pokemon/gen1recomp
1
0
Fork 0
mirror of https://github.com/bryanthaboi/gen1recomp.git synced 2026-08-16 08:11:35 +02:00
Code Issues Packages Projects Releases Wiki Activity

G2 support

Browse Source
This commit is contained in:
bryanthaboi
2026-08-11 11:52:56 -04:00
parent 79ed37699e
commit ae6cac89e1
489 changed files with 226677 additions and 1798 deletions
Download Patch File Download Diff File Expand all files Collapse all files
src/mods/Schemas.lua
+864 -8
View File
@@ -214,7 +214,15 @@ end
-- records are a feature -- but a patch key that is only a case/underscore
-- variant of a schema field is the classic typo and gets rejected with a
-- suggestion.
function Schemas.check(spec, registryName, id, value, mode)
--
-- `generation` is optional and only ever narrows: passing it resolves the
-- per-generation shape first (Schemas.shapeFor), and omitting it validates
-- against the Gen 1 shape, which is what every Gen 1 call site wants and what
-- a caller already holding a derived spec has anyway.
function Schemas.check(spec, registryName, id, value, mode, generation)
if generation ~= nil then
spec = Schemas.shapeFor(registryName, spec, generation)
end
if mode == "remove" or spec == nil then return true end
-- register and patch are synonyms on a deep registry, so a partial
-- payload is the normal case there and only override is a full value
@@ -323,10 +331,10 @@ local function refsFor(spec, name, id, value)
return refs
end
-- a structured target (battle_anims' per-kind subtables) hides its ids one
-- level down, so the pristine scan asks the spec instead of the raw keys
local function baseEntries(registry, base)
local spec = registry.spec
-- a structured target (battle_anims' per-kind subtables, Gold's trainer
-- classes) hides its ids one level down, so the pristine scan asks the spec
-- instead of the raw keys
local function baseEntries(spec, base)
if not spec.baseIds then return pairs(base) end
local ids = spec.baseIds(base)
local i = 0
@@ -357,13 +365,26 @@ function Schemas.crossValidate(loader, data)
end
end
for name, registry in pairs(loader.content) do
local spec = registry.spec
-- the shape this boot's generation validates by, so a Gen 2 record's
-- refs are read out of the Gen 2 fields (a species' `into`, not
-- `species`) instead of being missed entirely
local spec = Schemas.shapeFor(name, registry.spec, loader.generation)
for id in pairs(registry.ops) do
local value = registry:get(id)
if value ~= nil and registry.owners[id] ~= Schemas.ENGINE then
for _, ref in ipairs(refsFor(spec, name, id, value)) do
local refRegistry = Schemas.REGISTRIES[ref.registry]
and loader.content[ref.registry]
-- A registry with no home in this generation has no id space to
-- check against: its base view resolves to nothing, so EVERY
-- reference into it would read as dangling. Gold's species carry a
-- growthRate and an evolution method like Red's do; the ids are
-- fine, it is the Gen 1 `growth_rates` / `evolution_methods`
-- namespaces that are not there to confirm them. Skipped for the
-- same reason an undeclared registry is: unknown, not wrong.
if refRegistry and Schemas.gatedFor(ref.registry, loader.generation) then
refRegistry = nil
end
if refRegistry and refRegistry:get(ref.ref) == nil then
problems[#problems + 1] = {
owner = registry.owners[id],
@@ -377,7 +398,7 @@ function Schemas.crossValidate(loader, data)
if removed then
local base = registry.base and registry.base()
if base then
for id, value in baseEntries(registry, base) do
for id, value in baseEntries(spec, base) do
if registry.ops[id] == nil then
for _, ref in ipairs(refsFor(spec, name, id, value)) do
local set = tombstoned[ref.registry]
@@ -408,9 +429,307 @@ Schemas.ALIASES = { scripts = "map_scripts", ui = "screens" }
-- pass skips it and stays zero-work on a mod-free boot
Schemas.ENGINE = "engine"
-- ------- generation routing
--
-- Registry NAMES are shared across generations on purpose: a mod writes
-- mod.content.pokemon whichever game is running, and mod.content.encounters
-- means "wild encounters" in both. What can differ is the Data path the
-- merge lands on, because Gold namespaces the tables whose Gen 1 counterpart
-- means something else (data.gen2Palettes beside data.palettes).
--
-- One routing table per generation, read through Schemas.routing, and both
-- are read the same way:
--
-- absent -> keeps spec.target in that generation
-- mapped to a path -> merges there instead
-- mapped to false -> no home in that generation; the write is taken,
-- dropped and reported
--
-- That last case is the whole point of the manifest's gen2compat opt-in: a
-- mod that claims Gen 2 gets told which registry has no home there instead of
-- merging into a table nothing reads and appearing to work. It runs in both
-- directions, because the catalog now holds content BOTH ways round: Gold has
-- systems Red never had (the phone book, the decorations, the radio dial), and
-- those registries are the mirror image of the rows below -- declared once,
-- gated under GEN ONE, reported to a Red mod in the same sentence a Gold mod
-- gets about `tokens`. Schemas.GEN1 below Schemas.GEN2 carries them.
--
-- The `false` rows used to have three causes and now have one. The first is
-- gone: Gold's overworld tables no longer load off disk into World fields --
-- src/core/Game2.lua:load reads every one of them into game.data BEFORE it
-- calls mods:load(self.data), and src/world/gen2/World.lua:dataTable takes
-- them by reference and never copies, so a routed row merges into the very
-- table the world walks. The second is gone too: a registry whose Gen 2
-- records are shaped differently now says so in its own spec (the gen2Fields /
-- gen2Keys layer below, resolved by Schemas.shapeFor), so routing it validates
-- a mod's record against the GEN 2 shape rather than against Red's. What is
-- left is the systems Gold has not reimplemented through a registry at all.
Schemas.GEN2 = {
-- Namespaced on Gold and merged there. The registry NAME stays shared --
-- mod.content.maps means "maps" in both games -- and only the Data path
-- underneath it differs, which is the whole reason this table maps to paths
-- rather than renaming anything. Two id-space notes an author needs, and
-- docs/mod-api-gen2-compat.md spells out: Gold's `text` ids are ROM pointer
-- strings ("55:4067") rather than TEXT_* names, and a Gen 2 tileset carries
-- its walkability as `collision` where Gen 1 says `walkable`.
maps = "gen2Maps", tilesets = "gen2Tilesets", sprites = "gen2Sprites",
text = "gen2Text",
-- Namespaced AND differently shaped, and the shape is what these waited on.
