Remove unusful datasets after manipulation-CodePudding

Is there any straightforward way to remove data that I no longer need from the environment, instead of using the rm() function? The example here employs the join, but it can also be the basic merge() Here is a simple example with only 2 datasets, but I actually have much more than that.

library(tidyverse)
library(lubridate)

x <- lubridate::lakers %>%
  mutate(Month = lubridate::month(lubridate::as_date(date)))

y <- datasets::airquality 

z <- y %>% 
  dplyr::inner_join(x, by = "Month")

rm(x,y)

CodePudding user response：

It is possible to do this, but it is really bad practice to automatically delete data from the calling frame as a side effect of a function.

Please don't ever use this IRL - it's only to demonstrate that it can be done, not that it should be done.

First we start with an empty workspace:

ls()
#> character(0)

Now we define two data frames we are going to join in our function:

df1 <- data.frame(x = 1:5, y = 1:5)
df2 <- data.frame(x = 6:10, y = 6:10)
ls()
#> [1] "df1" "df2"

The following function removes the input data frames from the calling environment and returns the two joined in a simple rbind:

dangerous_function <- function(data1, data2) {
  data3 <- rbind(data1, data2)
  rm(list = c(deparse(substitute(data1)), 
       deparse(substitute(data2))), envir = parent.frame())
  data3
}

df3 <- dangerous_function(df1, df2)

And we can see that df1 and df2 have indeed been deleted.

ls()
#> [1] "dangerous_function" "df3"

You could even write the function so that it deletes itself after it has been used, which would probably be for the best.

CodePudding user response：

You can achieve this by avoiding creating objects. What I mean access dataframes within dplyr expression:

z <- datasets::airquality %>% 
  dplyr::inner_join(lubridate::lakers %>%
                    mutate(Month = lubridate::month(lubridate::as_date(date))), by = "Month")

or simplified version without specifying the packages in the code:

z <- airquality %>% inner_join(lakers %>%
                      mutate(Month = month(as_date(date))), by = "Month")

Note: the simplified version assumes you have loaded the packages prior, while longer version assumes you have not loaded packages and thus have to speicify with "dplyr::" in front of the function or dataset.