v0.42.0
CombatSolver 0.42.0
本日志汇总自 0.41.0 发布以来的改动。
简体中文
- 新增可选设置“预知战后药水奖励”,默认关闭。开启后,求解器会预估战后是否掉落药水;药水栏已满时,会把可能补回的药水计入用药取舍。无法确定的奖励会显示为未知,不会当作确定掉落。
- 路线摘要现在分开显示本场已经使用的药水、后续路线还要使用的药水,以及开启预测时的战后掉落结果。跨回合续用路线时,这些信息会随实际用药进度更新。
- 修复金斧、电流相生、扯碎、亡魂牵引、谋杀和超质量体依赖整场历史计数时,搜索可能把不同路线误判为相同状态的问题。本次修复针对搜索开始时已在牌组中的这些牌;战斗中途生成的同类牌,以及搜索开始时已在消耗堆、之后又被回收的牌,仍不在修复范围内。
- 减少相关战斗中反复生成卡牌时的重复目录查询,并避免为未选中的路线提前构造跨回合续用信息。具体耗时改善取决于牌组和战斗,不代表所有战斗都会同幅提速。
- 改善内存余量不足时的搜索处理;特别长的搜索会限制持续增长的历史状态记录。达到限制后仍继续搜索,但最终路线可能与旧版不同。此项尚未做广泛整场质量或可见游戏性能对照。
感谢 時穂(@s1f102500012) 贡献药水奖励预测 #109、代码维护 #111 / #112,以及整场历史状态修复 #113。感谢 洛天琉璃依(@ltlly) 贡献搜索性能与内存改进 #114;其中先行接入的改动另有整合记录 #115。
English
- Added an optional “Foresee post-combat potion reward” setting, off by default. When enabled, the solver forecasts whether a potion will drop after combat and accounts for a possible replacement when the potion belt is full. Rewards it cannot determine remain unknown rather than being treated as guaranteed drops.
- Route summaries now distinguish potions already used in this combat, potions still planned for the route, and the forecast reward when enabled. These details advance with actual potion use when a route continues into later turns.
- Fixed search states with different combat-wide histories being merged when Gold Axe, Voltaic, Tear Asunder, Pull From Below, Murder, or Supermassive depends on those counts. This fix covers these cards when they are in the deck zones at the start of the search. Cards generated later in combat, or already Exhausted at the search root and recovered later, remain outside its scope.
- Reduced repeated card-catalog work in relevant card-generation paths and delayed continuation data creation until a route is selected. The effect on search time depends on the deck and encounter; it is not a blanket speedup claim.
- Improved behavior under low memory headroom and limited the growth of retained search history in exceptionally long searches. Search continues after the limit is reached, but the selected route may differ from older versions. Broad full-combat quality and visible-game performance comparisons have not been completed for this limit.
Thanks to @s1f102500012 for potion reward forecasting #109, maintenance work #111 / #112, and the combat-history state fix #113. Thanks to @ltlly for search performance and memory work #114; an earlier integration of part of that work is also recorded in #115.