CODEWITHSUNDEEP
rfunctionfilterdplyr
Irakli
Irakli

Reputation: 1009

R: how to pass a list of filters to a function

I have a list of custom filters that I need to subset my dataframe by. For example, for mtcars I have this list: 

 filters=c(mpg>15, wt<2, carb>2 & am==0)

I want to pass this list to a function containing dplyr/pipes

fmean <- function(filter_x)   mtcars %>% filter(filter_x) %>% summarise(mean(disp))

My expected output after passing the list to the fmean is:

subset               mean(disp)
mpg>15                  192   
wt<2                    80.2
carb>2 & am==0          324

How to obtain the above output?

EDIT: found a tidyverse solution thanks to @alistaire, and others who replied here:

 library(tidyverse)
 filters <- c("mpg > 15", "wt < 2", "carb > 2 & am==0")

 fmean <- function(filter_x)   { mtcars %>%
                                   filter_(filter_x) %>%
                                   summarise(mean(disp)) %>% 
                                   mutate(subset=filter_x) %>%
                                   select(subset, everything()) }
 filters %>% map_df(fmean)

output:

 subset               mean(disp)
 mpg>15                  192.3   
 wt<2                    80.2
 carb>2 & am==0          324.5

Upvotes: 3

Views: 1827

Answers (3)

HubertL
HubertL

Reputation: 19544

In base R:

fmean <- function(filter_x)  data.frame(
  subset=filter_x, 
  do.call(rbind,
          lapply(filter_x, function(x) 
            mtcars %>% filter_(x) %>% summarise(mean(disp)))))

fmean(filters)
            subset mean.disp.
1         mpg > 15   192.3115
2           wt < 2    80.2250
3 carb > 2 & am==0   324.4600

Upvotes: 1

Chris
Chris

Reputation: 6372

data.table way, with eval(parse())

library(data.table)
mt_dt <- data.table(mtcars)

filters <- c("mpg > 15", "wt < 2", "carb > 2 & am==0")

out <- sapply(filters, function(x){mt_dt[eval(parse(text = x)), mean(disp)]})
out
# mpg > 15           wt < 2     carb > 2 & am==0 
# 192.3115          80.2250             324.4600

We are looping over our filters, and for each filter we subset and apply our aggregation function

This results in a named vector, which is quite flexible to work with. If you prefer a table, you can use:

data.table(subset = names(out), `mean(disp)` = out)
#              subset mean(disp)
# 1:         mpg > 15   192.3115
# 2:           wt < 2    80.2250
# 3: carb > 2 & am==0   324.4600

Upvotes: 1

Mark Timms
Mark Timms

Reputation: 606

The most straightforward way to accomplish this is probably to use the purrr package which, along with dplyr, is part of the tidyverse package:

library(tidyverse)

filters <- c("mpg > 15", "wt < 2", "carb > 2 & am==0")

fmean <- function(filter_x) {
  # Create list of means
  means <- filter_x %>% 
    map(~ mtcars %>% filter_(.dots = .x) %>% summarise(mean(disp)))

  # Create tibble from means
  tibble(subset = filter_x, means = unlist(means))
}   

fmean(filters)

Additionally, you want to use filter_ instead of filter which allows you to pass the subsetting conditions as strings rather than as unquoted text.

Upvotes: 4

Related Questions