-- Each carries a Gen 2 record schema in its catalog entry now, so the id
-- space is the one Gold actually keys by: the encounter KIND (.grass), the
-- trainer CLASS (one level into .classes, through gen2Write), a species id
-- or an ICON_ sheet name, and for palettes / battle_anims / constants the
-- target's own subtable names.
encounters = "gen2Encounters", trainers = "gen2Trainers",
palettes = "gen2Palettes", icons = "gen2Icons",
battle_anims = "gen2BattleAnims", constants = "gen2Constants",
-- Gold reimplements the system, and reads its rules back through the same
-- registry: src/battle/gen2/Battle.lua:statusRecordFor / moveEffectRecordFor,
-- Catching.recordFor, Ai.layersFor, Evolution.methodFor and
-- src/core/gen2/ItemEffects.lua:recordFor each read the merged table here
-- and fall back to their own module records when no loader ran. The vanilla
-- records at these paths are GOLD's, not Red's -- src/mods/Builtins.lua
-- swaps the registrant per generation, which it has to: both games call it
-- GREAT_BALL.
statuses = "gen2Statuses", move_effects = "gen2MoveEffects",
item_effects = "gen2ItemEffects", balls = "gen2Balls",
ai_classes = "gen2AiClasses", evolution_methods = "gen2EvolutionMethods",
-- The Gen 2-only six. They have no Gen 1 target to keep (their specs carry
-- none), so the routed path IS the only path they ever have, and the
-- Schemas.GEN1 rows below are what makes writing to one on Red a reported
-- drop rather than a merge into a table Red has never heard of. Two of them
-- merge onto a table that already exists when mods:load runs -- landmarks
-- onto the cache's own gen2Landmarks.landmarks, held_items onto the view
-- src/core/Game2.lua builds from data.items -- and the other four come
-- into existence AS the merge, seeded from their module's literals by
-- src/mods/Builtins.lua the way the battle-rule six are.
held_items = "gen2HeldItems", phone_contacts = "gen2PhoneContacts",
decorations = "gen2Decorations", apricorns = "gen2Apricorns",
landmarks = "gen2Landmarks.landmarks", radio_channels = "gen2RadioChannels",
-- Still no Gen 2 home. Every one of these is a system Gold reimplements
-- WITHOUT reading a registry: the Gen 1 target is still built and merged
-- into, but nothing in a Gold boot ever looks at it. Closing one is a
-- consumer change in the Gen 2 module first and a row here second, which is
-- exactly how battle_sprite_scales and render_pipelines came off this list
-- (see the note under it).
-- rulesets no Gen 2 ruleset dispatch exists
-- transitions Gold draws its own battle intro
-- (src/ui/gen2/BattleTransition.lua) and its STYLES table
-- is a boolean SET of the four cart wipes, not the
-- { frames, draw, sound, flash } record this registry
-- carries; there is no styleDef lookup for a mod id to
-- reach, so a registered style would fall back to vanilla
-- field the overworld grab bag; Gold's equivalents live in
-- gen2Maps and the VM's own tables
-- text_pointers Gen 1's TEXT_* indirection; Gold's text IS pointers
-- link_fields link play is Gen 1 only
rulesets = false, transitions = false,
field = false, text_pointers = false, link_fields = false,
-- battle_sprite_scales and render_pipelines are ABSENT from this table on
-- purpose: both keep the shared Gen 1 target because Gold reads the merged
-- table at that exact path.
-- battle_sprite_scales src/ui/gen2/BattleState.lua:imageScale reads
-- data.battle_sprite_scales with the same
-- image-then-species-then-default order as Gen 1's
-- BattleState.imageBattleScale / resolveBattleScale,
-- skipping the `_owners` bookkeeping row the same
-- way. Only the DEFAULT differs and neither side
-- reads it from here: Red's 32x32 back pics draw at
-- 2x, Gold's 48x48 ones fill their 6x6 box at 1x.
-- render_pipelines src/core/Game2.lua:load calls Pipelines.install
-- AFTER the merge, so data.render_pipelines is the
-- merged table, and Game2:draw composites `present`
-- through Pipelines.wantsPresent / Pipelines.present.
-- The `drawWorld` half is not composited yet (Gold's
-- overworld draws straight to the window rather than
-- into a canvas), and Game2 RETIRES a restored
-- drawWorld-only level rather than leaving it
-- switched on and rendering nothing; a mod that
-- registers only drawWorld is therefore inert on
-- Gold, which docs/mod-api-gen2-compat.md says in
-- those words. Gold also has no OPTION row for a
-- pipeline (Pipelines.rows is read only from
-- src/ui/OptionsMenu.lua), so a Gold player reaches
-- one by its hotkey.
-- src/script/gen2/Vm.lua is a bytecode VM over the cart's own opcodes, not
-- the Gen 1 row-list runner. `commands` IS routed (it is absent from this
-- table, so it keeps the shared data.commands target): the VM dispatches the
-- Opcodes.MOD_COMMAND row -- an op name with no cart byte behind it --
-- through that merged table, so mod.commands:register works on both games.
-- data.gen2Scripts is that bytecode pool keyed by ROM pointer, so
-- `map_scripts` has no home there: a Lua row list merged into it is not
-- something the VM can run. (`tokens` used to sit on this line and does
-- not belong there -- TextBox.new runs TextBox.substitute on EVERY box in
-- both generations and substitute reads game.data.tokens, so the shared
-- target was already live on Gold. It keeps that target, absent from this
-- table, and src/core/Game2.lua seeds data.tokens with a copy of
-- TextBox.TOKENS so the merge cannot mutate the module table.)
map_scripts = false,
-- Everything not listed keeps its Gen 1 target and works on Gold today:
-- pokemon, moves, items, type_chart, audio + music/sfx/cries/map_songs,
-- screens (the Gen2* ids in src/ui/Screens.lua), strings, font and commands.
}
-- The mirror of Schemas.GEN2: what a GEN 1 boot does with the registries that
-- only exist because Gold exists. Same three readings as the table above, and
-- only the third is used today -- there is no Gen 2-only registry with a
-- useful Red target to reroute to, because the systems themselves are absent
-- from Red rather than spelled differently there.
