We implement simple state machine iterators to iterate through every split type encounter array, and more finely control the path we iterate through. And, by using generics, we can have the compiler generate optimized code to avoid virtual calls.
In addition to this, we shift away from the big-5 encounter types and not inherit from an abstract class. This allows for creating a PK* of a specific type and directly writing properties (no virtual calls). Plus we can now fine-tune each encounter type to call specific code, and not have to worry about future game encounter types bothering the generation routines.
Inline logic for Meowstic form branch evo
Pass levelup byte value directly instead of recompute
With EvolutionSet all shared, reorder the constructor so the primary ctor doesn't need manually ordered properties (8 byte total struct size)
Hoopa cannot know both moves due to form reversion
Kyurem cannot know both moves due to form reversion
Kyurem cannot know Scary Face when fused in past games
Add dexnav for AtAnyTime/HOME lookback when original moveset deleted (no longer in local relearn moves)
BDSP Underground encounters w/ egg moves can be shared by other means
Relocate checksum adders to Checksums class, improve performance by eliminating slice calls
Improve HOME sharing for GO8 and IMoveset encounters: actually sanity check GO8, and skip other non-PK7/PB7 cases.
Move enum -> ushort instead of int
Redo handling of HOME Volt Tackle (disallow on SWSH Cap Pikachu)
Pass spans instead of strings to use span methods
Reset encounter filters on early abort
Reformat all pickles to the same serialization format (latest)
(format ignores all empty evo entries, to skip allocation of dead entries)
346KB -> 60.3KB, -200 lines of code
Nintendo + ILCA = no nicknames allowed ;)
While we're here, streamline Gen4/5 evo pickles too -- trim out unused bytes, no longer need to scan for array length (perf). Adjust method IDs for gen4 so we can use the same code for gen4 & gen5.
Add sound for adding all XY fashion items (was silent)
Fix Gen7 Resort PKM loading (4 bytes extra per slot)
Fix Gen6 HallOfFame TID/SID validation dropping leading zeroes
Fix Gen6 Best Friends ribbon validation
Treat GO8 as server date restricted, give detailed message outside window
HasTracker => must have HT name
No backwards transfer to LGP/E, and this behavior is only encountered if you turn off specific filtering settings as the generator filters them out afterwards.
Improve cctor times for Breeding; direct calls for splitbreed checks; inlined binary searches via generated IL instead of hashset
Fix permitting alpha ribbon on non-alphas in Gen8/8a/8b, disallow Gen8a/8b ribbons for Gen7 Alolan Raichu
Improve some IL generation of EvoChain logic
Add xmldoc for new evotree additions
GO imports that are created with their PK7 and PB7 movesets can bleed from their initial moves.
Example: Muk-Alola level 20 with Minimize can bleed into SV at level 20 (below the normal learn at level 21).
Like move validation, evolutions are the earliest thing we wish to traverse when determining what encounters may have originated the current Pokémon. To determine the permitted species-form-levels a Pokémon could originate with, we must devolve a Pokémon by traveling down-generation to origin. Once we have an encounter, we can then evolve it to the current species, traversing upwards from origin to the current format.
* Fix Pokemon tooltip for LA incorrectly referring to SwSh location table
* Fixes to comments referring to incorrect generation/version/game
---------
Co-authored-by: James Park <5295838+pencilethics@users.noreply.github.com>
* Heavily rewrites the `PKH` abstractions.
* Uses HOME's core-side classes as the transfer middlemen instead of direct A->B transfers.
* Revises logic to account for most of HOME's quirks (scale/height copying, safe refuge PLA)
Future revisions hinge on better handling of evotree (need better metadata about existing as specific evolutions in each game).
---------
Co-authored-by: sora10pls <17801814+sora10pls@users.noreply.github.com>
Co-authored-by: Lusamine <30205550+Lusamine@users.noreply.github.com>
https://projectpokemon.org/home/forums/topic/63429-bdsp-slate/
The "Give All" cheat will skip adding unobtainable items.
Since the BDSP-exclusive Treasure (slate/shard) are unable to held, they were also skipped. Revise the logic to allow them to be added via "Give All", but still disallowed as held items (separate arrays needed).
Also condense the logic for BCAT-Dmax crystal IDs since they're all sequential. Just iterate a range instead of a span.
More readonlyspan for the rest of the defined ushort[] arrays, less dictionary/hashset
Simplify some slot-empty checks, makes it easier to see an api for slot interaction (future?)
Perhaps one of the silliest things PKHeX could account for when it comes to legality 🥒🚫
Data has been added to PGET, but the issue has not been fixed yet, and Beast Ball is still up in the air. More to come?
- Add HeldItems_XY
- Fix that HeldItem_AO pointed to what should have been HeldItems_XY
- Fix HeldItem_AO being inconsistent with the rest of the HeldItems
- Correct HeldItems_AO only having XY held items and not ones added in ORAS
* Uses LearnSource more throughout the codebase when appropriate, rather than loosely coupled pivot methods.
* Hides Learnset/EggMove data inside the LearnSource classes.
* Extracts functionality from the large Legal class & partial Table*.cs files into better performing helper classes.
* Cleans up some code from prior LearnSource commits.
* Extract/encapsulate inventory legal arrays to interface+class
Hiding them behind methods allows these to be left as ReadOnlySpan<ushort> and thus never allocate on the heap.
