Help writing test where bobobase_modification_time changes
I'll try to explain as briefly as possible. I've got a working cache pattern that keeps a cached version of a Zope file's content and only when it changes (by > bobobase_modification_time) the cache is invalidated and recreated. Here's the gist of my test class Test(ZopeTestCase.FunctionalTestCase): def test_recreateCache(self): """ bla """ self.folder.manage_addFile('foo.js','FOO data') url = self.folder.mytool.cacheFile('foo.js') # pseudo code this time path = urlparse(url)[2] response = self.publish(path) self.assert... tests on response... _file = getattr(self.folder, 'foo.js') _file.manage_edit(_file.title, _file.content_type, filedata='new data') newurl = self.folder.mytool.cacheFile('foo.js') assert newurl != url, "url hasn't changed :(" The reason it fails on the last line is that the time difference between the first time and the second time is nil. Adding a time.sleep(1) won't change anything because the bobobase_modification_time() doesn't change. How can I solve this? Make one big transaction two without having to use transaction.get().commit() which I don't think I can use in unit tests as per Dieter's advice. Grateful for help. Peter -- Peter Bengtsson, work www.fry-it.com home www.peterbe.com hobby www.issuetrackerproduct.com
Peter Bengtsson wrote at 2008-4-8 19:08 +0100:
... The reason it fails on the last line is that the time difference between the first time and the second time is nil. Adding a time.sleep(1) won't change anything because the bobobase_modification_time() doesn't change.
How can I solve this?
Not easy. "bobobase_modification_time()" is in fact a "DateTime" representation of "_p_serial". And "_p_serial" is the transaction id (also a timestamp) that has committed the object state belonging to this object incarnation. Unfortunately, you must not commit a transaction in a test suite (as this may cause interference with other tests). There is a small chance that a savepoint changed "_p_serial" -- but the chance is only small... You might be able to write "_p_serial" yourself and thereby trick "bobobase_modification_time()". -- Dieter
Peter Bengtsson wrote:
How can I solve this? Make one big transaction two without having to use transaction.get().commit() which I don't think I can use in unit tests as per Dieter's advice.
I'd suggest monkeypatching bobobase_modification_time on the objects you're testing with. You're testing your cache code, not the implementation of those objects, so replacing them with (partially) dummy objects is great for avoiding this kind of problem and is a pattern I've used frequently... cheers, Chris -- Simplistix - Content Management, Zope & Python Consulting - http://www.simplistix.co.uk
Thanks for the tips Chris and Dieter, I actually solved it in a very simple way: from OFS.Image import File from time import time class MockFile(File): """ This file works the same as a normal File expect that in manage_edit() the internal modification time is set immediately. Normally Zope objects get their internal modification set when the transaction manager commits. """ def manage_edit(self, title, content_type, precondition='', filedata=None, REQUEST=None): # here's the mock hack, we set the _p_mtime NOW instead of letting # the transaction manager do it at the end. self._p_mtime = time() return File.manage_edit(self, title, content_type, precondition=precondition, filedata=filedata, REQUEST=REQUEST) This worked beautifully in the integration test. I'm happy. On 10/04/2008, Chris Withers <chris@simplistix.co.uk> wrote:
Peter Bengtsson wrote:
How can I solve this? Make one big transaction two without having to use transaction.get().commit() which I don't think I can use in unit tests as per Dieter's advice.
I'd suggest monkeypatching bobobase_modification_time on the objects you're testing with. You're testing your cache code, not the implementation of those objects, so replacing them with (partially) dummy objects is great for avoiding this kind of problem and is a pattern I've used frequently...
cheers,
Chris
-- Simplistix - Content Management, Zope & Python Consulting - http://www.simplistix.co.uk
-- Peter Bengtsson, work www.fry-it.com home www.peterbe.com hobby www.issuetrackerproduct.com
participants (3)
-
Chris Withers -
Dieter Maurer -
Peter Bengtsson