--
-- held_items Red's items carry no held attributes at all; the whole
-- hold/trigger machinery is Gen 2 (src/battle/gen2/Battle.lua
-- heldEffect)
-- phone_contacts no Pokegear, no phone
-- decorations no bedroom PC decoration menu
-- apricorns no Kurt, no apricorn balls
-- landmarks Red's town map is a Gen 1 town-map table, not the
-- LANDMARK_* index space the Pokegear and the #DEX AREA
-- page share
-- radio_channels no radio
Schemas.GEN1 = {
held_items = false, phone_contacts = false, decorations = false,
apricorns = false, landmarks = false, radio_channels = false,
}
-- The routing table for a generation: which one is consulted is the only
-- difference between the two directions. An unknown generation routes
-- nothing, so every registry keeps its catalog target.
local NO_ROUTING = {}
function Schemas.routing(generation)
if generation == 2 then return Schemas.GEN2 end
if generation == 1 then return Schemas.GEN1 end
return NO_ROUTING
end
-- The Data path `name` merges into for a generation, or nil when the registry
-- has no home there.
function Schemas.targetFor(name, spec, generation)
local routed = Schemas.routing(generation)[name]
if routed == nil then return spec.target end
return routed or nil
end
-- true when the registry exists but this generation has nowhere to put it,
-- which is a different diagnostic from a registry that has no target at all
function Schemas.gatedFor(name, generation)
return Schemas.routing(generation)[name] == false
end
-- ------- per-generation record shapes
--
-- Routing says WHERE a registration lands; this says what a record at that
-- path LOOKS like. The two are separate questions and only the second one is
-- gating the rest of the catalog: Gold's tables are the Gen 2 ROM's own
-- layout, so a species carries specialAttack/specialDefense where Red carries
-- one `special`, wild encounters key by encounter kind and time of day rather
-- than by map, and the palette table is GBC four-colour rows in a dozen named
-- subtables. Validating any of those against the Gen 1 schema judges a mod's
-- record against the wrong shape, which is worse than refusing the write.
--
-- So beside `value` / `fields` / `keys` / `keyValue` a spec may carry
-- `gen2Value` / `gen2Fields` / `gen2Keys` / `gen2KeyValue`, and beside
-- `semantics` / `extra` / `write` / `baseAt` / `baseIds` / `example` /
-- `notes` the matching `gen2*`. Absent means "the Gen 1 shape is right here
-- too", which is the common case and why most registries carry none of this.
-- The registry NAME, the verbs and (wherever the id space allows it) the ids
-- stay shared, exactly as the routing table keeps them shared.
--
-- `gen2X = false` CLEARS the Gen 1 slot rather than setting it, the same
-- reading `false` has in the routing table above: battle_anims' Gen 1 `write`
-- routes ids into per-kind subtables by prefix, and under Gen 2 the ids ARE
-- the subtables, so the right Gen 2 write is the default one. No slot here
-- ever carries a meaningful `false` (they are functions, strings and tables),
-- so the two readings cannot collide.
--
-- Schemas.shapeFor resolves it. It hands back the spec unchanged for Gen 1
-- and for any registry with no Gen 2 shape; otherwise a derived spec with the
-- gen2* keys folded onto the canonical names, memoized per spec so the
-- resolve is one table lookup after the first call. Everything downstream --
-- Schemas.check, Registry's fold and baseAt, the loader's merge and write --
-- then reads one spec and never learns about generations.
local GEN2_SHAPE = {
gen2Value = "value", gen2Fields = "fields", gen2Keys = "keys",
gen2KeyValue = "keyValue", gen2Extra = "extra",
gen2Semantics = "semantics", gen2Write = "write",
gen2BaseAt = "baseAt", gen2BaseIds = "baseIds",
gen2Example = "example", gen2Notes = "notes",
}
-- Schemas.check reads these four in a fixed order (keys/keyValue, then value,
-- then fields), so a Gen 2 shape that describes its records with `keys` must
-- clear the Gen 1 `value` rather than sit beside it: otherwise the first
-- branch that matches wins and the new schema is never consulted.
local VALUE_SLOTS = { value = true, fields = true, keys = true, keyValue = true }
-- Weak keys: a derived spec lives exactly as long as the catalog entry it
-- came from, which in a headless harness is per require rather than forever.
-- Keyed by spec alone, which is sound because the catalog gives every
-- registry its own table -- the two ALIASES resolve to the canonical name
-- before anything reaches here, and `target` is the only name-dependent
-- field a derived spec carries.
local derivedSpecs = setmetatable({}, { __mode = "k" })
-- does this registry describe its Gen 2 records differently at all?
function Schemas.hasGen2Shape(spec)
if type(spec) ~= "table" then return false end
for source in pairs(GEN2_SHAPE) do
if spec[source] ~= nil then return true end
end
return false
end
-- The spec to validate and merge `name` with under `generation`. Idempotent:
-- a derived spec carries no gen2* keys, so resolving one again returns it.
function Schemas.shapeFor(name, spec, generation)
if generation ~= 2 or not Schemas.hasGen2Shape(spec) then return spec end
local hit = derivedSpecs[spec]
if hit then return hit end
local out = {}
for key, value in pairs(spec) do out[key] = value end
local replacesValue = false
for source, slot in pairs(GEN2_SHAPE) do
if spec[source] ~= nil and VALUE_SLOTS[slot] then replacesValue = true end
end
if replacesValue then
for slot in pairs(VALUE_SLOTS) do out[slot] = nil end
end
for source, slot in pairs(GEN2_SHAPE) do
out[source] = nil
-- `or nil` is the clear: gen2Write = false leaves the slot empty
if spec[source] ~= nil then out[slot] = spec[source] or nil end
end
-- self-describing: a derived spec's `target` is the routed one, so a caller
-- holding it alone never reads the Gen 1 path by accident. targetFor stays
-- authoritative and stays idempotent over the result.
out.target = Schemas.targetFor(name, spec, generation)
derivedSpecs[spec] = out
return out
end
local R = {}
Schemas.REGISTRIES = R
-- ------- shared Gen 2 leaves
--
-- The ROM name spaces Gold's tables key by. They are enums rather than
-- f.str so a typo ("MORNING") fails at register time instead of writing a
-- subtable nothing ever reads; the ordered lists themselves ship as
-- data.gen2Constants (eggGroupOrder, trainerTypeOrder, ...).
