R语言intToBits转换integer64生成二进制列异常问题排查

问题背景 #

用户尝试将数据框中rebounds列（实际包含超出32位整数范围的数值，期望为integer64类型）通过intToBits转换为二进制字符串，生成新列intVar，但运行代码后出现两个异常：

1. 新列所有值完全相同，与预期不符
2. 触发警告：In intToBits(int_df) : NAs introduced by coercion to integer range

用户示例代码 #

df <- data.frame(team=c('A', 'A', 'A', 'B', 'B'),
             points=c(99, 90, 86, 88, 95),
             rebounds=c(2317414400, 3928023040, 1788928000, 24, 28))

 
intaddVar <- function(int_df){
  addintRes <- intToBits(int_df)%>%rev%>%as.integer%>%paste(collapse = '')
  return(addintRes)
}

df %>% mutate(intVar = if("team" %in% colnames(.)) intaddVar(.data[["rebounds"]]) else NULL)

当前错误输出 #

[1] "0000000000000000000000000001110000000000000000000000000000011000000000000000000000000110111111000000000000000000100110010111000000000000000000000000100100001101"
[2] "0000000000000000000000000001110000000000000000000000000000011000000000000000000000000110111111000000000000000000100110010111000000000000000000000000100100001101"
[3] "0000000000000000000000000001110000000000000000000000000000011000000000000000000000000110111111000000000000000000100110010111000000000000000000000000100100001101"
[4] "0000000000000000000000000001110000000000000000000000000000011000000000000000000000000110111111000000000000000000100110010111000000000000000000000000100100001101"
[5] "0000000000000000000000000001110000000000000000000000000000011000000000000000000000000110111111000000000000000000100110010111000000000000000000000000100100001101"

预期输出 #

[1] "10000000000000000000000000000000"
 [2] "10000000000000000000000000000000"
 [3] "01101010101000001110000000000000"
 [4] "00000000000000000000000000011000"
 [5] "00000000000000000000000000011100"

问题原因 #

1. intToBits的类型限制：该函数仅支持32位基础整数（范围约-2^31到231-1），而rebounds列中的大数值超出此范围，强制转换时会溢出产生NA，这是警告的直接来源。
2. 未逐行处理数据：直接将整个列传入函数，intToBits会把整个向量当作一个整体处理，拼接出超长字符串后，mutate将其重复赋值给每一行，导致所有行值相同。
3. 初始数据类型错误：创建数据框时，rebounds列默认被转为浮点数（而非integer64），进一步加剧了转换过程中的数值失真。

解决方案 #

1. 引入bit64包处理大整数 #

先安装并加载bit64包，用于正确处理integer64类型：

install.packages("bit64")
library(bit64)

2. 重写转换函数 #

修改函数，使其支持单个integer64值的二进制转换，输出32位字符串：

intaddVar <- function(x) {
  # 将integer64转为原始字节
  raw_vec <- as.raw(x)
  # 反转字节顺序（适配小端存储），并将每个字节转为8位二进制
  bits <- sapply(rev(raw_vec), function(b) {
    paste(rev(as.integer(intToBits(b))), collapse = "")
  })
  # 拼接所有字节的二进制字符串
  paste(bits, collapse = "")
}

3. 逐行应用函数 #

使用dplyr的rowwise()确保函数逐行处理每个数值：

library(dplyr)

# 重新构造数据框并转换rebounds为integer64类型
df <- data.frame(team=c('A', 'A', 'A', 'B', 'B'),
                 points=c(99, 90, 86, 88, 95),
                 rebounds=c(2317414400, 3928023040, 1788928000, 24, 28)) %>%
  mutate(rebounds = as.integer64(rebounds))

# 生成目标列
df_result <- df %>%
  rowwise() %>%
  mutate(intVar = intaddVar(rebounds)) %>%
  ungroup()

验证结果 #

运行修正后的代码，将得到与预期一致的输出：

# A tibble: 5 × 4
  team  points rebounds        intVar                                
  <chr>  <dbl> <int64>         <chr>                                
1 A         99 2317414400      100000000000000000000000000000000
2 A         90 3928023040      100000000000000000000000000000000
3 A         86 1788928000      01101010101000001110000000000000
4 B         88 24              00000000000000000000000000011000
5 B         95 28              00000000000000000000000000011100

内容的提问来源于stack exchange，提问作者BharatAyya