Skip to content

scenes.md

Latest commit

 

History

History
163 lines (134 loc) · 10.1 KB

scenes.md

File metadata and controls

163 lines (134 loc) · 10.1 KB
Raw

Scene engine

Source: src/CTLD_sceneManager.lua, src/scenes/*.lua Classes: CtldScene (one running instance), CTLDSceneManager (singleton registry + engine)

CTLDSceneManager executes time-sequenced deployments of DCS statics and ground groups from a declarative model. It is the backend for every FARP, FOB and minefield deployment: a scene model lists ordered steps, and the engine spawns each step's objects relative to a position snapshot, waiting a configurable delay between them.

Scene models live one per file under src/scenes/ and self-register at load time via CTLDSceneManager.getInstance():registerSceneModel(model). CTLDSceneManager:_registerBuiltins() is intentionally empty — there is no hard-coded model list.

Internal data model

CtldScene (one instance per active deployment)
  ├── _name        : "<modelName>#<counter>"  (unique per deployment)
  ├── _modelName   : original model name (used for registry / pack lookups)
  ├── _unit        : trigger DCS Unit (or a mock unit for playSceneAtPos)
  ├── _steps       : model.steps (shared reference)
  ├── _stepIndex   : current step pointer (0 before the first step)
  ├── _timeMarker  : absolute timer.getTime() target for the next step
  ├── _spawnedObjs : { DCSObject, … }  (every object spawned so far)
  ├── _params      : runtime bag forwarded to step funcs (e.g. farpName, repackData)
  ├── _onComplete  : callback fired after the last step (or model.onComplete)
  ├── _aborted     : true once abort() is called
  ├── _coalitionId / _countryId       : snapshot at init
  └── _refX / _refZ / _refAlt / _refHdgRad / _magDecDeg : position + heading snapshot

CTLDSceneManager (singleton, via getInstance())
  ├── _models[modelName] : registered model tables
  └── _active[sceneName] : CtldScene instances currently deployed

_models is keyed by model name; _active is keyed by the per-deployment _name ("<modelName>#<counter>"), so several instances of the same model can be active at once. Both tables are in-memory only — _active does not survive a mission restart or a full CTLD re-injection during development.

The reference position and heading are captured once in CtldScene:init() from the trigger unit (unit:getPoint() and ctld.utils.getHeadingInRadians). Every subsequent step is positioned relative to that snapshot, so the scene deploys coherently even if the unit moves or leaves. A step's func may overwrite _refX / _refZ / _refAlt on ctx.scene before later spawn steps run.

Step execution

The step machine is driven by timer.scheduleFunction. CtldScene:_execute() schedules the first step after steps[1].delayAfterPreviousStep seconds; _runNextStep() then executes the current step and schedules the next. Each step is one of three shapes:

Step type Key fields Behaviour
polar polar = { distance, angle }, relativeHeadingInDegrees, relativeAltitudeInMeters, registryKey Deterministic world position derived from the snapshot via ctld.utils.getRelativeCoords.
axis axis = { count, safeDistance, spacing }, registryKey A random axis around the unit; count objects distributed along it via ctld.utils.getSpawnObjectPositions.
func func = function(ctx) … end No spawn — runs the callback only (post-spawn hook / custom placement).

Every step also carries delayAfterPreviousStep (seconds): after step N runs, the engine waits that many seconds before step N+1. The same field on step 1 is the initial delay from the trigger.

For a spawning step (registryKey present, spawn not skipped), _runNextStep():

  1. Runs the optional preFunc(ctx) with ctx = { unit, step, scene }. Returning false skips this step's spawn (the scene continues); calling ctx.scene:abort(reason) stops the scene entirely.
  2. Looks up the descriptor with CTLDObjectRegistry.get(step.registryKey) and auto-injects circleRadius when the descriptor uses a circle formation.
  3. Spawns via CTLDObjectRegistry.spawnObject(registryKey, coalitionId, countryId, x, z, hdg, overrides) — the engine never calls coalition.addStaticObject / coalition.addGroup directly (a func-only step may, as the countryside FARP does for its boundary tyres).
  4. Appends every spawned object to _spawnedObjs.
  5. If step.critical is set and nothing spawned, calls abort() rather than continuing with a broken partial scene.
  6. Runs the optional func(ctx) with ctx = { unit, spawnedObj, step, scene }, where spawnedObj is the last object spawned this step (nil for a skipped or func-only step).

Both hooks are wrapped in pcall; an error is logged (ERROR) and the scene proceeds. When the last step finishes, _onComplete(scene) fires (also under pcall). A model may set model.onComplete; a caller-supplied callback to playScene overrides it.

