[R] change col types of a df/tbl_df

[Duncan Murdoch]

On 10/12/2015 6:12 AM, arnaud gaboury wrote:
> Here is a sample of my data frame, obtained with read_csv2 from readr package.
>
> myDf <- structure(list(X15 = c("30.09.2015", "05.10.2015", "30.09.2015",
>
> "29.09.2015", "10.10.2015"), X16 = c("02.10.2015", "06.10.2015",
> "01.10.2015", "01.10.2015", "13.10.2015"), X17 = c("Grains",
> "Grains", "Grains", "Grains", "Grains"), X18 = c("Soyabeans",
> "Soyabeans", "Soyabeans", "Soyabeans", "Soyabeans"), X19 = c("20,000",
> "20,000", "20,000", "29,930", "26,000")), .Names = c("X15", "X16",
> "X17", "X18", "X19"), class = c("tbl_df", "data.frame"), row.names = c(NA,
> -5L))
>
> gabx at hortensia [R] str(myDf)
> Classes ‘tbl_df’ and 'data.frame': 5 obs. of  5 variables:
>   $ X15: chr  "30.09.2015" "05.10.2015" "30.09.2015" "29.09.2015" ...
>   $ X16: chr  "02.10.2015" "06.10.2015" "01.10.2015" "01.10.2015" ...
>   $ X17: chr  "Grains" "Grains" "Grains" "Grains" ...
>   $ X18: chr  "Soyabeans" "Soyabeans" "Soyabeans" "Soyabeans" ...
>   $ X19: chr  "20,000" "20,000" "20,000" "29,930" ...
>
> I want to change date to date class and numbers (X19) to numeric, and
> keep the class of my object.
>
> This code works:
>
> myDf$X19 <- as.numeric(gsub(",", "", myDf$X19))
> myDf$X15 <- as.Date(myDf$X15, format = "%d.%m.%Y"))
> myDf$X16 <- as.Date(myDf$X16, format = "%d.%m.%Y"))
>
> Now, as I have more than 5 columns, this can be fastidious and slowing
> code (?), even if I can group by type. Columns are only types of char,
> num and Date, so it could be OK.
>
> I tried with lapply for the Date columns. It works BUT will place NA
> in any columns with numbers as characters.
> The reuslt will be this for X19:  num NA NA NA NA NA NA NA NA NA NA ..
>
> How can I target my goal with something else than lapply or writing a
> line for each type ?

I don't see how a function could reliably detect the types, but it might 
be good enough to use a regular expression, possibly just on the first 
line of the result.  Once you've identified columns, e.g.

  numcols <- 19
  datecols <- c(15:16)

etc, you can use lapply:

myDf[,numcols] <- lapply(myDf[, numcools, drop=FALSE], function(x) 
as.numeric(gsub(",", "", x)))

You can simplify myDf[,numcols] to myDf[numcols] if you want, but I 
think it makes it less clear.

Duncan Murdoch