R 转换列类/类型时维护日期/POSIXct列

我有一个400列的数据框，其中有多个日期列。 在下面的代表性示例中，我希望实现以下目标：

将因子转换为数字、字符或POSIXct

在合适的情况下，将字符转换为数字/整数

将包含日期的任何列转换为POSIXct，无论它是因子、字符还是日期

set.seed(123)
df1 <- data.frame(
A = as.numeric(1:10),
B = sample(seq(as.POSIXct('2000/01/01'), as.POSIXct('2018/01/01'), by="day"), size=10),
C = as.numeric(sample(20:90, size = 10)),
D = sample(c("yes", "no"), size=10, replace = TRUE),
E = as.factor(sample(1000:2000, size = 10)),
F = as.character(c("test","test2","test3","test4","test5","test6","test7","test8","test9","test10")),
G = as.factor(c("test","test2","test3","test4","test5","test6","test7","test8","test9","test10")),
H = as.character(sample(seq(as.POSIXct('2000/01/01'), as.POSIXct('2018/01/01'), by="day"), size=10)),stringsAsFactors=FALSE
)
df1
A                   B  C   D    E      F      G                   H
1   1 2005-03-06 00:00:00 87  no 1963   test   test 2002-07-27 23:00:00
2   2 2014-03-11 00:00:00 51  no 1902  test2  test2 2007-06-17 23:00:00
3   3 2007-05-11 23:00:00 66  no 1690  test3  test3 2007-06-11 23:00:00
4   4 2015-11-22 00:00:00 58  no 1793  test4  test4 2006-08-20 23:00:00
5   5 2016-12-02 00:00:00 26  no 1024  test5  test5 2002-09-27 23:00:00
6   6 2000-10-26 00:00:00 79  no 1475  test6  test6 2002-06-30 23:00:00
7   7 2009-06-30 23:00:00 35  no 1754  test7  test7 2004-03-11 00:00:00
8   8 2016-01-19 00:00:00 22  no 1215  test8  test8 2008-05-17 23:00:00
9   9 2009-11-30 00:00:00 40 yes 1315  test9  test9 2004-10-12 00:00:00
10 10 2008-03-17 00:00:00 85 yes 1229 test10 test10 2015-06-03 23:00:00

unlist(lapply(df1, class))
  A          B1          B2           C           D           E           F           G           H 
  "numeric"   "POSIXct"    "POSIXt"   "numeric" "character"    "factor" "character"    "factor" "character" 

set.seed（123）
df1这可能不是最优雅的方式，但对我来说似乎很管用

#install.packages("tidyverse")
#install.packages("dataCompareR")
library("tidyverse")
library("dataCompareR")



# create reproducible df
set.seed(123)
df1 <- data.frame(
  A = as.numeric(1:10),
  B = sample(seq(as.POSIXct('2000/01/01', tz = "UTC"), as.POSIXct('2018/01/01', tz = "UTC"), by="day"), size=10),
  C = as.numeric(sample(20:90, size = 10)),
  D = sample(c("yes", "no"), size=10, replace = TRUE),
  E = as.factor(sample(1000:2000, size = 10)),
  F = as.character(c("test","test2","test3","test4","test5","test6","test7","test8","test9","test10")),
  G = as.factor(c("test","test2","test3","test4","test5","test6","test7","test8","test9","test10")),
  H = as.character(sample(seq(as.POSIXct('2000/01/01', tz = "UTC"), as.POSIXct('2018/01/01', tz = "UTC"), by="day"), size=10)),stringsAsFactors=FALSE
)
df1 #look at df

unlist(lapply(df1, class)) #look at df classes


df1_clean <- df1 %>% mutate_all(funs(type.convert(as.character(.), as.is = TRUE))) #reassign classes by running type.convert (input are all variables from the df but as.character)
unlist(lapply(df1_clean, class)) #look at df classes now

#check if a column is a Date - https://stackoverflow.com/questions/18178451/is-there-a-way-to-check-if-a-column-is-a-date-in-r
tmp=sapply(df1_clean, function(x) !all(is.na(as.Date(as.character(x),format="%Y-%m-%d", tz = "UTC")))) 

