1 #
2 # Copyright (c) 2001 Bizar Software Pty Ltd (http://www.bizarsoftware.com.au/)
3 # This module is free software, and you may redistribute it and/or modify
4 # under the same terms as Python, so long as this copyright message and
5 # disclaimer are retained in their original form.
6 #
7 # IN NO EVENT SHALL BIZAR SOFTWARE PTY LTD BE LIABLE TO ANY PARTY FOR
8 # DIRECT, INDIRECT, SPECIAL, INCIDENTAL, OR CONSEQUENTIAL DAMAGES ARISING
9 # OUT OF THE USE OF THIS CODE, EVEN IF THE AUTHOR HAS BEEN ADVISED OF THE
10 # POSSIBILITY OF SUCH DAMAGE.
11 #
12 # BIZAR SOFTWARE PTY LTD SPECIFICALLY DISCLAIMS ANY WARRANTIES, INCLUDING,
13 # BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS
14 # FOR A PARTICULAR PURPOSE. THE CODE PROVIDED HEREUNDER IS ON AN "AS IS"
15 # BASIS, AND THERE IS NO OBLIGATION WHATSOEVER TO PROVIDE MAINTENANCE,
16 # SUPPORT, UPDATES, ENHANCEMENTS, OR MODIFICATIONS.
17 #
18 #$Id: back_anydbm.py,v 1.29 2002-02-25 14:34:31 grubert Exp $
19 '''
20 This module defines a backend that saves the hyperdatabase in a database
21 chosen by anydbm. It is guaranteed to always be available in python
22 versions >2.1.1 (the dumbdbm fallback in 2.1.1 and earlier has several
23 serious bugs, and is not available)
24 '''
26 import whichdb, anydbm, os, marshal
27 from roundup import hyperdb, date, password
28 from blobfiles import FileStorage
30 #
31 # Now the database
32 #
33 class Database(FileStorage, hyperdb.Database):
34 """A database for storing records containing flexible data types.
36 Transaction stuff TODO:
37 . check the timestamp of the class file and nuke the cache if it's
38 modified. Do some sort of conflict checking on the dirty stuff.
39 . perhaps detect write collisions (related to above)?
41 """
42 def __init__(self, config, journaltag=None):
43 """Open a hyperdatabase given a specifier to some storage.
45 The 'storagelocator' is obtained from config.DATABASE.
46 The meaning of 'storagelocator' depends on the particular
47 implementation of the hyperdatabase. It could be a file name,
48 a directory path, a socket descriptor for a connection to a
49 database over the network, etc.
51 The 'journaltag' is a token that will be attached to the journal
52 entries for any edits done on the database. If 'journaltag' is
53 None, the database is opened in read-only mode: the Class.create(),
54 Class.set(), and Class.retire() methods are disabled.
55 """
56 self.config, self.journaltag = config, journaltag
57 self.dir = config.DATABASE
58 self.classes = {}
59 self.cache = {} # cache of nodes loaded or created
60 self.dirtynodes = {} # keep track of the dirty nodes by class
61 self.newnodes = {} # keep track of the new nodes by class
62 self.transactions = []
64 def __repr__(self):
65 return '<back_anydbm instance at %x>'%id(self)
67 #
68 # Classes
69 #
70 def __getattr__(self, classname):
71 """A convenient way of calling self.getclass(classname)."""
72 if self.classes.has_key(classname):
73 if hyperdb.DEBUG:
74 print '__getattr__', (self, classname)
75 return self.classes[classname]
76 raise AttributeError, classname
78 def addclass(self, cl):
79 if hyperdb.DEBUG:
80 print 'addclass', (self, cl)
81 cn = cl.classname
82 if self.classes.has_key(cn):
83 raise ValueError, cn
84 self.classes[cn] = cl
86 def getclasses(self):
87 """Return a list of the names of all existing classes."""
88 if hyperdb.DEBUG:
89 print 'getclasses', (self,)
90 l = self.classes.keys()
91 l.sort()
92 return l
94 def getclass(self, classname):
95 """Get the Class object representing a particular class.
