将列名作为函数参数传递-R

Question

I am trying to find mean and median across categories "a" and "b" under y variable. I am trying to write a function to do this calculation. This is following sample dataset:

sample_data <- data.frame(x = 1:10, y = c("a","b"))
library(data.table)
sample_data_dt <- as.data.table(sample_data)

I have tried following methods but I am not able to find any elegant/simple method to pass column names as function parameters in both a data.table and a data.frame.

One working script for data.table sample_data_dt is:

apply_statistics_4 <- function(df, on_col, by_col){
df[, list(mean_value = mean(get(on_col)), median_value = median(get(on_col))), by = get(by_col)]}
apply_statistics_4(sample_data_dt, "x", "y") #works

However, similar script does not work for data.frame on ddply function:

apply_statistics_5 <- function(df, on_col, by_col){
ddply(df,.(get(by_col)), summarize, mean1 = mean(get(on_col)), median1 = median(get(on_col)))}
apply_statistics_5(sample_data, "x", "y") #Does not work
#  Error in get(by_col) : object 'y' not found 

One working script that I found for data.frame using ddply function is:

apply_statistics <- function(df, on_col, by_col){
df$y1 <- eval(substitute(by_col), df)
df$x1 <- eval(substitute(on_col), df)
ddply(df,.(y1), summarize, mean1 = mean(x1), median1 = median(x1))}
d <- apply_statistics(sample_data, x, y) #Works

If you know of any other method to use column names as function parameters in R for both a data.table and a data.frame, please do share with explanations.

Thanks.

Answer 1

You can reference the column names as follows:

sample_data[["y"]]
sample_data_dt[["y"]]

Another command that works similarly (although not identically) for both types is subset , eg

on_col <- "x"
subset(sample_data, select=get(on_col))
subset(sample_data_dt, select=get(on_col))
by_col <- "y"
subset(sample_data, subset=get(by_col)=="a")
subset(sample_data_dt, subset=get(by_col)=="a")

Note that the row numbers are output differently by data.table 's version of subset and the base R version, but otherwise they are pretty much interchangeable (although data.table is of course much faster).

Answer 2

It doesn't seem like it is a ddply problem but something related to the function environment. I had some test here, if you define the variables in the global environment, ddply can accept and get the result, but something curious happens when you pass the string as a variable to the function.

m <- "x"
n <- "y"
apply_statistics_5 <- function(df, m, n){
    ddply(df, n, summarise, mean1 = mean(get(m)), median1 = median(get(m)))
}
apply_statistics_5(sample_data, "x", "y")
  y mean1 median1
1 a     5       5
2 b     6       6

This will not work if m and n don't exist in the global environment.

Update : It might have something to do with the scoping issue of plyr package mentioned here .