R6类中标准化只读主动绑定通用函数实现问题

问题原因 #

之前的实现报错核心是两点：

• 生成的绑定函数把传入的表达式存在默认参数里，默认参数的求值环境是工厂函数的内部环境，不是R6实例的运行环境，找不到实例中的private对象
• 没有把访问私有字段的逻辑直接注入到绑定函数的函数体中，运行时求值表达式时上下文已经丢失

方案1：无依赖Base R实现（最简便） #

R6会自动把active绑定函数的封闭环境设置为实例环境，函数运行时可以直接访问self和private，不需要复杂的表达式操作，只要传私有字段名即可：

# 只读主动绑定工厂
active_readonly <- function(private_field, warn_msg = "This field is read-only") {
  function(value) {
    if (missing(value)) {
      # 运行时直接从实例的private环境取对应字段
      private[[private_field]]
    } else {
      warning(warn_msg)
    }
  }
}

类定义时直接传入对应私有字段名即可：

randomObject <- R6::R6Class("randomObject",
  public = list(
    initialize = function(size){
      private$.size <- size
      private$.num <- sample(1:100,size)
      private$.letter <- sample(letters,size)
      private$.place <- sample(names(islands),size)
    },
    print = function(){
      cat("Random Object with ",private$.size," items:\n")
      cat("number(s): ",paste(private$.num, collapse = ", "),"\n")
      cat("letter(s): ",paste(private$.letter, collapse = ", "),"\n")
      cat("place(s): ",paste(private$.place, collapse = ", "),"\n")
    }
  ),
  private = list(
    .size = NULL, .num = NULL, .letter = NULL, .place = NULL
  ),
  active = list(
    size = active_readonly(".size", "Field size cannot be changed"),
    num = active_readonly(".num", "Field num cannot be changed"),
    letter = active_readonly(".letter", "Field letter cannot be changed"),
    place = active_readonly(".place", "Field place cannot be changed")
  )
)

测试运行完全符合预期：

test <- randomObject$new(5)
test$size
#> [1] 5
test$place <- "someplace"
#> Warning message:
#> In (function (value)  : Field place cannot be changed

如果字段数量特别多，还可以写个辅助函数批量生成active列表，避免重复书写：

make_readonly_actives <- function(field_names) {
  setNames(
    lapply(field_names, function(fname) {
      active_readonly(paste0(".", fname), paste("Field", fname, "cannot be changed"))
    }),
    field_names
  )
}

# 直接批量生成所有绑定
active = make_readonly_actives(c("size", "num", "letter", "place"))

方案2：rlang准引用实现（支持直接传表达式） #

如果你习惯直接传访问表达式而非字符串，可以用rlang的准引用把表达式直接注入到生成的函数体中，避免环境问题：

library(rlang)
active_access <- function(priv_expr, warn.msg = "Field cannot be modified"){
  # 捕获传入的访问表达式
  priv_expr <- enexpr(priv_expr)
  # 直接把表达式注入函数体，不依赖外部环境
  new_function(
    args = pairlist(value = ),
    body = expr({
      if(missing(value)){
        !!priv_expr
      }else{
        warning(!!warn.msg)
      }
    })
  )
}

使用方式和最初的写法几乎一致：

active = list(
  size = active_access(private$.size, "Field size cannot be changed"),
  num = active_access(private$.num, "Field num cannot be changed"),
  letter = active_access(private$.letter, "Field letter cannot be changed"),
  place = active_access(private$.place, "Field place cannot be changed")
)

内容的提问来源于stack exchange，提问作者Athe