`prev_state` starts off as a copy of `swap_state` and then `swap_state`
collects `item_to_place`. Collecting an item must never reduce
accessibility (otherwise generation breaks horribly), so it is
guaranteed that `swap_state` will always be able to reach at least as
many locations as `prev_state`, so `new_loc_count >= prev_loc_count` is
always `True`.
As a sideeffect of this change, this fixes generation of Pokemon Emerald
with locally shuffled Badges/HMs when there are worlds with unconnected
entrances present in the multiworld e.g. KH1. This is because this
location counting did not respect `single_player_placement=True` and
counted reachable locations across the entire multiworld.
Fixes#4834 as a sideeffect of removing the redundant code.
* Core: Speed up fill_restrictive item_pool pop loop
Items from `reachable_items` are placed in last-in-first-out order, so
items being placed will be towards the end of `item_pool`, but the
iteration to find the item was iterating from the start of `item_pool`.
Now also uses `del` instead of `.pop()` for an additional, tiny,
performance increase.
It is unlikely for there to be a noticeable difference in most cases.
Only generating with many worlds with a high percentage of progression
items and fast access rules is likely to see a difference with this
change.
--skip_output generation of 400 template A Hat in Time yamls with
progression balancing disabled goes from 76s to 43s (43% reduction) for
me with this patch. This placed 43200 progression items out of 89974
items total (48% progression items).
* Fix comment typo
"be" was missing.
---------
Co-authored-by: Exempt-Medic <60412657+Exempt-Medic@users.noreply.github.com>
The base state passed to fill_restrictive should be as maximal as
possible otherwise fill_restrictive has to repeatedly re-sweep and
collect from advancement locations that were reachable from before
fill_restrictive has placed a single item.
This is not added within fill_restrictive itself because it is common
for fills to be performed using a partial 'all_state', which is already
a maximum exploration state.
With --skip_output generation of every template yaml, except FF, KH
and Shivers, this prevented repeatedly re-sweeping 576 advancement
locations in every sweep within progression fill, reducing the
generation time from 124s to 113s for me (8.8% reduction, averaged over
5 generations each).
* Actually take item from pool when plandoing from_pool
* Remove the awkward index thing
* oops left a comment in
* there wasn't a line break here before
* Only remove if actually found, check against player number
* oops
* Go back to index based system so we can just remove at the end
* Comment
* Fix error on None
* Update Fill.py
Co-authored-by: Exempt-Medic <60412657+Exempt-Medic@users.noreply.github.com>
---------
Co-authored-by: Exempt-Medic <60412657+Exempt-Medic@users.noreply.github.com>
* fix single player item links
* Make a variable and fix weird spacing
* use advancement instead of classification
---------
Co-authored-by: Exempt-Medic <60412657+Exempt-Medic@users.noreply.github.com>
* Rename sweep_for_events to sweep_for_advancements
* more event->advancement renames
* oops accidentally deleted the deprecation thing in the force push
* Update TestDungeon.py
* Update BaseClasses.py
* Update BaseClasses.py
* oops
* utils.deprecate
* treble, you had no idea how right you were
* Update test_panel_hunt.py
* Update BaseClasses.py
Co-authored-by: Fabian Dill <Berserker66@users.noreply.github.com>
---------
Co-authored-by: Fabian Dill <Berserker66@users.noreply.github.com>
* Core: dump all item placements for generation failures
* pass the multiworld from remaining fill
* change how the args get handled to fix formatting
---------
Co-authored-by: NewSoupVi <57900059+NewSoupVi@users.noreply.github.com>
* Core: allow for single player state sweeping
* Fill: have distribute items use single_player fill when it can
* oop
* pass locations to sweep_for_events instead of the player
* finally found the diff that was breaking swap
* LTTP fills everyone's dungeons at once, not just a single player's
---------
Co-authored-by: NewSoupVi <57900059+NewSoupVi@users.noreply.github.com>
* Fill: add test for swap error with item rules
https://discord.com/channels/731205301247803413/731214280439103580/1167195750082560121
* Fill: fix swap error found in CI
Swap now assumes the unplaced items can be placed before the to-be-swapped item.
Unsure if that is safe or unsafe.
