26
$\begingroup$

Using Chicago as an example;

According to the current rules for applying daylight saving time, at 02:00:00 on the first Sunday of November, clocks in Chicago go back 1 hour (from Central Daylight Time CDT to Central Standard Time CST).

In 2007, this transition occurred on the 4th of November. So that means, at 02:00:00 CDT on this date, clocks switched to 01:00:00 CST.

In[37]:= tzChicago = Entity["TimeZone", "America/Chicago"]

Out[37]= Entity["TimeZone", "America/Chicago"]

Chicago timezone entity

Let's try and construct DateObjects 1 second before 01:00:00 and at 01:00:00. (I cannot specify whether the latter is CDT or CST, but Mathematica should be able to handle it through the timezone specification.)

In[351]:= do005959CST = 
 DateObject[{2007, 11, 4, 0, 59, 59}, TimeZone -> tzChicago]

Out[351]= DateObject[{2007, 11, 4, 0, 59, 
  59}, "Instant", "Gregorian", "America/Chicago"]

In[352]:= do010000CST = 
 DateObject[{2007, 11, 4, 1, 0, 0.}, TimeZone -> tzChicago]

Out[352]= DateObject[{2007, 11, 4, 1, 0, 
  0.}, "Instant", "Gregorian", "America/Chicago"]

1s before 01:00

Notice, however, that Mathematica automatically assumes the latter time is CST, not CDT. But that means there is no way I can specify a time between 01:00:00 CDT and 02:00:00 CDT (a real interval of time that will be experienced by people living in Chicago, at the end of which, the clocks will be put back an hour). Any time given to Mathematica in this interval is automatically interpreted as CST, i.e. after the clocks have been put back.

I tried using an explicit timezone shift from GMT, which produces mixed results.

In[354]:= DateObject[{2007, 11, 4, 1, 0, 0.}, TimeZone -> -5]

Out[354]= DateObject[{2007, 11, 4, 1, 0, 
  0.}, "Instant", "Gregorian", -5.]

In[355]:= DateDifference[
 DateObject[{2007, 11, 4, 0, 59, 59}, "Instant", "Gregorian", 
  "America/Chicago"], 
 DateObject[{2007, 11, 4, 1, 0, 0.`}, "Instant", 
  "Gregorian", -5.`], "Seconds"]

Out[355]= Quantity[1., "Seconds"]

explicit timezone shift

So this looks correct. I appear to have managed to specify a time in the last hour of Chicago DST. But then this produces the following inconsistent result.

In[356]:= TimeZoneConvert[
 DateObject[{2007, 11, 4, 1, 0, 0.`}, "Instant", 
  "Gregorian", -5.`], tzChicago]

Out[356]= DateObject[{2007, 11, 4, 1, 0, 
  0.}, "Instant", "Gregorian", "America/Chicago"]

In[357]:= DateDifference[
 DateObject[{2007, 11, 4, 0, 59, 59}, "Instant", "Gregorian", 
  "America/Chicago"], 
 DateObject[{2007, 11, 4, 1, 0, 0.`}, "Instant", "Gregorian", 
  "America/Chicago"], "Seconds"]

Out[357]= Quantity[3601., "Seconds"]

timezone conversion

The application of TimeZoneConvert should return a point in time which is physically simultaneous with the input, yet we've now shifted an hour. Is this a bug or am I missing something?

$\endgroup$
7
  • 1
    $\begingroup$ Although the results of this code is best put in an image format, it would be helpful to have the code as it is easily copy & paste-able. If you can, please, provide such code for future users & those who wish to answer this question? $\endgroup$ – CA Trevillian Feb 6 at 8:48
  • $\begingroup$ Have added the code as requested. $\endgroup$ – RobertNathaniel Feb 6 at 13:08
  • $\begingroup$ Have you reported this? $\endgroup$ – CA Trevillian Feb 8 at 6:04
  • $\begingroup$ Yes. Will report back if I hear anything. In the meantime the only workaround I can see is to use hard-coded timezone shifts (i.e. versus GMT) and avoid the use of timezone entities and strings. $\endgroup$ – RobertNathaniel Feb 8 at 13:29
  • 3
    $\begingroup$ Got an acknowledgement from Wolfram. Looks like it has been accepted as an issue and has been forwarded to the developers. $\endgroup$ – RobertNathaniel Feb 11 at 13:36
22
$\begingroup$

I think this looks like a bug, though I have no authority to declare it. In particular, this looks like a hot mess:

Subtracting one second causes a UnixTime loss of 3601 seconds

That's the following:

tz = Entity["TimeZone", "America/Chicago"]
With[{t = DateObject[{2007, 11, 4, 1, 0, 0}, TimeZone -> tz]},
 {t, UnixTime[t]}]

{ DateObject[{2007, 11, 4, 1, 0, 0}, "Instant", "Gregorian", "America/Chicago"], 1194159600 }
With[{t = 
   DateObject[{2007, 11, 4, 1, 0, 0}, TimeZone -> tz] - 
    Quantity[1, "Seconds"]},
 {t, UnixTime[t]}]

{ DateObject[{2007, 11, 4, 0, 59, 59.`}, "Instant", "Gregorian", "America/Chicago"], 1194155999 }

Subtracting one second has caused a change of 3601 seconds in UnixTime space. I have so far been unable to imagine a possible world where this is correct. But this does rather suggest that date/time handling is terminally borked in Mathematica, which I struggle to believe.

$\endgroup$
1
  • 5
    $\begingroup$ Oh boy, this is pretty broken. If you instead travel forward in time 1 second, from 00:59:59 to 01:00:00 you get to be in both CDT & 3601 seconds ahead in UnixTime space! $\endgroup$ – CA Trevillian Feb 6 at 9:07

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.