97 If 'classname' is not a valid class name, a KeyError is raised.
98 """
99 if hyperdb.DEBUG:
100 print 'getclass', (self, classname)
101 return self.classes[classname]
103 #
104 # Class DBs
105 #
106 def clear(self):
107 '''Delete all database contents
108 '''
109 if hyperdb.DEBUG:
110 print 'clear', (self,)
111 for cn in self.classes.keys():
112 for type in 'nodes', 'journals':
113 path = os.path.join(self.dir, 'journals.%s'%cn)
114 if os.path.exists(path):
115 os.remove(path)
116 elif os.path.exists(path+'.db'): # dbm appends .db
117 os.remove(path+'.db')
119 def getclassdb(self, classname, mode='r'):
120 ''' grab a connection to the class db that will be used for
121 multiple actions
122 '''
123 if hyperdb.DEBUG:
124 print 'getclassdb', (self, classname, mode)
125 return self._opendb('nodes.%s'%classname, mode)
127 def _opendb(self, name, mode):
128 '''Low-level database opener that gets around anydbm/dbm
129 eccentricities.
130 '''
131 if hyperdb.DEBUG:
132 print '_opendb', (self, name, mode)
133 # determine which DB wrote the class file
134 db_type = ''
135 path = os.path.join(os.getcwd(), self.dir, name)
136 if os.path.exists(path):
137 db_type = whichdb.whichdb(path)
138 if not db_type:
139 raise hyperdb.DatabaseError, "Couldn't identify database type"
140 elif os.path.exists(path+'.db'):
141 # if the path ends in '.db', it's a dbm database, whether
142 # anydbm says it's dbhash or not!
143 db_type = 'dbm'
145 # new database? let anydbm pick the best dbm
146 if not db_type:
147 if hyperdb.DEBUG:
148 print "_opendb anydbm.open(%r, 'n')"%path
149 return anydbm.open(path, 'n')
151 # open the database with the correct module
152 try:
153 dbm = __import__(db_type)
154 except ImportError:
155 raise hyperdb.DatabaseError, \
156 "Couldn't open database - the required module '%s'"\
157 "is not available"%db_type
158 if hyperdb.DEBUG:
159 print "_opendb %r.open(%r, %r)"%(db_type, path, mode)
160 return dbm.open(path, mode)
162 #
163 # Nodes
164 #
165 def addnode(self, classname, nodeid, node):
166 ''' add the specified node to its class's db
167 '''
168 if hyperdb.DEBUG:
169 print 'addnode', (self, classname, nodeid, node)
170 self.newnodes.setdefault(classname, {})[nodeid] = 1
171 self.cache.setdefault(classname, {})[nodeid] = node
172 self.savenode(classname, nodeid, node)
174 def setnode(self, classname, nodeid, node):
175 ''' change the specified node
176 '''
177 if hyperdb.DEBUG:
178 print 'setnode', (self, classname, nodeid, node)
179 self.dirtynodes.setdefault(classname, {})[nodeid] = 1
180 # can't set without having already loaded the node
181 self.cache[classname][nodeid] = node
182 self.savenode(classname, nodeid, node)
184 def savenode(self, classname, nodeid, node):
185 ''' perform the saving of data specified by the set/addnode
186 '''
187 if hyperdb.DEBUG:
188 print 'savenode', (self, classname, nodeid, node)
189 self.transactions.append((self._doSaveNode, (classname, nodeid, node)))
191 def getnode(self, classname, nodeid, db=None, cache=1):
192 ''' get a node from the database
193 '''
194 if hyperdb.DEBUG:
195 print 'getnode', (self, classname, nodeid, db)
196 if cache:
197 # try the cache
198 cache = self.cache.setdefault(classname, {})
199 if cache.has_key(nodeid):
200 return cache[nodeid]
202 # get from the database and save in the cache
203 if db is None:
204 db = self.getclassdb(classname)
205 if not db.has_key(nodeid):
206 raise IndexError, "no such %s %s"%(classname, nodeid)
207 res = marshal.loads(db[nodeid])
208 if cache:
209 cache[nodeid] = res
210 return res
212 def hasnode(self, classname, nodeid, db=None):
213 ''' determine if the database has a given node
214 '''
215 if hyperdb.DEBUG:
216 print 'hasnode', (self, classname, nodeid, db)
217 # try the cache
218 cache = self.cache.setdefault(classname, {})
219 if cache.has_key(nodeid):
220 return 1
222 # not in the cache - check the database
223 if db is None:
224 db = self.getclassdb(classname)
225 res = db.has_key(nodeid)
226 return res
228 def countnodes(self, classname, db=None):
229 if hyperdb.DEBUG:
230 print 'countnodes', (self, classname, db)
231 # include the new nodes not saved to the DB yet
232 count = len(self.newnodes.get(classname, {}))
234 # and count those in the DB
235 if db is None:
236 db = self.getclassdb(classname)
237 count = count + len(db.keys())
238 return count
240 def getnodeids(self, classname, db=None):
241 if hyperdb.DEBUG:
242 print 'getnodeids', (self, classname, db)
243 # start off with the new nodes
244 res = self.newnodes.get(classname, {}).keys()
246 if db is None:
247 db = self.getclassdb(classname)
248 res = res + db.keys()
