Paradox 的数据文件格式
Paradox 是我很喜欢的一个游戏公司,在所谓 P 社 5 萌中,十字军之王和钢铁雄心都只有浅尝,但在维多利亚和群星上均投入了大量时间和精力。
这些游戏基于同一套引擎,所以数据文件格式也是共通的。P 社开放了 Mod ,允许玩家来修改游戏,所以数据文件都是明文文本存放在文件系统中,这给了我们一个极好的学习机会:对于游戏从业者,我很有兴趣看看成熟引擎是如何管理游戏数据和游戏逻辑的。
据我所接触到的国内游戏公司,包括我们自己公司在内,游戏数据大都是基于 excel 这种二维表来表达的。我把它称为 csv 模式。这种模式的特点是,基础数据结构基于若干张二维表,每张表有不确定的行数,但每行有固定了列数。用它做基础数据结构的缺陷是很明显的,比如它很难表达树状层级结构。这往往就依赖做一个中间层,规范一些使用格式,在其上模拟出复杂数据结构。
另一种在软件行业广泛使用的基础数据结构是 json/xml 模式。json 比 xml 要简单。它的特点就是定义了两种基础的复合结构,字典和数组,允许结构嵌套。基于这种模式管理游戏数据的我也见过一些。不过对于策划来说,编辑树结构的数据终究不如 excel 拉表方便。查看起来也没有特别好的可视化工具,所以感觉用的人要少一些。
最开始,我以为 P 社的数据文件是偏向于后一种 json 模式。但实际研究下来又觉得有很大的不同。今天我尝试用 lpeg 写了一个简单的 parser 试图把它读进 lua vm ,写完 parser 后突然醒悟过来,其实它就是基于的嵌套 list ,不正是 lisp 吗?想明白这点后,有种醍醐灌顶的感觉,的确 lisp 模式要比 json 模式简洁的多,并不比 csv 模式复杂。但表达能力却强于它们两者,的确是一个更好的数据组织方案。
我们来看一个从群星中随便摘录的例子(有点长,但挺有代表性):
country_event = { id = primitive.16 hide_window = yes trigger = { is_country_type = primitive has_country_flag = early_space_age #NOT = { has_country_flag = recently_advanced } OR = { AND = { exists = from from = { OR = { is_country_type = default is_country_type = awakened_fallen_empire } } } years_passed > 25 } } mean_time_to_happen = { years = 100 modifier = { factor = 0.6 has_country_flag = acquired_tech } } immediate = { remove_country_flag = early_space_age set_country_flag = primitives_can_into_space set_country_type = default change_country_flag = random if = { limit = { is_species_class = MAM } set_graphical_culture = mammalian_01 } if = { limit = { is_species_class = REP } set_graphical_culture = reptilian_01 } if = { limit = { is_species_class = AVI } set_graphical_culture = avian_01 } if = { limit = { is_species_class = ART } set_graphical_culture = arthropoid_01 } if = { limit = { is_species_class = MOL } set_graphical_culture = molluscoid_01 } if = { limit = { is_species_class = FUN } set_graphical_culture = fungoid_01 } change_government = { authority = random civics = random } set_name = random if = { limit = { home_planet = { has_observation_outpost = yes } } home_planet = { observation_outpost_owner = { country_event = { id = primitive.17 } } } } add_minerals = 1000 # enough for a spaceport and then some add_energy = 500 add_influence = 300 capital_scope = { every_tile = { limit = { has_blocker = yes NOR = { has_blocker = tb_decrepit_dwellings has_blocker = tb_failing_infrastructure } } remove_blocker = yes } while = { limit = { num_pops < 8 any_tile = { has_grown_pop = no has_growing_pop = no has_blocker = no } } random_tile = { limit = { has_grown_pop = no has_growing_pop = no has_blocker = no } create_pop = { species = owner } } } random_tile = { limit = { has_grown_pop = yes OR = { has_building = "building_primitive_farm" has_building = "building_primitive_factory" has_building = no } } clear_deposits = yes add_deposit = d_mineral_food_deposit set_building = "building_capital_2" } random_tile = { limit = { has_grown_pop = yes OR = { has_building = "building_primitive_farm" has_building = "building_primitive_factory" has_building = no } } clear_deposits = yes add_deposit = d_mineral_deposit set_building = "building_mining_network_1" } random_tile = { limit = { has_grown_pop = yes OR = { has_building = "building_primitive_farm" has_building = "building_primitive_factory" has_building = no } } clear_deposits = yes add_deposit = d_mineral_deposit set_building = "building_mining_network_1" } random_tile = { limit = { has_grown_pop = yes OR = { has_building = "building_primitive_farm" has_building = "building_primitive_factory" has_building = no } } clear_deposits = yes add_deposit = d_farmland_deposit set_building = "building_hydroponics_farm_1" } random_tile = { limit = { has_grown_pop = yes OR = { has_building = "building_primitive_farm" has_building = "building_primitive_factory" has_building = no } } clear_deposits = yes add_deposit = d_farmland_deposit set_building = "building_hydroponics_farm_1" } random_tile = { limit = { has_grown_pop = yes OR = { has_building = "building_primitive_farm" has_building = "building_primitive_factory" has_building = no } } clear_deposits = yes add_deposit = d_energy_deposit set_building = "building_power_plant_1" } random_tile = { limit = { has_grown_pop = yes OR = { has_building = "building_primitive_farm" has_building = "building_primitive_factory" has_building = no } } clear_deposits = yes add_deposit = d_energy_deposit set_building = "building_power_plant_1" } random_tile = { limit = { has_grown_pop = yes OR = { has_building = "building_primitive_farm" has_building = "building_primitive_factory" has_building = no } } clear_deposits = yes add_deposit = d_energy_deposit set_building = "building_power_plant_1" } remove_all_armies = yes create_army = { name = random owner = PREV species = owner_main_species type = "defense_army" } create_army = { name = random owner = PREV species = owner_main_species type = "defense_army" } create_army = { name = random owner = PREV species = owner_main_species type = "defense_army" } create_army = { name = random owner = PREV species = owner_main_species type = "defense_army" } } random_owned_ship = { limit = { is_ship_size = primitive_space_station } fleet = { destroy_fleet = THIS } } } }
起初,我很疑惑在这个格式中,为啥赋值和相等都用的 = ,这不是容易引起歧义么?但是你从 lisp 的角度来看就简单了。等于号只是为了便于策划书写和阅读的一个变形。所谓 id = primitive.16 你可以理解为 ( id, primitive.16 ) 而 iscountrytype = default 一样可以理解为 ( iscountrytype , default ) 。 而
create_army = { name = random owner = PREV species = owner_main_species type = "defense_army" }
本质上是 ( create_army , ( ( name, random ) , (owner, PREV), (species, owner_main_species), (type, "defense_army") ) )
。
基础数据结构只要能表达出来,怎么理解这些 list 是更上层的工作,这就和我们在 csv 中去模拟树结构是一样的道理。只不过 years_passed > 25
这样的东西,被翻译成 ( years_passed, > , 25 )
有三个元素。上层解析的时候,如果确定它是一个逻辑表达式,就很容易在 2 个元素的 list 中间插入一个 = 补全。
这种结构很容易描述一些控制结构,比如上面例子中的 if 。我还在其它数据中发现了 repeat while 等控制结构,这些都是上层的工作,和底层数据模型无关。但不得不说,lisp 模式比 csv 模式更容易做此类控制结构。
把这种数据结构翻译成 lua 也很容易:只需要用 lua table 的 array 来保存即可。但为了使用方便,可以加一个代理结构。如果上层业务想把一个 list 解析成字典,就在 cache 中临时生成一个 hash 表加快查询即可。我们甚至可以把它直接存在 C 内存中,只在 lua 中暴露出遍历以及高层的访问方法。所谓高层的访问方法指,可以直接读取 if repeat 等控制结构,或是把带 AND OR 这样的复合 list 直接翻译成一个条件表达式。
8 月 8 日补充:
我实现了一个简单的 lua parser : https://github.com/cloudwu/pdxparser 。
Comments
Posted by: anAtheist | (15) May 15, 2024 12:36 AM
Posted by: anAtheist | (14) May 15, 2024 12:36 AM
Posted by: adzhang | (13) August 28, 2017 01:11 PM
Posted by: Ax72864 | (12) July 21, 2017 10:22 AM
Posted by: TeslaLee | (11) July 14, 2017 01:36 PM
Posted by: jjwei | (10) July 12, 2017 02:47 PM
Posted by: kelvin | (9) July 8, 2017 04:49 PM
Posted by: aduhupu | (8) July 7, 2017 04:56 PM
Posted by: Ian | (7) July 7, 2017 12:43 PM
Posted by: Cloud | (6) July 7, 2017 11:25 AM
Posted by: zbb | (5) July 7, 2017 10:56 AM
Posted by: Cloud | (4) July 7, 2017 10:48 AM
Posted by: pig09 | (3) July 7, 2017 10:19 AM
Posted by: pig09 | (2) July 7, 2017 10:19 AM
Posted by: dwing | (1) July 7, 2017 09:52 AM