feat: add minecraft-plugin-development skillFeat/minecraft plugin development (#1454)

* feat: add minecraft-plugin-development skill

* docs: expand minecraft plugin skill patterns

* docs: add minecraft skill examples

* docs: generalize minecraft skill patterns

* docs: expand minecraft progression guidance

* docs: add minecraft plugin validation workflow

---------

Co-authored-by: jiang <helloworld@jiang.cn>
This commit is contained in:
Zixuan Jiang
2026-04-28 09:41:41 +08:00
committed by GitHub
parent 7b9e8229fb
commit ca56e9577d
10 changed files with 1317 additions and 0 deletions

View File

@@ -0,0 +1,120 @@
# Config, Data, And Async Patterns
Use this reference when a task touches config loading, persistence, caches, scoreboards, or background refreshes.
## Config patterns from real plugins
### Multi-config minigame layer
Large minigames may use multiple config-oriented classes:
- `PluginConfigManager`
- `GameConfigurationManager`
- `KillMessagesConfig`
- `MessagesConfig`
- `LevelsConfig`
- `MapConfig`
- `PlayerConfig`
This pattern works well when gameplay systems each own a distinct config surface.
Tradeoff:
- easy to grow per-domain config
- harder to keep defaults and reload semantics consistent unless carefully managed
### Defensive wrapper config layer
Persistent brawl modes may use a defensive wrapper around the main config and separate loaders for other domain files such as progression and heroes.
Observed strengths:
- fallback values
- warning logs on invalid config
- one wrapper for core lobby and material settings
Guidance:
- if the project is still compact, prefer a typed wrapper with fallbacks
- split into many config managers only when domain complexity truly requires it
## Database and player data patterns
### Match-heavy minigame observations
- uses a database manager plus stats and brand-related storage
- flushes pending updates before disconnecting
### Persistent-brawl observations
- separates `DatabaseManager`, repository, and service
- keeps in-memory cache keyed by `UUID`
- tracks dirty players
- flushes asynchronously
- rebuilds leaderboards from cached data
- uses dirty sets for sidebar and rank UI so scoreboards are refreshed in batches
- snapshots leaderboard data before presenting it to gameplay or UI code
Guidance:
- use a repository or DAO boundary when persistence logic grows
- keep a cache for hot player stats
- record dirty entries instead of writing on every event
- flush on intervals and on shutdown
- debounce or batch leaderboard rebuilds when kills, deaths, or ranks can change frequently
- publish immutable snapshots or copies from async-maintained caches
## Async rules for Paper plugins
Observed safe persistent-brawl pattern:
- preload cache asynchronously
- when data is ready, switch to the main thread for Bukkit-side UI refreshes
- async leaderboard rebuilds avoid blocking gameplay
Recommended rules:
- async:
- SQL
- file-heavy processing
- leaderboard rebuilds
- cache recomputation
- main thread:
- teleport
- inventory changes
- entity or world mutation
- scoreboard and visible UI changes unless the API is explicitly thread-safe
## Background task patterns
Observed persistent-brawl patterns:
- async periodic leaderboard rebuild checks
- async dirty-player flush task
- throttled UI refresh strategy with dirty sets
- main-thread scoreboard flush from UUID dirty sets
- async cache preload followed by main-thread scoreboard, tab name, and leaderboard refresh
Observed match-heavy minigame patterns:
- main-thread periodic game-stage task
- per-session cooldown ticking
- time-sensitive display refreshes tied to game progression
Guidance:
- use async tasks for data maintenance
- use main-thread tasks for gameplay progression
- if a task can be event-driven plus dirty-flagged, prefer that over brute-force refreshing everything every tick
- copy mutable cached data before saving or sorting it asynchronously
- avoid starting async work from inside an async repeating task unless you need distinct lifecycle control
## Config contribution guidance
When generating config-aware code:
- provide defaults
- validate inputs
- document required keys
- avoid silently crashing on invalid material names or missing locations
- mention whether reload is safe for the specific subsystem