249 return res
252 #
253 # Files - special node properties
254 # inherited from FileStorage
256 #
257 # Journal
258 #
259 def addjournal(self, classname, nodeid, action, params):
260 ''' Journal the Action
261 'action' may be:
263 'create' or 'set' -- 'params' is a dictionary of property values
264 'link' or 'unlink' -- 'params' is (classname, nodeid, propname)
265 'retire' -- 'params' is None
266 '''
267 if hyperdb.DEBUG:
268 print 'addjournal', (self, classname, nodeid, action, params)
269 self.transactions.append((self._doSaveJournal, (classname, nodeid,
270 action, params)))
272 def getjournal(self, classname, nodeid):
273 ''' get the journal for id
274 '''
275 if hyperdb.DEBUG:
276 print 'getjournal', (self, classname, nodeid)
277 # attempt to open the journal - in some rare cases, the journal may
278 # not exist
279 try:
280 db = self._opendb('journals.%s'%classname, 'r')
281 except anydbm.error, error:
282 if str(error) == "need 'c' or 'n' flag to open new db": return []
283 elif error.args[0] != 2: raise
284 return []
285 journal = marshal.loads(db[nodeid])
286 res = []
287 for entry in journal:
288 (nodeid, date_stamp, user, action, params) = entry
289 date_obj = date.Date(date_stamp)
290 res.append((nodeid, date_obj, user, action, params))
291 return res
293 def pack(self, pack_before):
294 ''' delete all journal entries before 'pack_before' '''
295 if hyperdb.DEBUG:
296 print 'packjournal', (self, pack_before)
298 pack_before = pack_before.get_tuple()
300 classes = self.getclasses()
302 # TODO: factor this out to method - we're already doing it in
303 # _opendb.
304 db_type = ''
305 path = os.path.join(os.getcwd(), self.dir, classes[0])
306 if os.path.exists(path):
307 db_type = whichdb.whichdb(path)
308 if not db_type:
309 raise hyperdb.DatabaseError, "Couldn't identify database type"
310 elif os.path.exists(path+'.db'):
311 db_type = 'dbm'
313 for classname in classes:
314 db_name = 'journals.%s'%classname
315 db = self._opendb(db_name, 'w')
317 for key in db.keys():
318 journal = marshal.loads(db[key])
319 l = []
320 last_set_entry = None
321 for entry in journal:
322 (nodeid, date_stamp, self.journaltag, action,
323 params) = entry
324 if date_stamp > pack_before or action == 'create':
325 l.append(entry)
326 elif action == 'set':
327 # grab the last set entry to keep information on
328 # activity
329 last_set_entry = entry
330 if last_set_entry:
331 date_stamp = last_set_entry[1]
332 # if the last set entry was made after the pack date
333 # then it is already in the list
334 if date_stamp < pack_before:
335 l.append(last_set_entry)
336 db[key] = marshal.dumps(l)
337 if db_type == 'gdbm':
338 db.reorganize()
339 db.close()
342 #
343 # Basic transaction support
344 #
345 def commit(self):
346 ''' Commit the current transactions.
347 '''
348 if hyperdb.DEBUG:
349 print 'commit', (self,)
350 # TODO: lock the DB
352 # keep a handle to all the database files opened
353 self.databases = {}
355 # now, do all the transactions
356 for method, args in self.transactions:
357 method(*args)
359 # now close all the database files
360 for db in self.databases.values():
361 db.close()
362 del self.databases
363 # TODO: unlock the DB
365 # all transactions committed, back to normal
366 self.cache = {}
367 self.dirtynodes = {}
368 self.newnodes = {}
369 self.transactions = []
371 def _doSaveNode(self, classname, nodeid, node):
372 if hyperdb.DEBUG:
373 print '_doSaveNode', (self, classname, nodeid, node)
375 # get the database handle
376 db_name = 'nodes.%s'%classname
377 if self.databases.has_key(db_name):
378 db = self.databases[db_name]
379 else:
380 db = self.databases[db_name] = self.getclassdb(classname, 'c')
382 # now save the marshalled data
383 db[nodeid] = marshal.dumps(node)
385 def _doSaveJournal(self, classname, nodeid, action, params):
386 entry = (nodeid, date.Date().get_tuple(), self.journaltag, action,
387 params)
388 if hyperdb.DEBUG:
389 print '_doSaveJournal', entry
391 # get the database handle
392 db_name = 'journals.%s'%classname
393 if self.databases.has_key(db_name):
394 db = self.databases[db_name]
395 else:
396 db = self.databases[db_name] = self._opendb(db_name, 'c')
398 # now insert the journal entry
399 if db.has_key(nodeid):
400 s = db[nodeid]
401 l = marshal.loads(db[nodeid])
402 l.append(entry)
403 else:
404 l = [entry]
405 db[nodeid] = marshal.dumps(l)
407 def _doStoreFile(self, name, **databases):
408 # the file is currently ".tmp" - move it to its real name to commit
409 os.rename(name+".tmp", name)
411 def rollback(self):
412 ''' Reverse all actions from the current transaction.
