{"id":1046,"date":"2023-01-09T19:47:06","date_gmt":"2023-01-09T18:47:06","guid":{"rendered":"https:\/\/lorentzen.ch\/?p=1046"},"modified":"2023-01-09T19:47:07","modified_gmt":"2023-01-09T18:47:07","slug":"dplyr-style-without-dplyr","status":"publish","type":"post","link":"https:\/\/lorentzen.ch\/index.php\/2023\/01\/09\/dplyr-style-without-dplyr\/","title":{"rendered":"Dplyr-style without dplyr"},"content":{"rendered":"\n
One of the reasons why we love the “dplyr” package: it plays so well together with the forward pipe operator `%>%` from the “magrittr” package. Actually, it is not a coincidence that both packages were released quite at the same time, in 2014.<\/p>\n\n\n\n
What does the pipe do? It puts the object on its left as the first argument into the function on its right: iris %>% head()<\/code> is a funny way of writing head(iris<\/code>). It helps to avoid long function chains like f(g(h(x)))<\/code>, or repeated assignments.<\/p>\n\n\n\n
In 2021 and version 4.1, R has received its native forward pipe operator |><\/code> so that we can write nice code like this:<\/p>\n\n\n\nImagine this without pipe…<\/figcaption><\/figure>\n\n\n\n
Since version 4.2, the piped object can be referenced by the underscore _<\/code>, but just once for now, see an example below.<\/p>\n\n\n\n
To use the native pipe via CTRL-SHIFT-M in Posit\/RStudio, tick this:<\/p>\n\n\n\n<\/figure>\n\n\n\n
Combined with the many great functions from the standard distribution of R, we can get a real “dplyr” feeling without even loading dplyr. Don’t get me wrong: I am a huge fan of the whole Tidyverse! But it is a great way to learn “Standard R”.<\/p>\n\n\n\n
Data chains<\/h3>\n\n\n\n
Here a small selection of standard functions playing well together with the pipe: They take a data frame and return a modified data frame:<\/p>\n\n\n\n
\n
subset()<\/code>: Select rows and columns of data frame<\/li>\n\n\n\n
transform()<\/code>: Add or overwrite columns in data frame<\/li>\n\n\n\n
library(ggplot2) # Need diamonds\n\n# What does the native pipe do?\nquote(diamonds |> head())\n\n# OUTPUT\n# head(diamonds)\n\n# Grouped statistics\ndiamonds |> \n aggregate(cbind(price, carat) ~ color, FUN = mean)\n\n# OUTPUT\n# color price carat\n# 1 D 3169.954 0.6577948\n# 2 E 3076.752 0.6578667\n# 3 F 3724.886 0.7365385\n# 4 G 3999.136 0.7711902\n# 5 H 4486.669 0.9117991\n# 6 I 5091.875 1.0269273\n# 7 J 5323.818 1.1621368\n\n# Join back grouped stats to relevant columns\ndiamonds |> \n subset(select = c(price, color, carat)) |> \n transform(price_per_color = ave(price, color)) |> \n head()\n\n# OUTPUT\n# price color carat price_per_color\n# 1 326 E 0.23 3076.752\n# 2 326 E 0.21 3076.752\n# 3 327 E 0.23 3076.752\n# 4 334 I 0.29 5091.875\n# 5 335 J 0.31 5323.818\n# 6 336 J 0.24 5323.818\n\n# Plot transformed values\ndiamonds |> \n transform(\n log_price = log(price),\n log_carat = log(carat)\n ) |> \n plot(log_price ~ log_carat, col = \"chartreuse4\", pch = \".\", data = _)<\/pre><\/div>\n\n<\/div><\/div>\n\t\t<\/div>\n\n\nA simple scatterplot<\/figcaption><\/figure>\n\n\n\n
The plot does not look quite as sexy as “ggplot2”, but its a start.<\/p>\n\n\n\n
Other chains<\/h3>\n\n\n\n
The pipe not only works perfectly with functions that modify a data frame. It also shines with many other functions often applied in a nested way. Here two examples:<\/p>\n\n\n
![\"\"](\"https:\/\/lorentzen.ch\/wp-content\/uploads\/2023\/01\/Base-R-and-piping.png\")
![\"\"](\"https:\/\/lorentzen.ch\/wp-content\/uploads\/2023\/01\/image.png\")
<\/figure>\n\n\n\n![\"\"](\"https:\/\/lorentzen.ch\/wp-content\/uploads\/2023\/01\/scatter.jpeg\")
![\"\"](\"https:\/\/lorentzen.ch\/wp-content\/uploads\/2023\/01\/barplot-2.jpeg\")
<\/figure>\n\n\n\n