# if tmp is True, change according column to as.POSIXct 
for (i in 1:ncol(df1_clean)){
  if (tmp[i] == T){
    df1_clean[,i]<- as.POSIXct(df1_clean[,i], tz = "UTC")
  }
}

df1_clean #look at df
unlist(lapply(df1_clean, class)) #look at df classes


comp <- rCompare(df1, df1_clean) #compare your dfs before and after using the dataCompareR package
summary(comp) # check summary

#安装程序包（“tidyverse”）
#install.packages（“dataCompareR”）
图书馆（“tidyverse”）
库（“数据比较器”）
#创建可复制的df
种子集（123）
df1我不明白：您显式地将所有列转换为
字符
，然后当它们丢失其
POSIXct
-属性时，您会感到惊讶吗？在R（不考虑自动协同）中，类通常是互斥的：
整数不能是
字符
，而
字符
不能是
POSIXct
，因此当你说
as.character（.）
，你是自愿且明确地放弃它的日期时间属性，将它转换成字符串。您是否试图在不预先知道哪个是哪个的情况下动态执行此操作，或者您是否可以使用
mutate_at（vars（a，C，…），funs（as.character（.））
？谢谢！我并不奇怪所有的东西都变成了字符类型。convert
只是我遇到的唯一一种在R中动态重新分配类的方法。我不知道如何将其编写为代码，但一种可行的方法是：I）将字符日期（例如H列）改为
POSIXct
，ii）创建一个非
POSIXct
的变量列表，最后iii）使用
类型转换所有非
POSIXct
的变量。请注意，
type.convert
的帮助中没有提到
POSIXct
，因此如果您使用该函数，您只能依靠自己。你提前知道哪些栏目应该是哪些栏目吗？或者，您是否正在尝试提出一个通用函数，该函数将愉快/巧妙地为您实现这一点，而不管您如何使用它？如果您知道type.convert，我很乐意使用一个更好的替代品。是的，我想要的是一个通用函数，不管我怎么做，它都会分配正确的类。执行“自动类型确定”的函数通常是为了将数据读入R，而不是处理已经在R环境中的对象。如果这是来自CSV或类似文件，我是否可以建议使用
readr
软件包？它首先为您检查前1000行（可配置）和自动类型。
#install.packages("tidyverse")
#install.packages("dataCompareR")
library("tidyverse")
library("dataCompareR")



# create reproducible df
set.seed(123)
df1 <- data.frame(
  A = as.numeric(1:10),
  B = sample(seq(as.POSIXct('2000/01/01', tz = "UTC"), as.POSIXct('2018/01/01', tz = "UTC"), by="day"), size=10),
  C = as.numeric(sample(20:90, size = 10)),
  D = sample(c("yes", "no"), size=10, replace = TRUE),
  E = as.factor(sample(1000:2000, size = 10)),
  F = as.character(c("test","test2","test3","test4","test5","test6","test7","test8","test9","test10")),
  G = as.factor(c("test","test2","test3","test4","test5","test6","test7","test8","test9","test10")),
  H = as.character(sample(seq(as.POSIXct('2000/01/01', tz = "UTC"), as.POSIXct('2018/01/01', tz = "UTC"), by="day"), size=10)),stringsAsFactors=FALSE
)
df1 #look at df

unlist(lapply(df1, class)) #look at df classes


df1_clean <- df1 %>% mutate_all(funs(type.convert(as.character(.), as.is = TRUE))) #reassign classes by running type.convert (input are all variables from the df but as.character)
unlist(lapply(df1_clean, class)) #look at df classes now

#check if a column is a Date - https://stackoverflow.com/questions/18178451/is-there-a-way-to-check-if-a-column-is-a-date-in-r
tmp=sapply(df1_clean, function(x) !all(is.na(as.Date(as.character(x),format="%Y-%m-%d", tz = "UTC")))) 

# if tmp is True, change according column to as.POSIXct 
for (i in 1:ncol(df1_clean)){
  if (tmp[i] == T){
    df1_clean[,i]<- as.POSIXct(df1_clean[,i], tz = "UTC")
  }
}

df1_clean #look at df
unlist(lapply(df1_clean, class)) #look at df classes


comp <- rCompare(df1, df1_clean) #compare your dfs before and after using the dataCompareR package
summary(comp) # check summary