Another step in reducing the rigidity of FlightPlan and making it
testable.
There is one intentional behavior change here: escort flights no longer
request escorts. That actually has a very minimal effect because these
properties are only used for two things: determining if a package needs
escorts or not, and determining when the TARCAP should show up and
leave. Since escorts won't have been in the package when the first part
happens anyway, that has no effect. The only change is that TARCAP won't
show up earlier or stay later just because of a TOT offset for an escort
flight.
This doesn't need to be a part of FlightPlan, and it's easier to test if
it isn't. Move it out and add the tests.
It's pretty misleading to allow this in the core of the flight plan code
anything. This is an extremely unreliable estimate for most aircraft so
it should be more clearly just for briefing purposes.
Create a WaypointAction class that defines the actions taken at a
waypoint. These will often map one-to-one with DCS waypoint actions but
can also be higher level and generate multiple actions. Once everything
has migrated all waypoint-type-specific behaviors of
PydcsWaypointBuilder will be gone, and it'll be easier to keep the sim
behaviors in sync with the mission generator behaviors.
For now only hold has been migrated. This is actually probably the most
complicated action we have (starting with this may have been a mistake,
but it did find all the rough edges quickly) since it affects waypoint
timings and flight position during simulation. That part isn't handled
as neatly as I'd like because the FlightState still has to special case
LOITER points to avoid simulating the wrong waypoint position. At some
point we should probably start tracking real positions in FlightState,
and when we do that will be solved.
There's an ugly special case in flight simulation to handle hold points
because we don't differentiate between the total time between two
waypoints (which can include delays from actions like holding) and
travel time. Split those up and remove the special case.
Putting the ingress point directly on one end of the FLOT means that AI
flights won't start searching and engaging targets until they reach that
point. If the front line has advanced toward the flight's departure
airfield, it might overfly targets on its way to the IP.
Instead, place an IP for CAS the same way we place any other IP. The AI
will fly to that and start searching from there.
This also:
* Removes the midpoint waypoint, since it didn't serve any real purpose
* Names the FLOT boundary waypoints for what they actually are
Fixes https://github.com/dcs-liberation/dcs_liberation/issues/2231.
mypy update is needed for typing.Self support. It caught an existing bug
(missing @property on override), and fixed a bug so we can drop an
ignore.
Upgrading mypy requires upgrading pydantic to get the newest pydantic
mypy plugin, and since that's what's driving fastapi it's probably smart
to upgrade those together.
mypy struggles to prove this cast correct when there are two or'd
isinstance checks where both types coincidentally have properties of the
same name (but no defined protocol making that explicit). I'm not really
sure why mypy is happy with this in its current state, but it isn't
after a change I'm making.
All our isinstance use is a bit of an anti-pattern anyway, so extract a
method that exposes the data we care about.
The start/end times for tankers aren't actually used, so this could be
simplified even more, but that data _should_ be used.
Troops must be dropped inside this zone or they won't attack the target.
The zone needs to be drawn in the map so players don't break the flight
plan by accidentally moving the drop waypoint outside the DZ.
I've move the API for doing this out of `PatrollingFlightPlan` in favor
of a mixin so this is no longer presented as `engagement_distance` by
the flight plan. I don't love that it's still the `commit-boundary`
endpoint, but it's fine for now.
I don't know why mypy wasn't able to catch this. pycharm is also
struggling to understand this class.
The timing for these doesn't work. Sweep RTBs at the same time the
package reaches its TOT. The tanker won't be on station until 1m30s
before the package reaches the refueling point.
This isn't actually the data that callers usually want. Most of the
callers just want the bounds. The heading and length are trivially
computed from that. Add a class to contain the result so it's easier to
refactor.
The next step in splitting up the layout and scheduling phases. This
facilitates splitting flights into two classes where one has a full
flight plan, but one used in the earlier phases of planning has only a
layout. Layout-only flights won't need TOTs, which will make them much
easier to work with once we've migrated TOTs from timedeltas to
datetimes.
Layout-only flights of course aren't actually usable, but it lets us
avoid dealing with the current sim time until we're certain the Flight
will even survive planning.
I'm not actually sure if we'll be able to split the two phases any more,
but this ends up being a nice cleanup anyway.
We can always estimate a startup time now. Remove the nullability from
the result, cleanup the callsites, and eliminate
TotEstimator.mission_start_time since it no longer does anything useful.
Flights without a meaningful TOT make the code around startup time (and
other scheduling behaviors) unnecessarily complicated because they have
to handle unpredictable flight plans. We can simplify this by requiring
that all flight plans have a waypoint associated with their TOT. For
custom flight plans, we can just fall back to the takeoff waypoint. For
RTB flight plans (which are only synthetic flight plans injected for
aborted flights), we can use the abort point.
This also means that all flight plans now have, at the very least, a
departure waypoint. Deleting this waypoint is invalid even for custom
flights, so that's no a problem.