Also refactors out some logic for checking if an item is legal.
End result feels more maintainable and is less bloaty (no more passing in a nullable func)
Batch editing
* Add HasType filter
```
=HasType=11
.HeldItem=Meadow Plate
```
slaps a meadow plate on any pkm with grass type
Use `=PersonalType1=11` for only primary grass types; only-secondary-type grass will not match it.
Gimmighoul can't increment its form argument counter, it can only ever be 999 on Gholdengo after evolving.
Also reorder switch cases according to National Dex indexes.
Instead of looping, if the moveset is full and a new move is added, the game shifts all arr[1..] down one slot then adds the move at the end.
Since we don't need to keep track of PP/PP Ups, we can just defer the shifting and do n % 4 rotations at the end instead of n rotations (one on each move added).
Runtime/jit repoints these to the dll rather than heap if we're Little Endian (always, otherwise will allocate like before).
Eliminates quite a few static constructors, so even faster startup. Items later.
Only fetch egg moves when using them
Localize KOR gen2 string directly instead of switch expression
Remap FRE gen4 chars directly, hot path instead of dictionary lookup
Uppercase chars in place
Follow up to #3235, this time closing the book for good.
From a bulk analysis perspective, having multiple patterns[0, 17] of Vivillon for the same Gen6/7 OT is illegal. Maybe not for traded eggs?
Skips the bucket array resizing steps
a 256 key dict will garbage 5KB; StringConverter2KOR wastes 100KB of garbage.
Zukan7b: save 144 bytes and eliminate cctor
Plugins may have popups that were created on another thread as diagnostic messages; by trying to close these on our main GUI thread, the program crashes.
Just leave these other-thread forms open when we change save files instead of crashing or closing them on their proper thread.
- Allow paradox species to receive master rank ribbon
- Resize & center IV/AV rand button (localizations
were too wide for AVs)
- Highlight blue the most recently toggled Ribbon
If the criteria template is optimized, we will be able to source moves from sibling game pairs that require indications of it being traded (yay strict logic...)
ex: encgenerator now yields X/Y eggs when ORAS tutors requested in moveset and game is X/Y
Un-fix main panel, guess this might help for scaling the GUI via OS settings. Add an overload to rescale ItemSize for the vertical tab control; don't think it is ever called though.
Make EvoChain get method public for archit
Allow dragon ascent bitfetch for gen6/7
Fix hidden power type parse/trim
Remove */ from hidden power type calc
allow longer set lines (full EVs specified for Gen2 is 74 chars
allow set lines of length 1-2 to fully support trash sets for all languages
Tweak pb8->pk8 to be more straightforward
Updates from net46->net7, dropping support for mono in favor of using the latest runtime (along with the performance/API improvements). Releases will be posted as 64bit only for now.
Refactors a good amount of internal API methods to be more performant and more customizable for future updates & fixes.
Adds functionality for Batch Editor commands to `>`, `<` and <=/>=
TID/SID properties renamed to TID16/SID16 for clarity; other properties exposed for Gen7 / display variants.
Main window has a new layout to account for DPI scaling (8 point grid)
Fixed: Tatsugiri and Paldean Tauros now output Showdown form names as Showdown expects
Changed: Gen9 species now interact based on the confirmed National Dex IDs (closes#3724)
Fixed: Pokedex set all no longer clears species with unavailable non-base forms (closes#3720)
Changed: Hyper Training suggestions now apply for level 50 in SV. (closes#3714)
Fixed: B2/W2 hatched egg met locations exclusive to specific versions are now explicitly checked (closes#3691)
Added: Properties for ribbon/mark count (closes#3659)
Fixed: Traded SV eggs are now checked correctly (closes#3692)
51875d27b2
Previous change ^ was short circuiting logic for gen5 eggs (unavailable HA for Serperior)
probably need to rewrite this method or maybe just delete these lines since they should fall through.
All Pokémon in the Paldea Pokédex are eligible, except for Legendaries, Sub-Legendaries, and Paradox.
Currently ineligible foreigners are Charmander/Charmeleon/Charizard, Wooper-0/Quagsire, Meowth-2/Perrserker
Now matches the Mightiest mark and other "isMissing" state checks.
Also indicate for Alpha marks since the logic check is the same != style.
Closes#3640
ty @Kitcham
Adds support for Scarlet & Violet.
Co-Authored-By: SciresM <8676005+SciresM@users.noreply.github.com>
Co-Authored-By: Matt <17801814+sora10pls@users.noreply.github.com>
Co-Authored-By: Lusamine <30205550+Lusamine@users.noreply.github.com>
Can unroll the loop, but would look more ugly :)
JIT will probably do it for us.
Fix <= to < for IVs
Inline temp var (start is unused after assigning to low)
Big thanks to Parzival from the RNG discord community for chiseling the LCRNG search space down to the best performing implementation possible.
50x? Down from O(2^8) -> O(2^3) is 32x, but we no longer need to access two heap arrays (262KB no longer needed!). Everything can be calculated tightly with the stack.
f641f3eab2/RNG/LCG_Reversal.py (L31)
Rainbow tables is the only faster implementation. However, nobody is gonna hog many GB of RAM for O(1) reversals. This is ~O(2^3), down from O(2^8). Much better than the days of O(2^16)!