-- wild encounters, overworld palettes and the roof pair all bucket by time of
-- day; DARK is the fourth palette bucket and never an encounter one, so the
-- encounter maps take the three-value list (constants/time_of_day.asm)
local gen2Tod = f.enum{ "MORN", "DAY", "NITE" }
local gen2PaletteTod = f.enum{ "MORN", "DAY", "NITE", "DARK" }
-- one GBC colour as the extractor writes it: a positional {r,g,b} triple
-- already expanded from 5-bit BGR to 0..255 (src/render/GbcPalette.lua)
local gen2Color = f.list(f.int(0, 255))
-- a palette row. The OBJ rows the sprite and mon pics use carry two colours
-- (the cart supplies white and black), the BG rows carry all four.
local gen2PaletteRow = f.list(gen2Color)
R.pokemon = {
semantics = "record", target = "pokemon",
fields = {
@@ -445,7 +764,61 @@ R.pokemon = {
battleScaleFront = f.opt(f.numRange(0.25, 4.0)),
battleScaleBack = f.opt(f.numRange(0.25, 4.0)),
},
-- Same registry, same target (data.pokemon), same species ids: only the
-- record differs, and it differs in four places, every one of them a real
-- Gen 2 change rather than an extractor spelling. Gen 2 splits `special`
-- into specialAttack/specialDefense (BaseData in pokegold's
-- data/pokemon/base_stats/), names the level-up table `levelMoves` and the
-- pic size `picSize`, has no separate level-1 move list (level 1 rows live
-- in levelMoves), and points an evolution at `into` rather than `species`.
-- Beside that it carries the breeding block Gen 1 has no analogue for
-- (eggGroups/eggMoves/eggSteps, genderRatio) and a held-item pair.
--
-- Without this, mod.content.pokemon:register is unusable for a Gold
-- species -- every record fails on the missing `special` -- while patch
-- happens to work, which is the worst of both.
gen2Fields = {
id = f.str, name = f.str, dex = f.int(1),
index = f.opt(f.int(0, 255)),
types = f.list(f.id("type_chart")),
baseStats = f.rec{ hp = f.int(1, 255), attack = f.int(1, 255),
defense = f.int(1, 255), speed = f.int(1, 255),
specialAttack = f.int(1, 255),
specialDefense = f.int(1, 255) },
catchRate = f.int(0, 255), baseExp = f.int(0, 255),
growthRate = f.id("growth_rates"), growthRateId = f.opt(f.int(0, 255)),
levelMoves = f.list(f.rec{ level = f.int(1), move = f.id("moves") }),
tmhm = f.opt(f.list(f.id("moves"))),
-- the raw TM/HM bitfield bytes, kept beside the resolved list so a
-- re-export round-trips; the engine reads `tmhm`
tmhmRaw = f.opt(f.list(f.int(0, 255))),
evolutions = f.list(f.rec{ method = f.id("evolution_methods"),
into = f.id("pokemon"),
level = f.opt(f.int(1)),
item = f.opt(f.id("items")),
-- EVOLVE_HAPPINESS' window and EVOLVE_STAT's
-- attack-vs-defence test
time = f.opt(f.enum{ "ANYTIME", "MORNDAY",
"NITE" }),
comparison = f.opt(f.enum{ "ATK_LT_DEF",
"ATK_GT_DEF",
"ATK_EQ_DEF" }) }),
-- breeding: two egg groups (the raw byte packs both nibbles), the egg
-- move list, and the cycle count src/core/gen2/Breeding.lua counts down
eggGroups = f.opt(f.list(f.str)), eggGroupsRaw = f.opt(f.int(0, 255)),
eggMoves = f.opt(f.list(f.id("moves"))), eggSteps = f.opt(f.int(0)),
genderRatio = f.opt(f.int(0, 255)),
-- the two wild held items, in the ROM's own order (rare then common)
items = f.opt(f.list(f.id("items"))),
spriteFront = f.path, spriteBack = f.path, picSize = f.int(1, 7),
source = f.opt(f.str),
cry = f.opt(f.id("cries")), trueColor = f.opt(f.bool),
battleScaleFront = f.opt(f.numRange(0.25, 4.0)),
battleScaleBack = f.opt(f.numRange(0.25, 4.0)),
},
example = 'mod.content.pokemon:patch("MEW", { baseStats = { attack = 120 } })',
gen2Example = 'mod.content.pokemon:patch("TOTODILE", '
.. '{ baseStats = { specialAttack = 80 } })',
}
R.moves = {
@@ -504,8 +877,16 @@ R.maps = {
-- carries no palettes at all, so an id reference would fail validation for
-- a perfectly good mod wherever there is no imported dataset.
palette = f.opt(f.str),
-- destGroup / destMapNum are the ROM map-group pair Gen 2 carries beside
-- the destination it actually warps through (World:resolveWarp reads
-- destMap and destWarp, the same two keys Gen 1 does). Optional and
-- additive rather than a second warp shape: `maps` routes to
-- data.gen2Maps under Gen 2, so the records a mod patches there are the
-- extractor's own, and a strict rec would reject every one of them.
warps = f.opt(f.list(f.rec{ x = f.int(0), y = f.int(0),
destMap = f.str, destWarp = f.int(0) })),
destMap = f.str, destWarp = f.int(0),
destGroup = f.opt(f.int(0)),
destMapNum = f.opt(f.int(0)) })),
objects = f.opt(f.list(f.any)),
signs = f.opt(f.list(f.any)),
connections = f.opt(f.map(f.enum{ "north", "south", "east", "west" }, f.any)),
@@ -545,6 +926,36 @@ R.tilesets = {
example = 'mod.content.tilesets:register("MY_TILES", { image = "...", blocks = { ... } })',
}
-- ------- Gen 2 wild encounters
--
-- One wild slot. A fishing slot's species may be the literal 0 the ROM uses
-- for "no fish here, roll the map's water table instead" (pokegold
-- data/wild/fish.asm), which is why species is a union rather than a bare id.
local gen2Slot = f.rec{ level = f.int(1), species = f.id("pokemon") }
-- the sentinel row carries level 0 as well as species 0, so both floors drop
local gen2FishSlot = f.rec{ chance = f.int(0, 255), level = f.int(0),
species = f.union{ f.id("pokemon"), f.int(0, 0) } }
-- Headbutt/Rock Smash slots. species is optional and the level floor is 0
-- because TreeMonSet_Rock has no `rare` half in the ROM (pokegold
-- data/wild/treemons.asm ends the table after the common rows), so the four
-- Rock Smash maps that point at it carry a rare table read out of whatever
-- follows: levels past 100 and rows with no species at all. Rejecting it
-- would mean the extractor's own table could never be re-registered.
local gen2TreeSlot = f.rec{ chance = f.int(0, 255), level = f.int(0),
species = f.opt(f.id("pokemon")) }
-- a grass row: one encounter rate and one seven-slot table PER time of day,
-- which is the whole reason this cannot share the Gen 1 shape
local gen2GrassRow = f.rec{
map = f.opt(f.str),
rates = f.map(gen2Tod, f.int(0, 255)),
slots = f.map(gen2Tod, f.list(gen2Slot)),
}
-- water has no time-of-day split: one rate, one three-slot table
local gen2WaterRow = f.rec{
map = f.opt(f.str), rate = f.int(0, 255), slots = f.list(gen2Slot),
}
R.encounters = {
semantics = "record", target = "encounters",
fields = {
@@ -556,7 +967,47 @@ R.encounters = {
slots = f.list(f.rec{ level = f.int(1),
species = f.id("pokemon") }) }),
},
-- Gold keys wild encounters by encounter KIND first and by map second
-- (data.gen2Encounters.grass.ROUTE_29), because the cart ships one table
-- per kind and a map appears in as many of them as it has water, swarms,
-- fishing spots and headbuttable trees. There is no per-map record to key
-- the registry by, so the id is the kind and `patch` folds per map instead
-- of replacing the kind's whole table.
