get lhs object name when piping with dplyr

Question

I\'d like to have a function that can use pipe operator as exported from dplyr. I am not using magrittr.

df %>% my_function

How can I get df

Answer 1

Although the question is an old one, and the bounty has already been awarded, I would like to extend on gersht's excellent answer which works perfectly fine for getting the most lefthand-side object name. However, integrating this functionality in a dplyr workflow is not yet solved, apart from using this approach in the very last step of a pipe.

Since I'm using dplyr a lot, I have created a group of custom wrapper functions around the common dplyr verbs which I call metadplyr (I'm still playing around with the functionality, which is why I haven't uploaded it on github yet).

In essence, those functions create a new class called meta_tbl on top of a tibble and write certain things in the attributes of that object. Applied to the problem of the OP I provide a simple example with filter, but the procedure works on any other dplyr verb as well.

In my original function family I use slightly different names than dplyr, but the approach also works when 'overwriting' the original dplyr verbs.

Below is a new filter function which turns a data frame or tibble into a meta_tbl and writes the original name of the lhs object into the attribute .name. Here I am using a short version of gersht's approach.

library(dplyr)

 filter <- function(.data, ...) {

    if(!("meta_tbl" %in% class(.data))) {

      .data2 <- as_tibble(.data)

      # add new class 'meta_tbl' to data.frame  
      attr(.data2, "class") <- c(attr(.data2, "class"), "meta_tbl")

      # write lhs original name into attributes
      i <- 1
      while(!("chain_parts" %in% ls(envir=parent.frame(i)))) {
        i <- i+1
      }
      attr(.data2, ".name") <- deparse(parent.frame(i)$lhs)

    }

    dplyr::filter(.data2, ...)

}

For convenience it is good to have some helper function to extract the original name from the attributes easily.

.name <- function(.data) {
  if("meta_tbl" %in% class(.data)) {
  attr(.data, ".name")
  } else stop("this function only work on objects of class 'meta_tbl'")

}

Both functions can be used in a workflow in the following way:

mtcars %>% 
  filter(gear == 4) %>% 
  write.csv(paste0(.name(.), ".csv"))

This might be a bad example, since the pipe doesn't continue, but in theory, we could use this pipe including the original name and pipe it in further function calls.