麻烦结合双和字符

Question

我在合并数据集时遇到问题，我打算将许多数据集合并在一起，因此需要找出一种方法来自动解决以下错误：

"Error: Can't combine `C:/Users/gabri/AppData/Local/Cache/R/noaa_lcd/2006_72038163885.csv$HourlyWetBulbTemperature` <double> and `C:/Users/gabri/AppData/Local/Cache/R/noaa_lcd/2009_72038163885.csv$HourlyWetBulbTemperature` <character>."

我检查了数据，发现其中一个文件中的一些 NA 用 * 标记，所以我知道这就是问题所在。我想添加一个命令，将全部转换为字符或全部转换为数字，以便我可以合并，但是当我尝试添加 as.character 时，我收到此错误：

Error: Names repair functions can't return `NA` values

这是我尝试运行的相关代码，它会产生错误。

library(rnoaa)
library(tidyverse)
library(fs)

super_big_df <- map_df(my_files, read_csv, col_select = c(1,2,21,32,80), col_types = "cTddd", .id = "file")

这里是数据集相关列的 dput 输出

structure(list(STATION = c(72038163885, 72038163885, 72038163885
), DATE = structure(c(1230768000, 1230769200, 1230770400), tzone = "UTC", class = c("POSIXct", 
"POSIXt")), HourlyWetBulbTemperature = c("*", "38", "37"), DailyAverageWetBulbTemperature = c(NA, 
NA, NA), MonthlyWetBulb = c(NA, NA, NA)), row.names = c(NA, -3L
), class = c("tbl_df", "tbl", "data.frame"))

structure(list(STATION = c(72038163885, 72038163885, 72038163885
), DATE = structure(c(1146459600, 1146460800, 1146462000), tzone = "UTC", class = c("POSIXct", 
"POSIXt")), HourlyWetBulbTemperature = c(NA_real_, NA_real_, 
NA_real_), DailyAverageWetBulbTemperature = c(72, NA, NA), MonthlyWetBulb = c(NA, 
NA, NA)), row.names = c(NA, -3L), class = c("tbl_df", "tbl", 
"data.frame"))

总之，我想知道是否有一个命令可以放入 map_df() 中，它将所有内容都转换为相同的（字符或数字），以便命令的其余部分仍然运行。

Answer 1

未经测试，但正如@GregorThomas 建议的那样，最好的方法是第一次正确阅读它。在这种情况下，可能是这样的：

super_big_df <- map_df(
  my_files, read_csv, na = c("", "NA", "*"),
  col_select = c(1,2,21,32,80), col_types = "cTddd",
  .id = "file")

如果您需要事后修复它，那么您需要将它们读入list-of-frames，也许将map_df 更改为map，

super_big_df <- map(
  my_files, read_csv, na = c("", "NA", "*"),
  col_select = c(1,2,21,32,80), col_types = "cTddd",
  .id = "file")
bind_rows(super_big_df)
# Error: Can't combine `..1$HourlyWetBulbTemperature` <character> and `..2$HourlyWetBulbTemperature` <double>.

然后是类似的东西

library(dplyr) # in case you did not already have it loaded
purrr::map(super_big_df, ~ mutate(., HourlyWetBulbTemperature = suppressWarnings(as.numeric(HourlyWetBulbTemperature)))) %>%
  bind_rows()
# # A tibble: 6 x 5
#       STATION DATE                HourlyWetBulbTemperature DailyAverageWetBulbTemperature MonthlyWetBulb
#         <dbl> <dttm>                                 <dbl>                          <dbl> <lgl>         
# 1 72038163885 2009-01-01 00:00:00                       NA                             NA NA            
# 2 72038163885 2009-01-01 00:20:00                       38                             NA NA            
# 3 72038163885 2009-01-01 00:40:00                       37                             NA NA            
# 4 72038163885 2006-05-01 05:00:00                       NA                             72 NA            
# 5 72038163885 2006-05-01 05:20:00                       NA                             NA NA            
# 6 72038163885 2006-05-01 05:40:00                       NA                             NA NA            

这里的suppressWarnings 是因为我们知道在那一列的某处有一个非数字（"*"）。对于那一帧，它将修复该列；对于其他帧，它应该是空操作，因为该列已经是 as.numeric。

请注意，我在这里对名称进行了硬编码，因为我们提前知道它是什么。如果有更多列需要修复（即，修复此列后出现更多错误），那么采用更动态/程序化的方法可能是有利的（此处尚未介绍）。

---

数据

super_big_df <- list(
  structure(list(STATION = c(72038163885, 72038163885, 72038163885), DATE = structure(c(1230768000, 1230769200, 1230770400), tzone = "UTC", class = c("POSIXct", "POSIXt")), HourlyWetBulbTemperature = c("*", "38", "37"), DailyAverageWetBulbTemperature = c(NA, NA, NA), MonthlyWetBulb = c(NA, NA, NA)), row.names = c(NA, -3L), class = c("tbl_df", "tbl", "data.frame")),
  structure(list(STATION = c(72038163885, 72038163885, 72038163885), DATE = structure(c(1146459600, 1146460800, 1146462000), tzone = "UTC", class = c("POSIXct", "POSIXt")), HourlyWetBulbTemperature = c(NA_real_, NA_real_, NA_real_), DailyAverageWetBulbTemperature = c(72, NA, NA), MonthlyWetBulb = c(NA, NA, NA)), row.names = c(NA, -3L), class = c("tbl_df", "tbl", "data.frame"))
)