You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
{{ message }}
Repository navigation
Pandas 1.5 breaks trip purpose for some models #608
Following up on this thread. It seems the develop branch and the parking location choice PR (#620) are failing the tests due to this pandas FutureWarning.
My thoughts on options of solution:
Option 1: Change strict to False in test settings to not error out on warnings. This would be a quick work-around, but I can see reasons for keeping strict to True. In addition, in prototype_arc and prototype_semcog strict is set to False, but in prototype_mtc and others strict is set to True, at least we should make them consistent.
Option 2: Update the triggering line to conform with future pandas. To me, this would be the preferred solution. I'm not sure what's the protocol on keeping ActivitySim codes compatible with newer versions of dependencies, I guess people just update codes when FutureWarnings come up?
Option 3: Constrain pandas version. To me, this should be the least preferred option.
@jpn-- How is the main branch able to pass the tests with strict = True?
The main branch passed its tests because our last release of ActivitySim was 1.1.3 on Sept 13th. Pandas 1.5 was released a week later on Sept 19th. If the tests were re-run today on the main branch, they would fail.
Your option 2 is the correct answer for the long run; we want to stay compatible with the latest Pandas in general. As this is just a warning and not a "real" error we have not scrambled to fix this urgently. Option 3 is easy and has been used as a stopgap in some PRs until we get option 2 done.
At some point in the future the warning will become a bonafide error, and having the strict tests makes sure we know about these changes with plenty of time to repair before they get serious.
Activity
i-am-sijia commentedon Nov 2, 2022
Following up on this thread. It seems the develop branch and the parking location choice PR (#620) are failing the tests due to this pandas FutureWarning.
My thoughts on options of solution:
strictto False in test settings to not error out on warnings. This would be a quick work-around, but I can see reasons for keepingstrictto True. In addition, in prototype_arc and prototype_semcogstrictis set to False, but in prototype_mtc and othersstrictis set to True, at least we should make them consistent.@jpn-- How is the main branch able to pass the tests with
strict = True?jpn-- commentedon Nov 2, 2022
The main branch passed its tests because our last release of ActivitySim was 1.1.3 on Sept 13th. Pandas 1.5 was released a week later on Sept 19th. If the tests were re-run today on the main branch, they would fail.
Your option 2 is the correct answer for the long run; we want to stay compatible with the latest Pandas in general. As this is just a warning and not a "real" error we have not scrambled to fix this urgently. Option 3 is easy and has been used as a stopgap in some PRs until we get option 2 done.
At some point in the future the warning will become a bonafide error, and having the strict tests makes sure we know about these changes with plenty of time to repair before they get serious.
temporarily constrain pandas ActivitySim#608
jpn-- commentedon Dec 14, 2022
Closed by #629
BayDAG merge with ActivitySim v1.2 (#56)
BayDAG Contributions (#657)
maintenance release (#666)
Getting into sync with ActivitySim/develop (#8)
Input Checker (#753)