--
-- Semantics stay "record" rather than becoming "deep" even though the id is
-- a namespace: a slot table is an ORDERED list whose position is the
-- encounter roll, and Merge.deepMerge appends lists under "deep" semantics,
-- so a mod rewriting a seven-slot table would get a fourteen-slot one.
gen2Keys = {
grass = f.map(f.str, gen2GrassRow),
-- the swarm variants shadow their base table while a swarm is running
swarmGrass = f.map(f.str, gen2GrassRow),
water = f.map(f.str, gen2WaterRow),
swarmWater = f.map(f.str, gen2WaterRow),
-- fishing: a map's rod points at a named group, and the group carries a
-- chance-ordered table per rod
fishGroups = f.map(f.str, f.rec{
id = f.opt(f.str), index = f.opt(f.int(0, 255)),
chance = f.int(0, 255),
old = f.list(gen2FishSlot), good = f.list(gen2FishSlot),
super = f.list(gen2FishSlot) }),
-- headbutt: map -> tree set id, and the set's common/rare tables. rocks
-- is the same indirection for Rock Smash.
trees = f.map(f.str, f.str),
rocks = f.map(f.str, f.str),
treeSets = f.map(f.str, f.rec{ common = f.list(gen2TreeSlot),
rare = f.list(gen2TreeSlot) }),
-- the Bug-Catching Contest pool (min/max level, not one level per slot)
bugContest = f.list(f.rec{ species = f.id("pokemon"),
min = f.int(1), max = f.int(1),
chance = f.int(0, 255) }),
-- where a roaming beast may walk next, keyed by the map it is on
roamMaps = f.list(f.rec{ map = f.str, to = f.list(f.str) }),
source = f.str, generation = f.int(1),
},
example = 'mod.content.encounters:patch("ROUTE_1", { grass = { rate = 30 } })',
gen2Example = 'mod.content.encounters:patch("grass", '
.. '{ ROUTE_29 = { rates = { NITE = 40 } } })',
}
R.trainers = {
@@ -582,7 +1033,64 @@ R.trainers = {
-- battles. The victory jingle stays kind-based.
battleTheme = f.opt(f.id("music")),
},
-- Gold hangs its rosters off data.gen2Trainers.classes, one record per
-- trainer CLASS carrying every named trainer of that class. The id space
-- is still the class id, so the registry keeps the Gen 1 call shape --
-- mod.content.trainers:patch("BEAUTY", { baseMoney = 99 }) -- and only the
-- one level of indirection to `.classes` is new. That is the same trick
-- battle_anims plays with its per-kind subtables, and it is why these three
-- callbacks exist rather than a `classes` key nobody would guess.
gen2BaseAt = function(base, id)
return base.classes and base.classes[id] or nil
end,
gen2BaseIds = function(base)
local ids = {}
for id in pairs(base.classes or {}) do ids[#ids + 1] = id end
return ids
end,
gen2Write = function(target, registry)
local classes = target.classes
if not classes then
classes = {}
target.classes = classes
end
local tombstones = {}
for id in pairs(registry.ops) do
local value = registry:get(id)
if value == nil then
tombstones[#tombstones + 1] = id
else
classes[id] = value
end
end
for _, id in ipairs(tombstones) do classes[id] = nil end
end,
gen2Fields = {
id = f.opt(f.str), name = f.str,
index = f.opt(f.int(0, 255)),
baseMoney = f.opt(f.int(0)),
-- the class's battle theme; Gen 1 spells the same idea `battleTheme`,
-- but this is the extractor's own key and a strict rename would reject
-- every one of Gold's 66 classes
encounterMusic = f.opt(f.id("music")),
-- the items the class's AI may use mid-battle, and the seven raw AI
-- bytes behind them (pokegold data/trainers/attributes.asm)
items = f.opt(f.list(f.id("items"))),
attributes = f.opt(f.list(f.int(0, 255))),
-- one entry per named trainer of the class. trainerType decides which
-- optional party fields the cart actually stores, so `moves` and `item`
-- are optional here rather than four party shapes in a union.
trainers = f.list(f.rec{
id = f.opt(f.str), name = f.str, index = f.opt(f.int(0, 255)),
trainerType = f.opt(f.enum{ "TRAINERTYPE_NORMAL", "TRAINERTYPE_MOVES",
"TRAINERTYPE_ITEM",
"TRAINERTYPE_ITEM_MOVES" }),
party = f.list(f.rec{ level = f.int(1), species = f.id("pokemon"),
item = f.opt(f.id("items")),
moves = f.opt(f.list(f.id("moves"))) }) }),
},
example = 'mod.content.trainers:patch("OPP_BROCK", { baseMoney = 99 })',
gen2Example = 'mod.content.trainers:patch("BEAUTY", { baseMoney = 99 })',
}
R.sprites = {
@@ -605,7 +1113,32 @@ R.sprites = {
-- the ROM (which is what `source` documents on imported records).
paletteSource = f.opt(f.str),
},
-- Same name, same ids, and (unlike the rest of this section) already
-- routed: `sprites` merges into data.gen2Sprites and Gold walks that very
-- table. The Gen 1 schema accepts a Gen 2 record only because the extra
-- keys fall through as unknown-but-preserved, which means none of them is
-- checked -- a mod could write paletteId = "blue" and find out at draw
-- time. This types them.
