Rewrites a good amount of legality APIs pertaining to:
* Legal moves that can be learned
* Evolution chains & cross-generation paths
* Memory validation with forgotten moves
In generation 8, there are 3 separate contexts an entity can exist in: SW/SH, BD/SP, and LA. Not every entity can cross between them, and not every entity from generation 7 can exist in generation 8 (Gogoat, etc). By creating class models representing the restrictions to cross each boundary, we are able to better track and validate data.
The old implementation of validating moves was greedy: it would iterate for all generations and evolutions, and build a full list of every move that can be learned, storing it on the heap. Now, we check one game group at a time to see if the entity can learn a move that hasn't yet been validated. End result is an algorithm that requires 0 allocation, and a smaller/quicker search space.
The old implementation of storing move parses was inefficient; for each move that was parsed, a new object is created and adjusted depending on the parse. Now, move parse results are `struct` and store the move parse contiguously in memory. End result is faster parsing and 0 memory allocation.
* `PersonalTable` objects have been improved with new API methods to check if a species+form can exist in the game.
* `IEncounterTemplate` objects have been improved to indicate the `EntityContext` they originate in (similar to `Generation`).
* Some APIs have been extended to accept `Span<T>` instead of Array/IEnumerable
* Reuses move parse result objects for each encounter parsed in a LegalityCheck attempt, instead of creating a new object.
* Ensures the objects are never-null, and makes cleanup easier.
Slightly adjusts some other parts of the moveset validation to reduce allocations.
Change the data structure so that different locations are tied to the table, rather than separate usages of the table.
This de-bloats from 344KB->85KB and has less runtime memory consumption.
Since we have more metadata with move learn sourcing, we can check if it was traded to gen2 to get new moves / deleted.
Adjust call sites appropriately
might have some issues, to be ironed out maybe
EncounterArea now stores a more specific type'd array for encounter slots. Better iteration and less casting, as the nonspecific `Slots` fetch is rarely referenced.
EncounterType renamed to GroundTile to reflect how it actually works in Gen4. Was previously an ambiguous field that was clarified a little; we can describe it a little better now. Keep the GUI the same to not scare the end users.
Change Trash Byte properties to get/set a Span. Trash Byte legality checking easier on the garbage collector?
AltForm & Form & Forme => Form
GenNumber & Generation => Generation
Extract out SpeciesForm interface, and re-add IGeneration
For those using PKHeX as a dependency, this should be a pretty straightforward manual replacement... GenNumber and AltForm should be quick find-replace`s.
Change Ability array to IReadOnlyList, add method to check ability index in personal data
Suppress some message warnings
Change EvolutionChain short-circuit for VC to jump from gen6 directly down to gen2. There aren't any notradeback 1 situations, so a notradeback1 will always start with g=1, so no need for the other if-continue.
Simplify pk5 conversion
* Rework gen1 slot loading
Slot templates are precomputed from ROM data and just loaded straight in, with tight coupling to the encounter area (grouped by slot types).
* Revise fuzzy met check for underleveled wild evos
Example: Level 23 poliwhirl in RBY as a level 50 poliwhirl, will assume the chain is 25-50 for poliwhirl (as poliwag evolves at 25). Instead of revising the origin chain, just ignore the evo min level in the comparison.
Previous commit fixed it for gen1.
* Rework gen2-4 slot loading
Gen4 not finished, Type Encounter data and some edge encounters not recognizing yet...
* Add feebas slots for old/good encounters
* Begin moving properties
Great news! Gen5-7 need to be de-dumbed like Gen1-4.
Then I can remove the bang (!) on the Area accessor and ensure that it's never null!
* Split off XD pokespot slot encounter table type
* Set area in constructor
* Deduplicate g3 roaming encounters
* Deduplicate xd encounter locations (rebattle)
Only difference is met location; no need to create 500 extra encounter objects. A simple contains check is ok (rarely in gen3 format).
* Make all slots have a readonly reference to their parent area
* Minor clean
* Remove "Safari" slot type flag
Can be determined via other means (generation-location), allows us to reduce the size of SlotType member to a byte
Output of slot binaries didn't preserve the Safari flag anyway.
* Update SlotType.cs
* Handle type encounters correctly
* Merge safari area into regular xy area
* Merge dexnav accessor logic
* fix some logic so that tests pass again
rearrange g5 dw init to be done outside of static constructor (initializer instead)
PIDGenerator: friend safari slots now generate with required flawless IV count
* Add cianwood tentacool gift encounter
* Remove unnecessary abstractions
Fake area just returned a slot; since Slots have a non-null reference to the area, we can just return the slot and use the API to grab a list of possible slots for the chain.
Increase restrictiveness of location/type get-set operations
* Minor tweaks, pass parameters
DexNav observed state isn't necessary to use, only need to see if it's possible to dexnav. Now that we have metadata for slots, we can.
* Remove unused legality tables
Results in a little more code, but each path is less tangled
simplify some expressions
remove RBDragonair content in favor of a strict filter for catch rate
Move gen1 trade trainer names to stringconverter, since pk1/pk2 shouldn't refer to legality classes
V### names weren't enjoyable to work with; use similar verbose style as
the program message strings.
updating the translation files with the remapped variable names shortly
remap list: https://pastebin.com/jybkVDAK
Checks.cs initially started out small, but over the years it has grown
to handle multiple types of checks. With all these checks next to
eachother, it's hard to see the overall groups. Splitting them up
(potentially further?) allows for more focused maintenance &
understanding.
Not sure if I'm happy with the overall bandaids used (checks no longer
done within LegalityAnalysis so variable repointing is excessively
used), but I'm happier the way it is now compared to the huge Checks.cs
surf does not have another slot set; so they were all being set to
Morning.
Could have set all 3 time flags, but kept as 'any'.
will update release binaries with this one
Rewrote showdown set exporting to be much clearer, added auto-fixing of
relearn moves when importing showdown set (applies suggested relearn
moves)
Simplified main form loading routine (from ~160 to ~25 by extracting
methods)
exercise in deferred execution/state machine, only calculate possible
matches until a sufficiently valid match is obtained. Previous setup
would try to calculate the 'best match' and had band-aid workarounds in
cases where a subsequent check may determine it to be a false match.
There's still more ways to improve speed:
- precalculate relationships for Encounter Slots rather than iterating
over every area
- yielding individual slots instead of an entire area
- group non-egg wondercards by ID in a dict/hashtable for faster
retrieval
reworked some internals:
- EncounterMatch is always an IEncounterable instead of an object, for
easy pattern matching.
- Splitbreed checking is done per encounter and is stored in the
EncounterEgg result
- Encounter validation uses Encounter/Move/RelearnMove/Evolution to
whittle to the final encounter.
As a part of the encounter matching, a lazy peek is used to check if an
invalid encounter should be retained instead of discarded; if another
encounter has not been checked, it'll stop the invalid checks and move
on. If it is the last encounter, no other valid encounters exist so it
will keep the parse for the invalid encounter.
If no encounters are yielded, then there is no encountermatch. An
EncounterInvalid is created to store basic details, and the parse is
carried out.
Breaks some legality checking features for flagging invalid moves in
more detail, but those can be re-added in a separate check (if
splitbreed & any move invalid -> check for other split moves).
Should now be easier to follow the flow & maintain 😄
