Support pickle dump for DateTime (required by PyQt signals)

[touilleMan]

[touilleMan]

We said we use u64/i64

Yup, we should use u64/i64 when defining our structure... unless we have a good reason not to ;-)

Here we wrap chrono::Datetime that uses i32/u32, hence forcing to use u64/i64 here is useless and cumbersome (typically adding those conversion lead to adding a bug in the year type that was supposed to be signed)

[TimeEngineer]

We said we use u64/i64

Yup, we should use u64/i64 when defining our structure... unless we have a good reason not to ;-)

Here we wrap chrono::Datetime that uses i32/u32, hence forcing to use u64/i64 here is useless and cumbersome (typically adding those conversion lead to adding a bug in the year type that was supposed to be signed)

Same shit can happen for usize/isize, but we prefered to use u64/i64 in Vincent's PR :thinking:

[touilleMan]

Same shit can happen for usize/isize, but we prefered to use u64/i64 in Vincent's PR thinking

Just follow KISS then ;-)

This u64/i64 thing is really important for serialized structures (given otherwise we will fail to deserialize a msgpack document containing a big number). Otherwise it is useful as a rule of thumb for not having to wonder between u32/u64/usize everytime we need a number type. But when the number type has already been decided for us (like what is done with chrono::DateTime) or if there is a good enough reason (if we where to implement DateTime from scratch I guess we could argue month/day/hour/minute/second cannot be > 60 so we may want to use u8 here... who knows ^^)

[vxgmichel]

(required by PyQt signals)

I'm not sure pickle support is required for PyQt signals. I ran a quick test on master and it seems there's not problem with providing a parsec._parsec.Datetime to a pyqtSignal(object). Maybe I'm missing something?

[FirelightFlagboy]

Can't reproduce