Created on 2003-01-02 14:23 by amaury.forgeotdarc, last changed 2003-01-02 16:37 by tim.peters. This issue is now closed.
|msg13805 - (view)||Author: Amaury Forgeot d'Arc (amaury.forgeotdarc) *||Date: 2003-01-02 14:23|
While exploring the new datetime module, I got results like: >>> datetime.datetime.today() datetime.datetime(2003, 1, 2, 15, 14, 51, 480999) Shouldn't the microseconds be 637000 ? While not very important, this can lead you to believe that the timestamp is accurate to 1e-6 seconds, whereas only milliseconds are relevant (on W2K). since today() is equivalent to fromtimestamp(time.time()), I explored the results of time.time(): >>> time.time(),datetime.datetime.today() returns: (1041516891.4809999, datetime.datetime(2003, 1, 2, 15, 14, 51, 480999)) It seems that the float returned by time.time is truncated to 1e-6. I suggest that it should be rounded instead, so that the last digit (which is significant) is taken into account. Looking into the code, I think that the code to change is in datetimemodule.c: the function datetime_from_timestamp could add 5e-7 to the "us" variable before casting it to int.
|msg13806 - (view)||Author: Amaury Forgeot d'Arc (amaury.forgeotdarc) *||Date: 2003-01-02 15:18|
Logged In: YES user_id=389140 Sorry for the typo: in my sample, I would like the microseconds to show as 481000 of course.
|msg13807 - (view)||Author: Tim Peters (tim.peters) *||Date: 2003-01-02 16:37|
Logged In: YES user_id=31435 I agree it's an irritation, and changed the code to round to the nearest microsecond instead. Thanks for the report! Fixed in Modules/datetimemodule.c; new revision: 1.25