Faster trimws in R-CodePudding

I need convert all values that are "-" in my data frame to 0 however the "-" may have leading or trailing spaces so I'm using:

DF_To_Process[ trimws(DF_To_Process) == "-"] = 0

it seems to do the job but it's very slow plus I have repeat for several different instances such as "." and "N/A". Is there a faster way?

CodePudding user response：

Does your code work on the whole dataframe? Or only on one column at a time? If you're only looking at a single column, this approach may be a bit faster:

library(microbenchmark)
library(ggplot2)

# generate some 'test' data (700k rows, 1 col)
df <- data.frame(column = rep(c("fine", "   -   ", "  -", "-  ", ".", "N/A", "fine"), 1e5))

# Test current method:
df2 <- df
df2[trimws(df2) == "-",] = 0
head(df2, 9)
#>        column
#> 1        fine
#> 2        -   
#> 3           -
#> 4         -  
#> 5           .
#> 6         N/A
#> 7        fine
#> 8        fine
#> 9        -   

# This doesn't appear to have worked...


# try on one column at a time
trimws_func <- function(df) {
  df2 <- df
  df2$column[trimws(df2$column) == "-"] = 0
  df2$column[trimws(df2$column) == "."] = 0
  df2$column[trimws(df2$column) == "N/A"] = 0
  return(df2)
}

ifelse_func <- function(df) {
  df2 <- df
  df2$column <- ifelse(grepl("\\s*-\\s*|\\s*\\.\\s*|\\s*N/A\\s*",
                             df$column, perl = TRUE),
                       0, df2$column)
  return(df2)
}

head(trimws_func(df))
#>   column
#> 1   fine
#> 2      0
#> 3      0
#> 4      0
#> 5      0
#> 6      0

head(ifelse_func(df))
#>   column
#> 1   fine
#> 2      0
#> 3      0
#> 4      0
#> 5      0
#> 6      0

res <- microbenchmark(trimws_func(df), ifelse_func(df), times = 8)
res
#> Unit: milliseconds
#>             expr      min       lq     mean   median       uq      max neval
#>  trimws_func(df) 853.3805 878.1754 889.3538 887.9099 891.2959 946.6870     8
#>  ifelse_func(df) 296.7631 308.3129 328.5878 318.1445 350.8522 377.3202     8
#>  cld
#>    b
#>   a
autoplot(res)
#> Coordinate system already present. Adding new coordinate system, which will replace the existing one.

^{Created on 2022-10-06 by the reprex package (v2.0.1)}