用 R 替换 data.table 中的多个值

Question

I have tried multiple ways to replace two values in a data table with NA.

The data are here . There are two values 9223372036854775807 and 2147483647 which I intend to replace with NA

library(data.table)

data <- fread("https://raw.githubusercontent.com/Deborah-Jia/Complete_Analysis_da2/main/eg1.csv", integer64 = "numeric")

I tried:

data[data = 9223372036854775807|2147483647]

had error:

Error in [.data.table (data, , data = 9223372036854775808 | 2147483647, : unused argument (data = 9223372036854775808 | 2147483647)

I checked the structure of [i, j, by...] but couldn't find the cause. So, I use for loop instead:

# only these cols have 9223372036854775807 and 2147483647
special_col <- data %>% select(matches("price|size|room")) %>% colnames()

for ( icol in special_col) {
  data[icol == 9223372036854775807|2147483647, icol := NA] 
}

It didn't work as expected; I can still find 2147483647 in the data table.

I know I can use

data[total_room_count_high == 9223372036854775807|2147483647, total_room_count_high := NA] 

and replicate each column, but it is rather tiresome.

Before these methods, I also did across , filter_at and mapply combined with a function to process each column. But as long as I put col inside data[ ] , then data.table would think col is a column name rather than a variable representing all columns.

Answer 1

For comparison you should use == . You can use | as -

data <- read.csv2("https://raw.githubusercontent.com/Deborah-Jia/Complete_Analysis_da2/main/eg1.csv")
data[data == 2147483647 | data == 9223372036854775807] <- NA
data

Answer 2

an approach using set

values <- c(9223372036854775807, 2147483647)
for(col in names(data)) set(data, i = which(data[[col]] %in% values), j = col, value = NA_integer_)

Answer 3

Note that 9223372036854775807 is not "really" in your data but is the way that integer64 sometimes represent NAs.

You can just replace the integer columns.

data <- fread("https://raw.githubusercontent.com/Deborah-Jia/Complete_Analysis_da2/main/eg1.csv", integer64 = "numeric")
for (j in seq_along(data)) {
  vj <- .subset2(data, j)
  if (is.integer(vj)) {
    i <- which(vj == .Machine$integer.max)
    set(data, i = i, j = j, value = NA_integer_)
  }
}

Note that 2147483647 might be there to represent "positive infinity" using integers and so might be a better representation than NA . (For example, if you want to filter on all properties above a certain price, these properties will be erroneously filtered if you replace these values with NA ).