Functions to truncate time types

[theophile-scrive]

During tests I insert an entity in PostgreSQL (which has microsecond resolution) with a timestamp created with time (which has picosecond resolution). When I fetch the entity from the database, the timestamps obviously differ, because PostgreSQL has lower precision:

- expiration = Just 2025-08-22 11:59:23.370617513 UTC}]
+ expiration = Just 2025-08-22 11:59:23.370617 UTC}]

I have the option to tweak the record before handing it to the assertEqual funtion in my test suite. I was wondering if I could remove some precision from UTCTime in a principled way. Is that advised?

(btw, thank you @AshleyYakeley for your work on this library!)

[AshleyYakeley]

There really isn't an easy way other than something like \t -> t {utctDayTime = (fromInteger $ div' (utctDayTime t) 1E-6) * 1E-6}

[AshleyYakeley]

Could create a class for this though:

class TruncateTime a where
    truncateTime :: DiffTime -> a -> a

with instances for DiffTime, NominalDiffTime, UTCTime, LocalTime, ZonedTime, CalendarDiffTime.

[theophile-scrive]

@AshleyYakeley Thanks. It appears my team already had an internal helper type (aptly named UTCTimePG) that performs the conversion where the workhorse is:

    truncateDiffTime :: DiffTime -> DiffTime
    truncateDiffTime =
      -- [4] get back DiffTime from picoseconds
      picosecondsToDiffTime
        -- [3] convert back to picoseconds
        . (* factor)j
        -- [2] convert to microseconds (quot truncates the decimal part)
        . (`quot` factor)
        -- [1] get picoseconds from DiffTime
        . diffTimeToPicoseconds
where
    -- conversion factor: 1 microsecond = 1_000_000 picoseconds
    factor :: Integer
    factor = 1_000_000

[AshleyYakeley]

OK, not going to add TruncateTime for 1.15, as it's potentially impacted by the design of #280.

[AshleyYakeley]

Following #280, probably look like this:

class Torsor a => TruncateTorsor a where
    truncateT :: TorsorGroup a -> a -> a

[theophile-scrive]

@AshleyYakeley yeah nothing to add in the code, although maybe a guide for this situation would be helpful for others. :)

[AshleyYakeley]

@theophile-scrive I'm going to hold this issue open because I might want to do something in this space.

[theophile-scrive]

@AshleyYakeley sure! :)