.Key (Name)
Entry = SELF._GET_ENTRY (Key)
Entry._lock.acquire ()
TRY:
Value = Self._unpack (entry)
IS_NEW = FALSE
IF value is not_initialized:
Opened = Self.check (key, name, entry)
Value = Self.Build (key, name, opened, entry)
IS_NEW = TRUE
SELF._PACK (Entry, Value)
SELF.COMMIT ()
Else:
Opened = Self.check (key, name, entry)
IF opened is not none:
Value = Self.Build (key, name, opened, entry)
IS_NEW = TRUE
SELF._PACK (Entry, Value)
SELF.COMMIT ()
Return IS_NEW, Key, Value, Entry
Finally:
Entry._lock.release ()
DEF MRU (Self):
"" Returns the MOSTLY USED Key "" "
IF self._maxsize: self._lock.acquire ()
TRY:
Return self._head._previous._key
Finally:
Self._lock.release ()
Else:
Return none
DEF LRU (Self):
"" Returns The Least Recessly Used Key "" "
IF self._maxsize:
Self._lock.acquire ()
TRY:
Return Self._Head._Next._Key
Finally:
Self._lock.release ()
Else:
Return none
Def key (self, name):
"" "" OVERRIDE this Method to Extract a key from the name passed to the [] operator "" "
Return Name
Def Commit (Self):
"" "OVERRIDE this Method if You Want to do Something Each Time The Underlying Dictionary is Modified (E.G. make it it penness)." ""
PASS
Def Clear:
"" Clears the cache "" "
Self._lock.acquire ()
TRY:
Self._dict.clear ()
IF self._maxsize:
Self._Head._Next = SELF._HEAD
Self._head._previous = self._head
Finally:
Self._lock.release ()
Def Check (Self, Key, Name, Entry):
"" "" OVERRIDE this Method to Check WHETHER THE Entry with the Given Name Is Stale. Return None if it is Fresh
OR An Opened Resource if it is stale. The Object Returned Will Be Passed to the 'build' Method as the 'Opened' parameter.
Use the 'entry' parameter to store meta-data if required. Don't worry About Multiple Threads accessing the Same Name,
as this method isotly isolated.
"" "
Return none
Def Build (Self, Key, Name, Opened, Entry):
"" "" Build The Cached Value with The Given Name from The Given Opened Resource. Use entry to ketain or store meta-data if needed.
Don't Worry About Multiple Threads Accessing The Same Name, AS this Method Is Properly isolated.
"" "
Raise NotimplementError ()
DEF _ACCESS (Self, Entry): "INTERNAL USE" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "
if Entry._Next Is Not Self._Head:
IF entry._previous is not none:
# REMOVE The entry from the access list
Entry._previous._next = entry._next
Entry._next._previous = entry._previous
# insert the entry at the end of the access list
Entry._previous = self._head._previous
Entry._previous._next = entry
Entry._next = Self._Head
Entry._next._previous = entry
If self._head._next is self._head:
Self._Head._Next = Entry
Def_Checklru (Self):
"" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "INTER
if Len (Self._Dict)> Self._MaxSize:
LRU = Self._Head._Next
Lru._previous._next = lru._next
Lru._next._previous = lru._previous
DEL SELF._DICT [Lru._Key]
Def _pack (Self, Entry, Value):
"" "" "" "" ""
Entry._value = Value
Def _unpack (self, entry):
"" "Recover The value from the entry, returns not_initialized if it is not ok." ""
Return Entry._Value
Class Weakcache (Cache):
"" "" This cache holds weak references to the value it stores. WHENEVER A Value IS Not Longer
Normally Reference, IT IS Removed from the cache. Useful for sharing the result of long
Computations But Letting Them Go As Soon as They Are Not Needed by Anybody.
"" "
Def _pack (Self, Entry, Value):
Entry._value = weakref.ref (value, lambda ref: self .__ delitem __ (entry._key))
Def _unpack (self, entry):
If entry._value is not_initialized:
Return NOT_INITIALIZED
Value = entry._value ()
IF value is none:
Return NOT_INITIALIZED
Else:
Return Value
Class FileCache (Cache): "" "A File Cache. Returns The Content of The Files As A String, Given Their FileName.
What Cache is Updated.
Override the build method to obtain more intending behaviour.
