Safe Haskell  None 

Language  Haskell2010 
Synopsis
 data Forecast a = Forecast {}
 data OutsideForecastRange = OutsideForecastRange {}
 constantForecastOf ∷ Ticked a → WithOrigin SlotNo → Forecast a
 mapForecast ∷ (Ticked a → Ticked b) → Forecast a → Forecast b
 trivialForecast ∷ GetTip b ⇒ b → Forecast ()
 crossEraForecastBound ∷ WithOrigin SlotNo → SlotNo → Word64 → Word64 → SlotNo
Documentation
Forecast the effect of time ticking
data OutsideForecastRange Source #
OutsideForecastRange  

Instances
constantForecastOf ∷ Ticked a → WithOrigin SlotNo → Forecast a Source #
Forecast where the values are never changing
This is primarily useful for tests; the forecast range is infinite, but we do still check the precondition, to catch any bugs.
trivialForecast ∷ GetTip b ⇒ b → Forecast () Source #
Trivial forecast of values of type ()
performed by an instance of
GetTip
.
Specialization of constantForecast
.
Utilities for constructing forecasts
crossEraForecastBound Source #
∷ WithOrigin SlotNo  Current tip (the slot the forecast is at) 
→ SlotNo  Slot at which the transition to the next era happens 
→ Word64  Max lookeahead in the current era 
→ Word64  Max lookeahead in the next era 
→ SlotNo 
Compute the upper bound for a range for a forecast across eras.
We have to be very careful here in how we compute the maximum lookahead. As long as we are in a single era, things look like this:
/\   chain ...  block  block  block [block]   v ledger TIP VIEW
where TIP
is the current ledger tip and VIEW
is the last ledger view we
can forecast, because the next block [block]
to arrive will take effect in
the next leger state after VIEW
. Note that if the maximum lookahead is
zero, this looks like
chain ...  block  block  block [block]   ledger TIP
where [block]
can have immediate changes on the ledger, and so we can't
look ahead at all (of course, we always know the current TIP
).
Note that blocks arriving after [block]
can only take effect later than
[block]
, and so they are not relevant for computing the maximum slot number
we can compute a ledger view for.
Now, if we are near an era transition, this picture gets a bit more complicated. If the next block is still in this era (that is, unless we are right at the edge), then that imposes one constraint, as before. However, the first block in the next era imposes an additional constraint:
~ ~ /\ ~   / ~ \   ~    block [block] ~ [block']    ~ v v TIP ~ VIEW ~
There are no restrictions on the relative values of these two maximum lookahead values. This means that it's quite possible for the next era to have a smaller lookahead (to reiterate, since that era has not yet begun, the first block in that era is at the transition, and so the maximum lookahead applies from the transition point):
~ ~ /\ ~   / ~ \  ~    block [block] ~ [block']    ~ v v TIP ~ VIEW ~
Indeed, if the next era has zero lookahead, when the first block of the next era comes it, it can make changes immediately, and so we can't even know what the view at the transition point is.
Note that if there can be no more blocks in this era, the maximum lookahead of the current era is irrelevant:
~ ~ /\ ~   ~   ~   block ~ [block']   ~ v TIP ~ VIEW ~
We can therefore compute the earliest SlotNo
the next block in this era
(if any) can make changes to the ledger state, as well as the earliest
SlotNo
the first block in the next era can; their minimum
will serve as
an exclusive upper bound for the forecast range.