Support for negative timedeltas #23

Activity

1. SGmuwa commented on Apr 18, 2021

   

   SGmuwa

   on Apr 18, 2021

   More actions

   from pytimeparse.timeparse import timeparse
   from datetime import timedelta
   
   t = timedelta(days= -1, seconds=5)
   print(t) # -1 day, 0:00:05
   t = timedelta(seconds=timeparse(str(t)))
   print(t) # -2 days, 23:59:55
   t = timedelta(seconds=timeparse(str(t)))
   print(t) # -3 days, 0:00:05
   
   t = timedelta(days= -1, seconds=-5)
   print(t) # -2 days, 23:59:55
   t = timedelta(seconds=timeparse(str(t)))
   print(t) # -3 days, 0:00:05
   t = timedelta(seconds=timeparse(str(t)))
   print(t) # -4 days, 23:59:55
2. Yuren0312 commented on Oct 18, 2026

   

   Yuren0312

   on Oct 18, 2026

   More actions

   Thanks for the report! I can confirm this on current master: timeparse extracts the leading - and negates the whole field sum, but timedelta.__str__ negates only the day count while the clock stays non-negative. So str(timedelta(days=-1, seconds=5)) = -1 day, 0:00:05 parses as -86405 instead of -86395, and every round-trip drifts further (the repro in the thread: -86405 -> -259195 -> -259205 -> ...).

   I've opened #41 with a minimal fix: the second-count computation was factored into a _sum_fields(mdict, sign) helper, and for the day+clock form a leading minus now negates only the week/day portion while the clock stays positive, matching timedelta semantics. The DAYCLOCK (DD-HH:MM:SS) and unit forms keep their existing whole-expression sign behaviour.

   Now -1 day, 0:00:05 -> -86395 and the round-trip is stable. Two regression tests were added and the two old expectations that encoded the previous (incorrect) whole-expression negation were updated. Verification: python -m unittest discover -s pytimeparse/tests -> 62 tests, OK.

   Happy to adjust if you'd prefer a different approach.