"" "
DEF __INIT __ (Self, Max_Size = 0, Mode = 'RB'):
Cache .__ init __ (self, max_size)
Self.Mode = MODE
Def Check (Self, Key, Name, Entry):
TimeStamp = Stat (key) .st_mtime
If entry._value is not_initialized:
Entry._timestamp = TimeStamp
Return File (key, self.mode)
Else:
IF entry._timestamp! = TimeStamp:
Entry._timestamp = TimeStamp
Return File (key, self.mode)
Else:
Return none
Def Build (Self, Key, Name, Opened, Entry):
"" "Return The Content of The File as a string. Override this for better behaviour." "" "
TRY:
Return OPENED.READ ()
Finally:
Opened.close ()
Def Parserfc822Time (T):
Return Mktime (ParSedate (T))
RE_MAX_AGE = Re.Compile ('max-agn / s * = / s * (/ d )', Re.i)
Class httpentity (Object):
DEF __INIT __ (Self, Entity, Metadata):
Self.entity = Entity
Self.metadata = metadata
DEF __REPR __ (Self):
Return 'httpentity (% s,% s)'% (Repr (Self.Entity), Self.Metadata
DEF __STR __ (Self):
Return Self.Antity
Class httpcache (cache):
"" "" AN http cache. Returns The Entity Found At The Given URL.
Uses Expires, etag and last-modified headers to minimize bandwidth usage.
Partial Cache-Control Support (Only Max-Age IS Support).
"" "
Def Check (Self, Key, Name, Entry):
Request = urllib2.request (key)
TRY:
IF Time () Return none Except AttributeError: PASS TRY: Header, Value = entry._validator Request.headers [header] = ValueExcept AttributeError: PASS Opened = none TRY: Opened = urllib2.urlopen (Request) Headers = OPENED.INFO () # xiration handling Expiration = false TRY: Match = re_max_age.match (Headers ['Cache-Control']) IF match: Entry._expires = time () int (Match.Group (1)) Expiration = true Except (KeyError, ValueError): PASS IF NOT Expiration: TRY: Date = Parserfc822Time (Headers ['Date']) Expires = Parserfc822Time (Headers ['Expires']) Entry._expires = time () (Expires-Date) Expiration = true Except Keyerror: PASS # Validator HANDLING VALIDATION = FALSE TRY: Entry._validator = 'if-none-match', Headers ['etc "] Validation = TRUE Except Keyerror: PASS IF not validation: TRY: Entry._validator = 'if-modified-Since', Headers ['last-modified'] Except Keyerror: PASS Return Opened Except Urllib2.httperror, Error: if Opened: Opened: Opened.Close () IF error.code == 304: Return none Else: Raise Error Def Build (Self, Key, Name, Opened, Entry): TRY: Return httpensity (Opened.Read (), Dict (Opened.info ()))) Finally: Opened.close () RE_NOT_WORD = Re.Compile (r '/ w ') Class Modulecache (Filecache): "" "A Module Cache. Give It A File Name, IT Returns A Module Which Results from the Execution of the Python Script It Contains. This module is not inserted Into sys.modules. "" " DEF __INIT __ (Self, Max_Size = 0): Filecache .__ init__ (self, max_size, 'r') Def Build (Self, Key, Name, Opened, Entry): TRY: Module = new.module (RE_NOT_WORD.SUB ('_', key)) Module .__ file__ = key EXEC OPENED IN Module .__ DICT__ Return ModuleFinally: Opened.close () Class httpmodulecache (httpcache): "" "" A Module Cache. Give It An HTTP URL, IT RETURns A Module Which Results from the Execution of the Python Script It Contains. This module is not inserted Into sys.modules. "" " DEF __INIT __ (Self, Max_Size = 0): Httpcache .__ init __ (self, max_size) Def Build (Self, Key, Name, Opened, Entry): TRY: Module = new.module (RE_NOT_WORD.SUB ('_', key)) Module .__ file__ = key TEXT = OPENED.READ (). Replace ('/ r / n', '/ n') Code = Compile (Text, Name, 'Exec') Exec code in module .__ DICT__ Return Module Finally: Opened.close () Class FunctionCache (Cache): DEF __INIT __ (Self, Function, Max_Size = 0): Cache .__ init __ (self, max_size) Self.function = function DEF __CALL __ (Self, * args, ** kW): IF kW: # a dict is not Hashable So We Build a Tuple of (Key, Value) PAIRS KW = tuple (kw.iterItems ()) Return Self [Args, KW] Else: Return Self [args, ()] Def Build (Self, Key, Name, Opened, Entry): Args, kw = key Return Self.Function (* args, ** Dict (kW)) Discussion: Two Years ago I Was A Definite Java Fan (Writing in this language since "). NOW i ' m in love with Python. The trouble is that I left behind quite a few useful classes I wrote (thread-safe caches, pools, etc.), so I had to reimplement them in Python.A cache is a pretty simple dictionary-like object: you provide it an index or a name, it gives you back an object for example, for an HTTP cache, the index is an URL, the object is the data you can fetch from the URL.The trick is that the corresponding. object can be quite expensive (in CPU, bandwitdh, time or memory) to build, so you have to balance between building the object every time you need it, or pre-building all the objects you could require, knowing that the