gen2Fields = {
id = f.opt(f.str), image = f.path, frames = f.int(1),
walker = f.opt(f.bool), trueColor = f.opt(f.bool),
paletteSource = f.opt(f.str),
-- the OBJ palette this sprite draws with, by name and by the slot index
-- src/world/gen2/Palettes.lua indexes into (PAL_OW_RED is slot 0)
palette = f.opt(f.str), paletteId = f.opt(f.int(0, 7)),
-- how the overworld animates it: WALKING_SPRITE has the four facings and
-- a step cycle, STANDING_SPRITE only the facings, STILL_SPRITE one frame,
-- POKEMON_SPRITE the party-icon pair (pokegold constants/sprite_constants)
spriteType = f.opt(f.enum{ "WALKING_SPRITE", "STANDING_SPRITE",
"STILL_SPRITE", "POKEMON_SPRITE" }),
-- a POKEMON_SPRITE names the species it follows and the party icon it
-- borrows its art from
species = f.opt(f.id("pokemon")), icon = f.opt(f.str),
source = f.opt(f.str),
},
example = 'mod.content.sprites:register("SPRITE_HERO", { image = "...", frames = 6 })',
gen2Example = 'mod.content.sprites:patch("SPRITE_BEAUTY", '
.. '{ palette = "PAL_OW_RED", paletteId = 0 })',
}
R.text = {
@@ -839,7 +1372,48 @@ R.battle_anims = {
into[key] = registry:get(id)
end
end,
-- Gold's battle animations are the cart's own bytecode, not a Lua sequence:
-- data.gen2BattleAnims is a script POOL keyed by ROM pointer plus the name
-- tables that index into it (a move id or an ANIM_* id resolves to a
-- pointer), and the object/frameset/OAM/graphics tables the scripts spawn
-- from. src/battle/gen2/AnimRunner.lua walks exactly those. The id is the
-- table, and `patch` adds one object without restating the pool.
--
-- The Gen 1 write/baseAt/baseIds trio is cleared rather than reused: it
-- routes an id into a per-kind subtable by prefix, and here the ids ARE the
-- subtables, so the plain record placement is the correct one.
gen2Write = false, gen2BaseAt = false, gen2BaseIds = false,
gen2Keys = {
-- pointer -> the decoded command rows the runner steps; each row is a
-- verb string followed by its operands
scripts = f.map(f.str, f.list(f.list(f.any))),
-- the pool in ROM order, which is what a re-export writes back
scriptOrder = f.list(f.str),
-- move id -> script pointer, and ANIM_* id -> script pointer
moves = f.map(f.str, f.str),
ids = f.map(f.str, f.str),
-- an animation object: which graphics, palette, frameset and update
-- function it spawns with
objects = f.map(f.str, f.rec{ gfx = f.str, palette = f.str,
frameset = f.str, func = f.str,
fixY = f.opt(f.int(0, 255)),
flags = f.opt(f.int(0, 255)) }),
-- a frameset is its own little row list (frame / wait / delete)
framesets = f.map(f.str, f.list(f.list(f.any))),
-- OAM: the sprite rectangle an object draws, and the VRAM tile it starts
-- at. x/y are the cart's unsigned bytes, so 240 means -16.
oamsets = f.map(f.str, f.rec{ vtile = f.int(0, 255),
sprites = f.list(f.rec{
x = f.int(0, 255), y = f.int(0, 255),
tile = f.int(0, 255),
attr = f.int(0, 255) }) }),
gfx = f.map(f.str, f.rec{ image = f.path, tiles = f.int(1),
wide = f.int(1) }),
bank = f.int(0), source = f.str, generation = f.int(1),
},
example = 'mod.content.battle_anims:register("SHADOW_BALL", { seq = { ... } })',
gen2Example = 'mod.content.battle_anims:patch("moves", '
.. '{ SHADOW_BALL = "5e86" })',
}
R.transitions = {
@@ -1052,7 +1626,45 @@ R.palettes = {
return ("needs exactly 4 colors, got %d"):format(#colors)
end
end,
-- Gold's palette table is not a flat name -> four colours map: the GBC has
-- eight BG and eight OBJ slots and the cart reloads them per context, so
-- the extractor writes one subtable per context (mon pics with their shiny
-- twin, trainer pics, the BG rows a map's environment indexes into, the
-- overworld OBJ rows per time of day, the town roof pair, the HP and EXP
-- bars). The id is the context, so a mod that recolours one species
-- patches `pokemon` and leaves the other 250 alone. The Gen 1 four-colour
-- `extra` is cleared: it reads the record as one palette, and here a record
-- is a whole subtable of them.
gen2Extra = false,
gen2Keys = {
-- every species has both a normal and a shiny row; the shiny one is what
-- src/render/GbcPalette.lua swaps in on a shiny battler
pokemon = f.map(f.str, f.rec{ normal = gen2PaletteRow,
shiny = gen2PaletteRow }),
trainers = f.map(f.str, gen2PaletteRow),
-- the BG rows, indexed by number: `environments` names eight of them per
-- environment per time of day, which is how a map gets its palette
bg = f.list(gen2PaletteRow),
environments = f.map(f.str, f.map(gen2PaletteTod, f.list(f.int(0)))),
-- the eight overworld OBJ rows per time of day; a sprite's paletteId
-- indexes this
objects = f.map(gen2PaletteTod, f.list(gen2PaletteRow)),
-- one pair per roof group (keyed by the group number, 0 included), and
-- the BG slot the roof colours are written into
roofs = f.map(f.int(0), f.rec{ mornDay = gen2PaletteRow,
nite = gen2PaletteRow }),
roofSlot = f.int(0, 7),
hpBar = f.map(f.enum{ "green", "yellow", "red", "blue" }, gen2PaletteRow),
expBar = gen2PaletteRow,
partyMenu = f.list(gen2PaletteRow),
battleObjects = f.map(f.str, gen2PaletteRow),
-- the ordered name lists the numeric indices above resolve through
daytimes = f.list(f.str), slotNames = f.list(f.str),
source = f.str, generation = f.int(1),
},
example = 'mod.content.palettes:override("MEWMON", { {255,255,255}, ... })',
gen2Example = 'mod.content.palettes:patch("pokemon", '
.. '{ TOTODILE = { shiny = { {255,255,255}, {255,0,0} } } })',
}
-- keyed by species id, unlike the vanilla byDex array: a species past the
@@ -1061,10 +1673,63 @@ R.palettes = {
-- dex-indexed default. The value is a built-in icon NAME -- one of BALL, BIRD,
-- BUG, FAIRY, GRASS, HELIX, MON, QUADRUPED, SNAKE, WATER (uppercase) -- or a
-- { image = <bundled file path>, frames? } table of your own art.
-- Gold splits the same idea in two: data.gen2Icons.icons is the 39 icon
-- SHEETS (each its own two-frame image) and data.gen2Icons.species is the
-- species -> sheet name assignment. Both halves keep the Gen 1 id space --
-- a species id names an assignment, a sheet id names a sheet -- so one
-- registry serves both, routed by the ICON_ prefix every sheet name carries.