Entry points

Method Use
playScene(unit, modelName, params, onComplete) Start a scene from a live trigger unit. Rejects a nil/dead unit, an unknown model, or a _disabled model.
playSceneAtPos(modelName, pos, coalitionId, countryId, params) Start a scene with no live unit (e.g. parachute auto-unpack). Builds a minimal mock unit at pos facing north and delegates to playScene.

Both register the new CtldScene in _active before executing.

FARP pack flow

Packing tears a deployed FARP scene back down into crates while preserving its warehouse state. The flow is split across the scene manager and CTLDCrateManager:

Player selects "Pack Equipt → Pack <scene>"
  └── CTLDCrateManager:refreshPackEquiptSection(playerObj)
        └── CTLDSceneManager:findNearbyRepackableScenes(transport:getPoint(), 300)
              └── returns _active scenes within radius whose model defines onRepack
        └── on click, per scene:
              1. sc = CTLDSceneManager._active[sceneName]
              2. repackData = CTLDSceneManager:packScene(sc)
                    ├── model.onRepack(sc, repackData)   ← snapshot warehouse into repackData
                    ├── destroy every obj in sc._spawnedObjs
                    └── _active[sc._name] = nil
              3. spawn cratesRequired crates via CTLDCrateManager:spawnCrate(...)
                    └── crate.metadata.warehouseSnapshot = repackData.warehouseSnapshot

On crate unpack at a new site (auto-unpack path in CTLDCrateManager):
  └── sm:getModel(desc.unit) resolves the scene model
        └── CTLDSceneManager:playSceneAtPos(desc.unit, centroid, coa, cId, { repackData = … })
              └── warehouse step reads ctx.scene._params.repackData.warehouseSnapshot to restore fuel/inventory

packScene(scene) takes only the scene instance; onRepack(scene, repackData) is where a model writes its repackData.warehouseSnapshot. The snapshot travels on the crate through crate.metadata.warehouseSnapshot and is threaded back into the new scene's _params.repackData so the warehouse-stocking step can restore liquids and inventory instead of zeroing them.

The relevant identifiers (onRepack, findNearbyRepackableScenes, packScene, repackData, warehouseSnapshot) keep their spelling in code; the F10 label is ctld.tr("Pack %1", …).

Adding a new scene (dev checklist)

  1. Create src/scenes/CTLD_myScene.lua: a local model table with name and steps, ending with CTLDSceneManager.getInstance():registerSceneModel(myScene).
  2. Declare the crate on the model itself — myScene.crate = { weight, i18nKey, deployKey, cratesRequired, side, … }. It is auto-injected into the Request Equipment menu by CTLDCrateManager:_processSpawnableCrates / _injectSceneCrate; there is no unit = "…" entry to add in CTLD_userConfig.lua.
  3. Register any DCS objects the steps reference with CTLDObjectRegistry.registerIfAbsent(key, descriptor) and point each step's registryKey at them.
  4. Add the file to tools/build/listToMerge.txt (scenes are listed after all managers so the crate auto-injection resolves) and add the matching dofile line to tests/helpers/loader.lua.
  5. If the scene deploys a mod-based helipad FARP with a warehouse, add a func-only final step that stocks it via w:setLiquidAmount(fuelType, qty) (fuel types 0–3: jet fuel, aviation gasoline, MW50, diesel). Read levels with w:getLiquidAmount(fuelType), never getLiquid. Invisible FARP airbases return nil from getWarehouse(), so guard for it.
  6. For pack support, implement myScene.onRepack(scene, repackData) that reads w:getLiquidAmount(...) and w:getInventory() into repackData.warehouseSnapshot.
  7. To make the scene deployable as a FOB, set fobCompatible = true inside the crate table (myScene.crate.fobCompatible). A FOB scene typically also supplies a custom crate.unpack = function(unit, unitName, sceneName) … end delegating to CTLDFOBManager.
  8. If the scene needs a mod that cannot be probed (heliport types with probeSkip = true), set myScene.requiresMod = "<registryKey>" so _auditAfterModValidator emits a startup WARN.

src/scenes/CTLD_countrysideFarpScene.lua is the reference implementation (Invisible FARP + warehouse stocking + onRepack); src/scenes/CTLD_fobScene.lua shows the FOB variant.

Mod validation

After CTLDModValidator runs, CTLDCoreManager:init() calls CTLDSceneManager:_auditAfterModValidator(). It walks every registered model's steps, resolves each registryKey in CTLDObjectRegistry, and — skipping entries flagged probeSkip — checks the DCS type against the validator (isStaticInvalid / isGroundInvalid). A model with any missing type is marked _disabled, reported via trigger.action.outText, and purged from the Request Equipment menu (CTLDCrateManager:_purgeDisabledScenes). Models declaring requiresMod (which cannot be auto-validated) instead emit a WARN reminding mission makers that all clients need the mod. isSceneEnabled(name) reflects the result. See Architecture for the probeSkip rationale and the object registry.