target object can change in time (think of how an URL can point to different data over time). A cache is precisely a way to find a balance between these two extremities.This recipes provides you with an abstract Cache class, from which you can inherit, overriding And Build () Methods, And Four Specialisations. Filecache and Httpcache Are Quite WHA t their name describe. ModuleCache is an experimental specialisation of FileCache, which can come handy when playing with dynamic code, since it allow you to load any arbitrary file as a python module, and dynamically reload it each time the file is modified. FunctionCache is the good old function call cache (already presented many times in this Cookbook), with thread-safety included thanks to the Cache base class.Thread-safety of the cache is a must, since the purpose of a cache is to be shared a used By As Much Code As Possible ... a Multi-Threaded Application Such as a Web Application Server Has A Strong Need for Thread-Safe Cache and pool structures.for A Sample Usage of Fileche: >>> FC = Filecache (10) # 10 Files in Memory At MOST >>> F = Open ('Test.txt', 'W') >>> F.write ('Hello, World!') >>> f.close () >>> FC ['Test.txt'] 'Hello, World!' >>> FC ['Test.txt'] # this time the file is checked but not read'Hello, World! '>>> f = Open ('Test.txt', 'W') >>> F.Write ('Hello, ME!') >>> f.close () >>> FC ['Test.txt'] # this time the file is Checked and re-read'Hello, Me'A Sample Usage of httpcache: >>> hc = httpcache (1000) # maximum 1000 Documents in the cache >>> HC ['http://www.google.com/'] Httpensity (' [snipped] html>', ...) >>> HC ['http://www.google.com'] # The problem is, Google Don't want it.com Cached, So there is no gainhttpentity (' [snipped] ', ...) >>> HC ['http://www.google.com'] .metadata # That's why: no last -Modified, NO ETAG, NO EXPIRES Headers. {'Content-Length': '2360', 'Set-Cookie': 'pref = [snipped]; Expires = Sun, 17-JAN-2038 19:14:07 GMT; Path = /; domain = .google.fr ',' Server ':' GWS / 2.1 ',' Connection ':' Keep-Alive ',' Cache-Control ':' Private ',' Date ':' WED, 01 Sep 2004 21:21:50 GMT ',' Content-Type ':' Text / Html '} >>> HC [' http://diveintomark.org/xml/atom.xml '] httpensity (' XML Version = "1.0" Encoding = "UTF-8"?> [snipped] ', ...) >>> HC [' http://diveintomark.org/xml/atom.xml ' ] # The second call is much faster, Since Mark Put Some Cache Hint in Order To Save His Bandwidthhtt Prentity (' xml version = "1.0" encoding = "UTF-8"?> [snipped]', ...)> >> HC ['http://diveintomark.org/xml/atom.xml'] .metadata # here's why: nice expires, last-modified and etc Headers {'Content-length': '9785', 'Accept-Ranges ':' Bytes ',' Expires ':' THU, 02 Sep 2004 01:23:43 GMT ',' Vary ':' * ',' Server ':' Apache / 1.3.31 (Debian GNU / Linux), 'Last-Modified': 'WED, 01 SEP 2004 03:26:16 GMT', 'Connection': 'Close', 'ETAG': '"E80A6-2639-41354158"', 'Cache-Control': 'Max -AGE = 14400 ',' Date ':' WED, 01 Sep 2004 21:23:43 GMT ',' Content-Type ':' Application / XML '} Sample Usage of functioncache: >>> from time import sleep >> > def my_long_function (value): ... Sleep (5) ... Return Value 1 >>> my_long_function (2) # 5 seconds lat ... 3 >>> cached = functioncache (my_long_function, 10) # Keep The 10 Last Calls in Memory >>> Cached (2) # 5 SECONDS LATER ... 3 >>> Cached (2) # Immediate Answer3Add Comment Number of Comments: 4 Update, Nicolas Lehuen, 2004/09 / 07Changes: - added a WeakCache abstract class- all short member names such a entry._p, cache._d and so on have been renamed to longer, more explicit names Hopefully the code is more readable. now.- Cache.extract is renamed to Cache._unpack, and must return NOT_INITIALIZED if the entry is invalid. Cache._pack does the opposite of Cache._unpack. See WeakCache for an example use.- ModuleCache now return Module objects, which still Are PlaceHolder Classes But With A Better Repr (). Add CommentUpdate, Nicolas Lehuen, 2004/10 / 13IN __SETITITEM__, An Entry Lock Was Forgotten When Setting The value on an already existing entry.add Comment Update, Nicolas Lehuen, 2005/10 / 05Updated the code to the latest version I'm using The API for check () and build () has slightly changed:. The "key" parameter has been added The ModuleCache class now uses real. Module Objects INSTEAD OF Fake Ones.Add Comment Update, Nicolas Lehuen, 2006/02 / 08Compatibility improvements: this latest version is compatible with Python 2.2, and with Python versions which do not include thread support.Performance improvements: FileCache stat () the file to test if it is modified, and open () It online, it is it, previously it opened, Even if it isn't modified.add comment