* Test: clarify docstring and comments in fill swap test
* Test: clarify comments in fill swap test more
🤞
* map option objects to a `World.options` dict
* convert RoR2 to options dict system for testing
* add temp behavior for lttp with notes
* copy/paste bad
* convert `set_default_common_options` to a namespace property
* reorganize test call order
* have fill_restrictive use the new options system
* update world api
* update soe tests
* fix world api
* core: auto initialize a dataclass on the World class with the option results
* core: auto initialize a dataclass on the World class with the option results: small tying improvement
* add `as_dict` method to the options dataclass
* fix namespace issues with tests
* have current option updates use `.value` instead of changing the option
* update ror2 to use the new options system again
* revert the junk pool dict since it's cased differently
* fix begin_with_loop typo
* write new and old options to spoiler
* change factorio option behavior back
* fix comparisons
* move common and per_game_common options to new system
* core: automatically create missing options_dataclass from legacy option_definitions
* remove spoiler special casing and add back the Factorio option changing but in new system
* give ArchipIDLE the default options_dataclass so its options get generated and spoilered properly
* reimplement `inspect.get_annotations`
* move option info generation for webhost to new system
* need to include Common and PerGame common since __annotations__ doesn't include super
* use get_type_hints for the options dictionary
* typing.get_type_hints returns the bases too.
* forgot to sweep through generate
* sweep through all the tests
* swap to a metaclass property
* move remaining usages from get_type_hints to metaclass property
* move remaining usages from __annotations__ to metaclass property
* move remaining usages from legacy dictionaries to metaclass property
* remove legacy dictionaries
* cache the metaclass property
* clarify inheritance in world api
* move the messenger to new options system
* add an assert for my dumb
* update the doc
* rename o to options
* missed a spot
* update new messenger options
* comment spacing
Co-authored-by: Doug Hoskisson <beauxq@users.noreply.github.com>
* fix tests
* fix missing import
* make the documentation definition more accurate
* use options system for loc creation
* type cast MessengerWorld
* fix typo and use quotes for cast
* LTTP: set random seed in tests
* ArchipIdle: remove change here as it's default on AutoWorld
* Stardew: Need to set state because `set_default_common_options` used to
* The Messenger: update shop rando and helpers to new system; optimize imports
* Add a kwarg to `as_dict` to do the casing for you
* RoR2: use new kwarg for less code
* RoR2: revert some accidental reverts
* The Messenger: remove an unnecessary variable
* remove TypeVar that isn't used
* CommonOptions not abstract
* Docs: fix mistake in options api.md
Co-authored-by: Doug Hoskisson <beauxq@users.noreply.github.com>
* create options for item link worlds
* revert accidental doc removals
* Item Links: set default options on group
* change Zillion to new options dataclass
* remove unused parameter to function
* use TypeGuard for Literal narrowing
* move dlc quest to new api
* move overcooked 2 to new api
* fixed some missed code in oc2
* - Tried to be compliant with 993 (WIP?)
* - I think it all works now
* - Removed last trace of me touching core
* typo
* It now passes all tests!
* Improve options, fix all issues I hope
* - Fixed init options
* dlcquest: fix bad imports
* missed a file
* - Reduce code duplication
* add as_dict documentation
* - Use .items(), get option name more directly, fix slot data content
* - Remove generic options from the slot data
* improve slot data documentation
* remove `CommonOptions.get_value` (#21)
* better slot data description
Co-authored-by: black-sliver <59490463+black-sliver@users.noreply.github.com>
---------
Co-authored-by: el-u <109771707+el-u@users.noreply.github.com>
Co-authored-by: Doug Hoskisson <beauxq@users.noreply.github.com>
Co-authored-by: Doug Hoskisson <beauxq@yahoo.com>
Co-authored-by: black-sliver <59490463+black-sliver@users.noreply.github.com>
Co-authored-by: Alex Gilbert <alexgilbert@yahoo.com>
#1800 introduced a cleanup pass to "eject" unreachable items to hopefully get better error reporting of what could not be placed. The code to do so has an unnecessary amount of sweeps from pool.
* Tests: add test for mixing minimal and non-minimal
* Tests: minor cleanup in test_minimal_mixed_fill
* fix fill_restrictive for mixed minimal/non-minimal
The reason why this only happens for minimal is because it would not accept the solution it found otherwise.
Tracking and releasing unreachable items would be the better solution, but that's a lot harder to do.
* fix typo in fill_restrictive
* fix pep8 in fill_restrictive
* Fill: cleanup invalid unsafe placements, better comments
* Fill: more cleanup
* Core: Fix divide by zero error if all locations are priority
* Core: 100% makes more sense than 0% in this context
* Core: Revert a some "autoformatter" damage
* Core: Move comparisons a bit earlier.
Worlds that are 100% filled with locked locations should now be completely skipped in the progression balancing step now as we filter them out at the very beginning of the stage now. Also skips progression balancing if it turns out all locations are locked early before attempting to balance.
* Early Items option
* Early Items description update
* Change Early Items to dict
* Rewrite early items as extra fill steps
* Move if statement up
* Document early_items
* Move early_items handling before fill_hook
* Apply suggestions from code review
Co-authored-by: Doug Hoskisson <beauxq@users.noreply.github.com>
* Subnautica pre-fill moved to early_items
* Subnautica early items fix
* Remove unit test bug workaround
* beauxq's pr
Co-authored-by: Doug Hoskisson <beauxq@users.noreply.github.com>
