Thanks Lupa. I have tracked it down and created a patch which fixes the problem in general. Seems the ISO 8601 implementation had some warts. We are testing the patch now. When I am sure it's OK I'll submit it to the Collector with an "URGENT" tag. I'll also send you a copy. I suspect that the daylight savings time behavior is unrelated to this problem and is likely to be a bad entry in a table. Thanks for you helpful response. BTW if you want an alpha copy of the patch, let me know and I'll send it to you. The more testing the better. The patched version has the following semantics dates and times with no timezone tag resolve to the timezone of the host machine (Zope standard behavior) dates and times with a trailing Z are guaranteed to be GMT dates and times with a trailing offset have the offset as their time zone. It is illegal to combine a 'Z' and an offset. (Hacked in the code for 2.7 beta) I think this matches the spec at http://www.w3.org/TR/NOTE-datetime On Sun, 12 Jun 2005, +lupa+ wrote:
Hi Dennis, I'm responding directly to you as I'm not on the zope list. I see the same error on Zope 2.7.6 and with 2.3.5 python, not 2.4.x. I've narrowed the behavior down a bit more and it seems that
dd = '%d-%02d-%02d' % (y,m,d,) D = DateTime(dd)
converts to local time, while
ee = '%d/%d/%d' % (m, d, y) E = DateTime(ee)
converts to Mean time. I'm not absolutely sure of that, but in a quick look, I can see that the difference between D and E changes from -5 hours to -3 when Daylight Savings Time change comes around... which seems a bit weird in itself. I wrote and maintain CalendarX for Plone, so this behavior concerns me greatly and I'll be looking into it more very soon.
+lupa+ Lupa Zurven, CalendarX.org
At 12:00 PM 6/12/2005 -0400, you wrote:
From: Dennis Allison <allison@shasta.stanford.edu> Precedence: list Subject: [Zope] Datetime bug in Zope-2.7.6? Date: Sat, 11 Jun 2005 16:54:03 -0700 (PDT) To: zope@zope.org Message-ID: <Pine.LNX.4.44.0506111632260.3493-100000@shasta.stanford.edu> Content-Type: TEXT/PLAIN; charset=US-ASCII MIME-Version: 1.0 Message: 5
Zope 2.7.6 Python 2.4.1 (and yes, I know I am jumping the gun a bit)
The problem appears to be in the parse component --
Consider a test script:
for y in range(1995,2005): for d in (1, 3, 13, 25,): for m in (1,4,5,6,): dd = '%d-%02d-%02d' % (y,m,d,) ee = '%d/%d/%d' % (m, d, y) D = DateTime(dd) E = DateTime(ee) print dd, D.strftime('%Y-%m-%d') print ee, E.strftime('%Y-%m-%d') return printed
In all cases, each of the dates shoudl be the same. As can be seen, for dates sourced from a string of the form
1995-04-25
the resulatant date is off by one day
1995-04-24
The problem is not seen with Zope 2.6.4 running with Python 2.3.5.
A quick diff shows tht there are many changes between 2.6 and 2.7. Anyone else seen this problem?
------------------------------------ Partial Results:
1995-04-25 1995-04-24 4/25/1995 1995-04-25
1995-05-25 1995-05-24 5/25/1995 1995-05-25
1995-06-25 1995-06-24 6/25/1995 1995-06-25
-- Dennis Allison * Computer Systems Laboratory * Gates 227 * Stanford University * Stanford CA 94305 * (650) 723-9213 * (650) 723-0033 fax * allison@shasta.stanford.edu * allison@sumeru.stanford.edu
-- Dennis Allison * Computer Systems Laboratory * Gates 227 * Stanford University * Stanford CA 94305 * (650) 723-9213 * (650) 723-0033 fax * allison@shasta.stanford.edu * allison@sumeru.stanford.edu