413 '''
414 if hyperdb.DEBUG:
415 print 'rollback', (self, )
416 for method, args in self.transactions:
417 # delete temporary files
418 if method == self._doStoreFile:
419 if os.path.exists(args[0]+".tmp"):
420 os.remove(args[0]+".tmp")
421 self.cache = {}
422 self.dirtynodes = {}
423 self.newnodes = {}
424 self.transactions = []
426 #
427 #$Log: not supported by cvs2svn $
428 #Revision 1.28 2002/02/16 09:14:17 richard
429 # . #514854 ] History: "User" is always ticket creator
430 #
431 #Revision 1.27 2002/01/22 07:21:13 richard
432 #. fixed back_bsddb so it passed the journal tests
433 #
434 #... it didn't seem happy using the back_anydbm _open method, which is odd.
435 #Yet another occurrance of whichdb not being able to recognise older bsddb
436 #databases. Yadda yadda. Made the HYPERDBDEBUG stuff more sane in the
437 #process.
438 #
439 #Revision 1.26 2002/01/22 05:18:38 rochecompaan
440 #last_set_entry was referenced before assignment
441 #
442 #Revision 1.25 2002/01/22 05:06:08 rochecompaan
443 #We need to keep the last 'set' entry in the journal to preserve
444 #information on 'activity' for nodes.
445 #
446 #Revision 1.24 2002/01/21 16:33:20 rochecompaan
447 #You can now use the roundup-admin tool to pack the database
448 #
449 #Revision 1.23 2002/01/18 04:32:04 richard
450 #Rollback was breaking because a message hadn't actually been written to the file. Needs
451 #more investigation.
452 #
453 #Revision 1.22 2002/01/14 02:20:15 richard
454 # . changed all config accesses so they access either the instance or the
455 # config attriubute on the db. This means that all config is obtained from
456 # instance_config instead of the mish-mash of classes. This will make
457 # switching to a ConfigParser setup easier too, I hope.
458 #
459 #At a minimum, this makes migration a _little_ easier (a lot easier in the
460 #0.5.0 switch, I hope!)
461 #
462 #Revision 1.21 2002/01/02 02:31:38 richard
463 #Sorry for the huge checkin message - I was only intending to implement #496356
464 #but I found a number of places where things had been broken by transactions:
465 # . modified ROUNDUPDBSENDMAILDEBUG to be SENDMAILDEBUG and hold a filename
466 # for _all_ roundup-generated smtp messages to be sent to.
467 # . the transaction cache had broken the roundupdb.Class set() reactors
468 # . newly-created author users in the mailgw weren't being committed to the db
469 #
470 #Stuff that made it into CHANGES.txt (ie. the stuff I was actually working
471 #on when I found that stuff :):
472 # . #496356 ] Use threading in messages
473 # . detectors were being registered multiple times
474 # . added tests for mailgw
475 # . much better attaching of erroneous messages in the mail gateway
476 #
477 #Revision 1.20 2001/12/18 15:30:34 rochecompaan
478 #Fixed bugs:
479 # . Fixed file creation and retrieval in same transaction in anydbm
480 # backend
481 # . Cgi interface now renders new issue after issue creation
482 # . Could not set issue status to resolved through cgi interface
483 # . Mail gateway was changing status back to 'chatting' if status was
484 # omitted as an argument
485 #
486 #Revision 1.19 2001/12/17 03:52:48 richard
487 #Implemented file store rollback. As a bonus, the hyperdb is now capable of
488 #storing more than one file per node - if a property name is supplied,
489 #the file is called designator.property.
490 #I decided not to migrate the existing files stored over to the new naming
491 #scheme - the FileClass just doesn't specify the property name.