-- Two id forms in one registry is the same shape font and battle_anims use.
local function gen2IconIsSheet(id)
return tostring(id):match("^ICON_") ~= nil
end
R.icons = {
semantics = "record", target = "icons.bySpecies",
value = f.union{ f.str, f.rec{ image = f.path, frames = f.opt(f.int(1)) } },
gen2Value = f.union{
-- the assignment form: a species id mapped to a sheet name
f.str,
-- the sheet form: width/height are the sheet's pixel size, and every
-- vanilla sheet is a 16x32 two-frame strip
f.rec{ id = f.opt(f.str), index = f.opt(f.int(0, 255)), image = f.path,
width = f.int(1), height = f.int(1), frames = f.int(1) },
},
gen2Extra = function(id, value)
if gen2IconIsSheet(id) then
if type(value) ~= "table" then
return "an ICON_ id is a sheet and needs an image, width, height and frames"
end
elseif type(value) ~= "string" then
return "a species id takes the NAME of an ICON_ sheet, not a sheet"
end
end,
gen2BaseAt = function(base, id)
if gen2IconIsSheet(id) then return base.icons and base.icons[id] or nil end
return base.species and base.species[id] or nil
end,
gen2BaseIds = function(base)
local ids = {}
for id in pairs(base.icons or {}) do ids[#ids + 1] = id end
for id in pairs(base.species or {}) do ids[#ids + 1] = id end
return ids
end,
gen2Write = function(target, registry)
local sheets, species = target.icons, target.species
if not sheets then
sheets = {}
target.icons = sheets
end
if not species then
species = {}
target.species = species
end
for _, id in ipairs(registry.order) do
local into = gen2IconIsSheet(id) and sheets or species
into[id] = registry:get(id)
end
end,
example = 'mod.content.icons:register("MODMON", "QUADRUPED") -- a built-in name, or { image = mod.assets:path("icon.png"), frames = 2 }',
gen2Example = 'mod.content.icons:override("TOTODILE", "ICON_MONSTER")',
}
-- glyph codes are not bytes: the vanilla pages sit at $60/$80 but a
@@ -1178,6 +1843,44 @@ R.tokens = {
-- ------- deep registries: id is a top-level key of the target table
-- Gold's `constants` is not the Gen 1 rule block at all: it is the ROM's own
-- ordered name lists, one per enum the cart indexes by number. A script
-- opcode that says "special 12" or an animation that says "object 41" is
-- resolved through these, so replacing an entry renames what that number
-- means. Every one of them is a dense list of ids in ROM order, which is why
-- they can be built from a name list instead of restated one by one.
local GEN2_CONSTANT_ORDERS = {
"battleAnimBgPaletteOrder", "battleAnimFramesetOrder", "battleAnimFuncOrder",
"battleAnimGfxOrder", "battleAnimOamsetOrder", "battleAnimObPaletteOrder",
"battleAnimObjectOrder", "battleBgEffectOrder", "cmdQueueOrder",
"decoDescOrder", "eggGroupOrder", "environmentOrder", "evolveMethodOrder",
"fishGroupOrder", "floorOrder", "growthRateOrder", "heldEffectOrder",
"iconOrder", "itemMenuOrder", "itemOrder", "landmarkOrder",
"mapCallbackOrder", "mapOrder", "moveEffectOrder", "moveOrder", "musicOrder",
"paletteOrder", "phoneContactOrder", "pocketOrder", "sfxOrder", "spawnOrder",
"specialCallOrder", "specialOrder", "speciesOrder", "spriteOrder",
"stdScriptOrder", "tilesetOrder", "tradeDialogOrder", "tradeGenderOrder",
"trainerClassOrder", "trainerTypeOrder", "treeMonSetOrder",
}
local gen2ConstantKeys = {
-- the map table the group/number pair in a warp resolves through
mapGroups = f.list(f.rec{ group = f.int(0), map = f.int(0), name = f.str,
width = f.int(1), height = f.int(1) }),
-- class id -> its named trainers, in the order the class's table stores them
trainerClassMembers = f.map(f.str, f.list(f.str)),
-- type id -> its ROM byte; the only one of these that is a lookup rather
-- than an ordered list, because the type numbers are not contiguous
types = f.map(f.str, f.int(0)),
-- counts the extractor stamps beside the lists
itemNameCount = f.int(0), numOverworldSprites = f.int(0),
spritePokemon = f.int(0),
source = f.str, generation = f.int(1),
}
for _, name in ipairs(GEN2_CONSTANT_ORDERS) do
gen2ConstantKeys[name] = f.list(f.str)
end
-- The rules the engine used to hard-code as Kanto/Red literals. Keys the
-- importer does not stamp are seeded with their vanilla value at data load
-- (src/core/Data.lua) so a patch always has something to fold over.
@@ -1195,7 +1898,16 @@ R.constants = {
hmMoves = f.list(f.id("moves")),
encounterBuckets = f.list(f.int(1, 256)),
},
-- Gold's keys are ordered lists where position IS the id a script byte
-- resolves through, so they must replace rather than append -- which is
-- what "deep" semantics would do to them (Merge.deepMerge concatenates
-- lists there, and Gen 1's `field` rows genuinely want that). A key
-- neither catalog names is still a mod's own data and merges as-is.
gen2Semantics = "record",
gen2Keys = gen2ConstantKeys,
example = 'mod.content.constants:patch("levelCap", 80)',
gen2Example = 'mod.content.constants:patch("speciesOrder", '
.. '{ [252] = "MODMON" })',
}
-- The overworld's data grab bag. Only the keys this milestone routes are
@@ -1267,6 +1979,150 @@ R.text_pointers = {
example = 'mod.content.text_pointers:patch("PalletTown", { TEXT_PALLETTOWN_SIGN = { text = "_MySign" } })',
}
-- ------- Gen 2 only content
--
-- The mirror of the gated rows in Schemas.GEN2: six systems Gold has and Red
-- does not, so there is no Gen 1 table to share a target with and no Gen 1
-- consumer to read one. Each spec therefore carries NO `target` at all -- the
-- routed Schemas.GEN2 path is its only home -- and a Schemas.GEN1 row of
-- `false`, which is what turns a Red mod's write into the same reported drop a
-- Gold mod gets for `tokens` instead of a silent merge into a namespace
-- nothing on Red would ever read.
