* Implemented link creation based on link plans
* Improvised an implementation of door placement in DefaultDoorDecorator
for testing purposes. I'll provide a better version later.
Known issues:
1. Doors with one-way links to other rooms will generate a return door
at the destination. The return door doesn't actually lead back, it leads
to a new pocket. Need to disable pair generation for doors in mazes.
2. Consider weighing sections by the door capacity to avoid putting a
lot of doors into a small section. Also double-check that room selection
within sections is unbiased.
* Increased the chance of decorating unprotected rooms from 1/4 to 1/3.
* Made minor optimizations to the box building function.
* Added some important arguments to BaseDecorator.decorate(). Also
updated all decorators as a result of this change.
* Implemented TorchDecorator so that mazes have light sources in them.
* Finished implementing link planning for mazes. Doors aren't placed yet
because that's up to Decorators and those haven't been implemented yet.
* Added bounding walls to mazes.
* Added decay effects to mazes.
Tweaked maze section generation to use a random MAX_DISTANCE for
including rooms in a section. Also changed the code to perform room
removals as sections are processed rather than deferring them to the
end. Deferring removals would cause the algorithm to detect holes from
rooms that were going to be removed. This made section generation much
stricter than necessary.
* Changed saving code to create backups by moving existing files rather
than creating copies and deleting the originals.
* Removed final call to PocketManager.save() in PocketManager.unload().
Since we no longer check if the caller is the client or server and
unload() must be called from both, this prevents clients from trying to
save pocket data locally. A final save() call wasn't needed anyway.
Started implementing the placement of Dimensional Doors in mazes.
Currently, a design is guaranteed to have enough space for some doors.
MazeDesigner still needs more code to plan out which rooms will have
doors and where those doors will lead.
Made a minor change to PocketBuilder so that mazes generate instead of
regular pocket dimensions. I'm only doing this to test dungeon
generation - it'll get switched back once mazes are ready.
Rewrote portions of our maze generation code to use RoomData. This
provides an object that unifies all room data instead of having it
spread across various data structures and linked loosely by hash maps.
We'll need this to implement the remaining generation features.
Added a flag to NewDimData so that we can avoid writing dimensions to
disk if they haven't been modified. They're still rewritten when the
server shuts down.
Made saving a bit more robust, now we only delete ones that where not
modified. Also overhauled door placement, its all handled by the
eventHandler now.
* Changed DDTeleporter to stop us from generating exits to Witchery's
Spirit World - this would cause people to lose their items upon leaving
the dimension.
* Changed GatewayGenerator to stop us from generating gateways or rift
clusters in the Spirit World
Changed FillContainersOperation so that empty dispensers are not filled
with a stack of arrows on import. Just in case someone needs to use
empty dispensers in their design. As far as I can remember, this won't
affect any of our dungeons - arrow traps are rare. The only dungeon that
I can remember is one by Balgor and it was exported with all of its
dispensers loaded.
XombyCraft has been kind enough to research which provider and biome IDs
are used by major mods. He found some ranges of free IDs and suggested
changing our defaults to sidestep conflicts with Biomes o' Plenty.
Fixed a nasty crash to desktop. It happens when ChickenChunks is
installed and a new world is generated with HardcoreLimboEnabled = true.
It appears that ChickenChunks forces a chunk to generate in Limbo if
LimboProvider.canRespawnHere() = true, which is the case if hardcore
Limbo is enabled. Our Monolith and gateway generation code runs as a
tick handler instead of through standard world gen calls. I believe
Limbo is unloaded immediately after the chunks are generated because no
players are around. That would cause DimensionManager to return null for
Limbo because it's not loaded, which would crash our code. We probably
dealt with this for Monoliths by adding a check. Now it happened again
because we didn't take precautions while calling the gateway generation
method. I've added code to forcefully load Limbo if it's not loaded.
Changed the condition on LimboProvider.canRespawnHere() so that players
can respawn in Limbo even if LimboEnabled is false. LimboEnabled only
controls whether players are sent to Limbo when they die in a pocket. It
does not prevent players from ending up in Limbo because of Monoliths.
If Hardcore Limbo is enabled, it stands to reason that people should be
respawning in Limbo anyway.
After finally getting the dungeons to load, I've determined that only
half of the rooms should be included. The three remaining rooms should
definitely stay. That's not much for a pack so I'm going to make some
modifications to the schematics and include them as part of Ruins.
* Fixed Limbo gateways. I accidentally broke them while overhauling
gateways in general.
* Changed references to BaseItemDoor.placeDoorBlock() to use
ItemDoor.placeDoorBlock() instead. We should refer to the original class
that implements the function.
* Set up the necessary code in DungeonHelper so that Balgor is
registered along with the other bundled packs.
* Improved the code for registering bundled packs to reduce the number
of paths we need to hardcode and to crash DD if a pack fails to load. A
crash would be inevitable no matter what since bundled packs are
integral to DD.
* Corrected an invalid generation rule for Balgor. It's set to select
random dungeons infinitely for now. I'll add an exit room later and
change the rule to force an exit. Balgor is still unusable until its
schematics get proper doors.
* Fixed PocketBuilder to actually check the results of validating
schematics before we try to build them. This was causing cryptic error
messages when flawed schematics were loaded (e.g. rooms without proper
doors) and could have caused serious problems during dungeon
regeneration. Don't ignore validation!
* Split off schematic-related functionality into BaseSchematicGateway.
* Moved the default implementations of many methods to the base classes
because having a bunch of duplicate stubs all over the place was a
waste.
* Removed methods that were redundant or weren't being used for
anything.
* Added support for importing schematics while ignoring air blocks
through IBlockSetter - now our gateways are copied in without air by
default.
* Fixed bugs that would have prevented the sandstone gateway from
generating.
* Removed the code in generate() that would cause dungeon
pre-generation. We should solve this by attaching data to links instead
and it's not a feature that we're even using right now (everything uses
the default pack).
* Fully documented our functions - it's so beautiful...
Changed dd-rift and dd-random so that rifts are only placed after a
dungeon is generated successfully. We also delete the link if generation
fails to clean up after ourselves. Also changed
PocketBuilder.generateSelectedDungeonPocket() so that its checks are
stricter and we validate dungeons before allocating a dimension.
This resolves one of our old issues: "Rearrange workflow in dd-rift to
prevent rifts from being created if no dungeon gets loaded and to
prevent dimension registration if the dimension cannot be populated"
Changed the code for dungeon selection in various classes so that rather
than allocating and passing around the dimension where the dungeon will
be generated, we instead pass around the parent dimension. This
simplifies our code and moves us toward avoiding stray dims when dungeon
selection fails and to solving the dungeon pre-generation problem with
gateways.
Added a new setting to dungeon pack configs called
"DuplicateSearchLevels", which allows us to configure how many levels up
of the dungeon tree should be checked to avoid duplicating rooms used in
that subtree. In other words, it lets us avoid repeating rooms used in
neighboring branches of the dungeon. The setting has been added but it's
not fully supported yet - some additional code is needed in
DungeonHelper and it's not trivial to implement. I took a break because
doing it wrong could break dungeon selection.
Added a setting to DDProperties for controlling the max distance that
players can be moved randomly when they're sent to Limbo. We previously
had a setting for the same except for leaving Limbo. Cleaned up some of
the related code a little. Added another setting for the chance of rifts
dropping World Thread on block destruction. Also fixed a potential NPE
in EventHookContainer that could arise theoretically arise if a
non-Vanilla door was attached to a rift.
