R: Remove multiple empty columns of character variables

Question

I have a data frame where all the variables are of character type. Many of the columns are completely empty, i.e. only the variable headers are there, but no values. Is ther

Answer 1

A simple solution using the purrr package:

purrr::discard(my_data_frame, ~all(is.na(.)))

Answer 2

If you know the column indices, you can use

df[,-c(3, 5, 7)]

This will omit columns 3, 5, 7.

Answer 3

I have a similar situation -- I'm working with a large public records database but when I whittle it down to just the date range and category that I need, there are a ton of columns that aren't in use. Some are blank and some are NA.

The selected answer: https://stackoverflow.com/a/17672737/233467 didn't work for me, but this did:

df[!sapply(df, function (x) all(is.na(x) | x == ""))]

Answer 4

It depends what you mean by empty: Is it NA or "", or can it even be " "? Something like this might work:

df[,!apply(df, 2, function(x) all(gsub(" ", "", x)=="", na.rm=TRUE))]

Answer 5

You can do either of the following:

emptycols <- sapply(df, function (k) all(is.na(k)))
df <- df[!emptycols]

or:

emptycols <- colSums(is.na(df)) == nrow(df)
df <- df[!emptycols]

If by empty you mean they are "", the second approach can be adapted like so:

emptycols <- colSums(df == "") == nrow(df)

Answer 6

Here is something that can be modified to exclude columns containing any variables specied.

newdf= df[, apply(df, 2, function(x) !any({is.na(x) | x== "" | 
x== "-4"} ) )]