--
-- Names stay plain for the same reason hook and event names do: `decorations`
-- is what the thing is called, and a "gen2Decorations" registry NAME would be
-- a namespace no mod could ever share if Gen 1 grew the system later. Only
-- the Data path underneath carries the gen2 prefix.
-- data/items/attributes.asm's last two columns, split out of the item record
-- so a mod can give an item a held behaviour without owning the whole item.
-- src/core/Game2.lua seeds the merge target from data.items and writes the
-- merged rows back onto it, and src/battle/gen2/Battle.lua's heldEffect (the
-- one read all eight held-item sites go through, and the held_item.trigger
-- hook's own site) reads it from there.
R.held_items = {
semantics = "record",
fields = {
-- the HELD_* name the battle compares against, out of
-- data.gen2Constants.heldEffectOrder; a mod may invent its own and steer
-- it from the held_item.trigger hook
heldEffect = f.str,
-- ItemAttributes' parameter byte: the boost percentage, the heal amount,
-- the BrightPowder odds -- whatever the effect reads it as
heldParameter = f.opt(f.int(0, 255)),
},
example = 'mod.content.held_items:override("LEFTOVERS", '
.. '{ heldEffect = "HELD_LEFTOVERS", heldParameter = 0 })',
}
-- data/phone/phone_contacts.asm, one record per PHONE_* row. The id space is
-- data.gen2Constants.phoneContactOrder, so PHONE_YOUNGSTER_JOEY names the row
-- the cart calls PHONE_YOUNGSTER_JOEY; `index` is that row's byte, which is
-- what the save's contact list holds and what src/core/gen2/Phone.lua keys
-- every one of its own lookups by. The four PHONE_UNUSED const_skip holes are
-- not registered -- they are copies of the wrong-number filler row, and one id
-- cannot name four of them.
R.phone_contacts = {
semantics = "record",
fields = {
index = f.int(0),
-- non-trainer rows (MOM, BILL, ELM, the BIKE SHOP) carry a PHONECONTACT_*
-- number instead of a trainer; trainer rows carry the class and the
-- roster member, which is what the rematch machinery and the caller's
-- name are looked up by
number = f.opt(f.int(0, 255)),
class = f.opt(f.str), member = f.opt(f.str),
map = f.opt(f.id("maps")),
-- the SCRIPT1 / SCRIPT2 time masks: MORN | DAY | NITE, 0 for "never"
calleeTime = f.opt(f.int(0, 7)), callerTime = f.opt(f.int(0, 7)),
-- the script LABEL (Phone.SCRIPT_KEYS resolves it) and, once the cache
-- has been read, the "<bank>:<addr>" pointer it resolved to
callee = f.opt(f.str), caller = f.opt(f.str),
calleeKey = f.opt(f.str), callerKey = f.opt(f.str),
},
example = 'mod.content.phone_contacts:patch("PHONE_YOUNGSTER_JOEY", '
.. '{ map = "ROUTE_31" })',
}
-- data/decorations/attributes.asm, one record per DECO_* row. The cart's
-- decoration constants are a bare const_def block with no name table behind
-- them -- nothing in the ROM spells DECO_FEATHERY_BED -- so the id is the
-- attribute row's own index, written "deco:<n>" the way battle_anims writes
-- "subanim:<n>". That index IS wMenuSelection, which is what every caller
-- passes src/core/gen2/Decorations.lua.
R.decorations = {
semantics = "record",
fields = {
-- constants/deco_constants.asm decoration types: 1 PLANT, 2 BED,
-- 3 CARPET, 4 POSTER, 5 DOLL, 6 BIGDOLL. The type decides how GetDecoName
-- spells the row and whether `sprite` is a block id or a sprite one.
type = f.int(1, 6),
name = f.str,
-- DECOATTR_ACTION, as the Decorations.ACTIONS key rather than the
-- jumptable index; nil on the CANCEL row alone
action = f.opt(f.str),
-- DECOATTR_EVENT_FLAG: the wEventFlags bit that says the player owns it
flag = f.int(0),
-- DECOATTR_SPRITE: a BLOCK id for the four kinds the map paints, a
-- SPRITE_* byte for the four an object stands on
sprite = f.int(0, 255),
},
example = 'mod.content.decorations:patch("deco:2", { name = "COZY" })',
}
-- data/items/apricorn_balls.asm. Id = the apricorn item, because that is what
-- the player hands Kurt and what FindApricornsInBag walks the bag for;
-- `index` is the row's position in that table, which is load bearing twice
-- (Kurt's menu order and the checkevent chain in maps/KurtsHouse.asm).
R.apricorns = {
semantics = "record",
fields = {
apricorn = f.id("items"), ball = f.id("items"),
-- constants/event_flags.asm index of this apricorn's EVENT_GAVE_KURT_*
event = f.int(0),
index = f.int(1),
},
example = 'mod.content.apricorns:override("RED_APRICORN", '
.. '{ apricorn = "RED_APRICORN", ball = "ULTRA_BALL", event = 600, index = 1 })',
}
-- data/maps/landmarks.asm. The town-map places, which on Gold are one index
-- space shared by the Pokegear MAP card, the #DEX AREA page and every map
-- header's `landmark` byte. The merge lands inside the cache's own landmark
-- table (gen2Landmarks.landmarks), so a registered record is one the map card
-- can already draw.
R.landmarks = {
semantics = "record",
fields = {
id = f.opt(f.str),
-- the two-line name the town map prints, "\n" and all
name = f.str,
-- the marker's tile position on the 20x18 town map
x = f.int(0), y = f.int(0),
-- LANDMARK_*: the byte a map header carries, and what
-- src/core/gen2/Nests.lua's region split reads
index = f.int(0),
},
example = 'mod.content.landmarks:patch("LANDMARK_ROUTE_29", { x = 12 })',
}
-- PlayRadioStationPointers (engine/pokegear/pokegear.asm). Id = the station
-- the dial resolves to, which is the LoadStation_* id the show state machine
-- is keyed by; `channel` is its MAPRADIO_* dial position, the byte a wall
-- radio's `setval` passes to the MapRadio special. Position 0 is not a
-- station: it resolves by region and time of day, so no record claims it.
R.radio_channels = {
semantics = "record",
fields = {
channel = f.int(0, 255),
-- the name quoted in the text box; without one the Pokegear's own
-- STATION_NAMES row is used, which is where the vanilla eight get theirs
name = f.opt(f.str),
},
example = 'mod.content.radio_channels:register("PIRATE_RADIO", '
.. '{ channel = 9, name = "PIRATE RADIO" })',
}
-- ------- persistence
-- compose, keyed by the owning mod id: the runner walks each owner's chain