I don’t like cats much

I don’t love cats. I am not a member of @RCatLadies. So the fact that Tidyverse packages for dealing with factors and functional programming have cat-related names (forcats and purrr) does not endear them to me.

I knew there was a time when I would encounter an R problem that needed the power of for loops, so when I asked the Twittersphere whether there was an alternative (hopefully a tidyverse one) that would allow me to avoid loops a little longer, I was a bit disappointed to hear that it is the purrr package.

I think the time has come, I might not be able to avoid #forloops anymore. I want to run the same set of #dplyr operations (rename, select, filter etc) on a set of dataframes I have datapastaed into #rstats. Is there a way to say "do this" to files P001-P032 without using a loop?

— Dr Jenny Richmond (@JenRichmondPhD) September 4, 2018

Apparently the map() function in purrr is going to be my friend, but at the moment it is completely bewildering. I thought it might be helpful to paste all the links that the lovely rstats people on Twitter sent me so that I can keep them in one place.

If you want to learn purrr this is what the Twitter experts suggest.

Useful purrr related blog posts/resources

Alison Hill

Alison’s fave purrr learning resources are all from #rladies. I need to check out resources from

Jennifer Thompson https://github.com/jenniferthompson/RLadiesIntroToPurrr

Jenny Bryan https://jennybc.github.io/purrr-tutorial/

Charlotte Wickham https://github.com/cwickham/purrr-tutorial

also see: http://rstd.io/row-work

Jared Wilber

Jared Wilber suggests that Amber Thomas resources are good https://amber.rbind.io/blog/2018/03/26/purrr_iterations/

Tom Kelly

Tom Kelly pointed me towards the @swcarpentry resources

https://swcarpentry.github.io/r-novice-inflammation/03-loops-R/index.html https://swcarpentry.github.io/r-novice-inflammation/15-supp-loops-in-depth/

Miles McBain says…

You can use dplyr::bind_rows() instead of reduce(rbind()). BUT if you want them all in one frame at the end you probably just want purrr::map_dfr(), which is a map and bind combo function. So many options! that’s actually half the problem with going #noloops. My most commonly used fns in purrr are map(), pmap(), walk(), iwalk() and every() maybe that helps narrow it down a bit.

Hendrik vanB says…

This feels like an ideal purrr::map() use case. E.g., assuming .csv files:

purrr::map(filepaths, function(x) {
  readr::read_csv(x) %>%
    rename(...) %>%
    select(...)  %>%
    filter(...) %>%
    etc.
})

James Goldie says…

I use similar patterns a lot! You can:

1. map over the filename vector, immediately joining the dfs into 1 df using map_dfr and then doing your operations on the result

map_dfr(filepaths, read_csv) %>%
  select(...)

2. map over the filename vector, doing your operations on each one inside map and then joining the result

map_dfr(filepaths, function(x) {
  read_csv(x) %>%
  select(...)
})

3. split a dataframe up into a list of dfs and map over that

bigdataframe %>%
  split(list(bigdataframe$group1)) %>%
  map(function(df) {
    select(...)
  })

my own googling

A little Google searching turned out this post from Claus Wilke that outlines how to use map() to run read_csv for many files

https://serialmentor.com/blog/2016/6/13/reading-and-combining-many-tidy-data-files-in-R

Stay tuned for more purrr related posts…