从具有重复值的数据框中删除列

Question

I have the following data frame containing characters and numbers, and NA:

df <- data.frame(a=c("notfound","NOT FOUND","NOT FOUND"), b=c(NA,"NOT FOUND","NOT FOUND"), c=c("not found",2,3), d=c("not   found","NOT FOUND","NOT FOUND"), e=c("234","NOT FOUND",NA))

 abcde 1 notfound <NA> not found not found 234 2 NOT FOUND NOT FOUND 2 NOT FOUND NOT FOUND 3 NOT FOUND NOT FOUND 3 NOT FOUND <NA>

I would like to remove all the columns where all the entries are "not found", "NOT found", "NOT FOUND" "notfound". basically if tolower(gsub(" ","",df)=="notfound") . It seems like this operation does not work on data frames. Are there any alternatives?

The desired output would be:

 de 1 not found 234 2 2 NOT FOUND 3 3 <NA>

Answer 1

You can use grepl with a regular expression to search for strings matching that expression and keep only those columns where some elements don't show a match (indicated by FALSE grepl output) so that the number of matches for that column is less than nrow(df) . This pattern matches strings that start with "not" and end with "found", and grepl is set to be case-insensitive.

is_nf <- 
  sapply(df, grepl, pattern = '(?=^not).*found$', 
         perl = TRUE, ignore.case = TRUE)


df[colSums(is_nf) < nrow(df)]
#           b         c         e
# 1      <NA> not found       234
# 2 NOT FOUND         2 NOT FOUND
# 3 NOT FOUND         3      <NA>

I'm guessing you'd also want to remove columns where the only non "not found" is NA.

is_na <- is.na(df)

df[colSums(is_nf | is_na) < nrow(df)]
#           c         e
# 1 not found       234
# 2         2 NOT FOUND
# 3         3      <NA>