492 #
493 #Revision 1.18 2001/12/16 10:53:38 richard
494 #take a copy of the node dict so that the subsequent set
495 #operation doesn't modify the oldvalues structure
496 #
497 #Revision 1.17 2001/12/14 23:42:57 richard
498 #yuck, a gdbm instance tests false :(
499 #I've left the debugging code in - it should be removed one day if we're ever
500 #_really_ anal about performace :)
501 #
502 #Revision 1.16 2001/12/12 03:23:14 richard
503 #Cor blimey this anydbm/whichdb stuff is yecchy. Turns out that whichdb
504 #incorrectly identifies a dbm file as a dbhash file on my system. This has
505 #been submitted to the python bug tracker as issue #491888:
506 #https://sourceforge.net/tracker/index.php?func=detail&aid=491888&group_id=5470&atid=105470
507 #
508 #Revision 1.15 2001/12/12 02:30:51 richard
509 #I fixed the problems with people whose anydbm was using the dbm module at the
510 #backend. It turns out the dbm module modifies the file name to append ".db"
511 #and my check to determine if we're opening an existing or new db just
512 #tested os.path.exists() on the filename. Well, no longer! We now perform a
513 #much better check _and_ cope with the anydbm implementation module changing
514 #too!
515 #I also fixed the backends __init__ so only ImportError is squashed.
516 #
517 #Revision 1.14 2001/12/10 22:20:01 richard
518 #Enabled transaction support in the bsddb backend. It uses the anydbm code
519 #where possible, only replacing methods where the db is opened (it uses the
520 #btree opener specifically.)
521 #Also cleaned up some change note generation.
522 #Made the backends package work with pydoc too.
523 #
524 #Revision 1.13 2001/12/02 05:06:16 richard
525 #. We now use weakrefs in the Classes to keep the database reference, so
526 # the close() method on the database is no longer needed.
527 # I bumped the minimum python requirement up to 2.1 accordingly.
528 #. #487480 ] roundup-server
529 #. #487476 ] INSTALL.txt
530 #
531 #I also cleaned up the change message / post-edit stuff in the cgi client.
532 #There's now a clearly marked "TODO: append the change note" where I believe
533 #the change note should be added there. The "changes" list will obviously
534 #have to be modified to be a dict of the changes, or somesuch.
535 #
536 #More testing needed.
537 #
538 #Revision 1.12 2001/12/01 07:17:50 richard
539 #. We now have basic transaction support! Information is only written to
540 # the database when the commit() method is called. Only the anydbm
541 # backend is modified in this way - neither of the bsddb backends have been.
542 # The mail, admin and cgi interfaces all use commit (except the admin tool
543 # doesn't have a commit command, so interactive users can't commit...)
544 #. Fixed login/registration forwarding the user to the right page (or not,
545 # on a failure)
546 #
547 #Revision 1.11 2001/11/21 02:34:18 richard
548 #Added a target version field to the extended issue schema
549 #
550 #Revision 1.10 2001/10/09 23:58:10 richard
551 #Moved the data stringification up into the hyperdb.Class class' get, set
552 #and create methods. This means that the data is also stringified for the
553 #journal call, and removes duplication of code from the backends. The
554 #backend code now only sees strings.
555 #
556 #Revision 1.9 2001/10/09 07:25:59 richard
557 #Added the Password property type. See "pydoc roundup.password" for
558 #implementation details. Have updated some of the documentation too.
559 #
560 #Revision 1.8 2001/09/29 13:27:00 richard
561 #CGI interfaces now spit up a top-level index of all the instances they can
562 #serve.
563 #
564 #Revision 1.7 2001/08/12 06:32:36 richard
565 #using isinstance(blah, Foo) now instead of isFooType
566 #
567 #Revision 1.6 2001/08/07 00:24:42 richard
568 #stupid typo
569 #
570 #Revision 1.5 2001/08/07 00:15:51 richard
571 #Added the copyright/license notice to (nearly) all files at request of
572 #Bizar Software.
573 #
574 #Revision 1.4 2001/07/30 01:41:36 richard
575 #Makes schema changes mucho easier.
576 #
577 #Revision 1.3 2001/07/25 01:23:07 richard
578 #Added the Roundup spec to the new documentation directory.
579 #
580 #Revision 1.2 2001/07/23 08:20:44 richard
581 #Moved over to using marshal in the bsddb and anydbm backends.
582 #roundup-admin now has a "freshen" command that'll load/save all nodes (not
583 # retired - mod hyperdb.Class.list() so it lists retired nodes)
584 #
585 #