--- title: formatR subtitle: Format R code automatically author: Yihui Xie date: "`r Sys.Date()`" show_toc: true slug: formatr githubEditURL: https://github.com/yihui/formatR/edit/master/vignettes/formatR.Rmd output: knitr:::html_vignette: toc: yes vignette: > %\VignetteEngine{knitr::rmarkdown} %\VignetteIndexEntry{An Introduction to formatR} --- ```{js, echo=FALSE} // redirect from CRAN to my personal website if (location.protocol === 'https:' && location.href.match('yihui.org') === null) location.href = 'https://yihui.org/formatr/'; ``` ```{r setup, include=FALSE} options(formatR.indent = 4, width = 70) knitr::opts_chunk$set(tidy = TRUE) ``` # 1. Installation You can install **formatR** from [CRAN](https://cran.r-project.org/package=formatR), or yihui.r-universe.dev if you want to test the latest development version: ```{r eval=FALSE} install.packages('formatR', repos = 'http://cran.rstudio.com') # or development version options(repos = c( yihui = 'https://yihui.r-universe.dev', CRAN = 'https://cloud.r-project.org' )) install.packages('formatR') ``` Or check out the [Github repository](https://github.com/yihui/formatR) and install from source if you know what this means. This page is always based on the development version. ```{r} library(formatR) sessionInfo() ``` # 2. Reformat R code The **formatR** package was designed to reformat R code to improve readability; the main workhorse is the function `tidy_source()`. Features include: - Long lines of code and comments are reorganized into appropriately shorter ones; - Spaces and indentation are added where necessary; - Comments are preserved in most cases; - The number of spaces to indent the code (i.e., tab width) can be specified (default is 4); - An `else` statement on a separate line without the leading `}` will be moved one line back; - `=` as an assignment operator can be substituted with `<-`; - `%>%` can be substituted with `|>`; - The left brace `{` can be moved to a new line; - Arguments of a function call can start on a new line after the function name when they cannot fit on one line; - Lines can be wrapped after pipes (both magrittr pipes such as `%>%` and R's native pipe `|>` are supported). Below is an example of what `tidy_source()` can do. The source code is: ```{r example, eval=FALSE, tidy=FALSE} ## comments are retained; # a comment block will be reflowed if it contains long comments; #' roxygen comments will not be wrapped in any case 1+1 if(TRUE){ x=1 # inline comments }else{ x=2;print('Oh no... ask the right bracket to go away!')} 1*3 # one space before this comment will become two! 2+2+2 # only 'single quotes' are allowed in comments lm(y~x1+x2, data=data.frame(y=rnorm(100),x1=rnorm(100),x2=rnorm(100))) ### a linear model 1+1+1+1+1+1+1+1+1+1+1+1+1+1+1+1+1+1+1+1+1 # comment after a long line ## here is a long long long long long long long long long long long long long comment that may be wrapped ``` We can copy the above code to clipboard, and type `tidy_source(width.cutoff = 50)` to get: ```{r example, eval=FALSE, tidy.opts=list(width.cutoff=50)} ``` Two applications of `tidy_source()`: - `tidy_dir()` can reformat all R scripts under a directory - `usage()` can reformat the usage of a function, e.g. compare `usage()` with the default output of `args()`: ```{r collapse=TRUE} library(formatR) usage(glm, width = 40) # can set arbitrary width here args(glm) ``` # 3. The Graphical User Interface If the **shiny** packages has been installed, the function `tidy_app()` can launch a Shiny app to reformat R code like this (live demo at `https://yihui.shinyapps.io/formatR/`): ``` r formatR::tidy_app() ``` ```{r echo=FALSE, results='asis'} if (ignore_img <- !is.na(Sys.getenv('_R_CHECK_PACKAGE_NAME_', NA))) cat('') ``` # 4. Evaluate the code and mask output in comments It is often a pain when trying to copy R code from other people's code which has been run in R and the prompt characters (usually `>`) are attached in the beginning of code, because we have to remove all the prompts `>` and `+` manually before we are able to run the code. However, it will be convenient for the reader to understand the code if the output of the code can be attached. This motivates the function `tidy_eval()`, which uses `tidy_source()` to reformat the source code, evaluates the code in chunks, and attaches the output of each chunk as comments which will not actually break the original source code. Here is an example: ```{r comment=NA} set.seed(123) tidy_eval(text = c("a<-1+1;a # print the value", "matrix(rnorm(10),5)")) ``` The default source of the code is from clipboard like `tidy_source()`, so we can copy our code to clipboard, and simply run this in R: ```{r eval=FALSE} library(formatR) tidy_eval() # without specifying any arguments, it reads code from clipboard ``` # 5. Showcase We continue the example code in Section 2, using different arguments in `tidy_source()` such as `arrow`, `blank`, `indent`, `brace.newline` and `comment`, etc. ## Substitute `=` with `<-` ```{r example, eval=FALSE, echo=5, tidy.opts=list(arrow=TRUE)} ``` ## Discard blank lines Note the 5th line (an empty line) was discarded: ```{r example, eval=FALSE, echo=1:5, tidy.opts=list(blank = FALSE)} ``` ## Reindent code (2 spaces instead of 4) ```{r example, eval=FALSE, echo=5, tidy.opts=list(indent = 2)} ``` ## Start function arguments on a new line With `args.newline = TRUE`, the example code below ```{r, args-code, eval=FALSE} shiny::updateSelectizeInput(session, "foo", label = "New Label", selected = c("A", "B"), choices = LETTERS, server = TRUE) ``` will be reformatted to: ```{r, args-code, eval=FALSE, tidy.opts=list(args.newline=TRUE)} ``` ## The pipe operators `%>%` and `|>` Since **formatR** 1.9, code lines contains operators `|>`, `%>%`, `%T%`, `%$%`, and/or `%<>%` will be automatically wrapped after these operators. For example, ```{r, pipe-code, eval=FALSE, tidy=FALSE} mtcars %>% subset(am == 0) %>% lm(mpg~hp, data=.) ``` will be reformatted to: ```{r, pipe-code, eval=FALSE, tidy=TRUE} ``` ## Move left braces `{` to new lines ```{r example, eval=FALSE, echo=5, tidy.opts=list(brace.newline = TRUE)} ``` ## Do not wrap comments ```{r example, eval=FALSE, echo=11:12, tidy.opts=list(wrap = FALSE)} ``` ## Discard comments ```{r example, eval=FALSE, tidy.opts=list(comment = FALSE, width.cutoff = 50)} ``` # 6. Further notes The tricks used in this packages are very dirty. There might be dangers in using the functions in **formatR**. Please read the next section carefully to know exactly how comments are preserved. The best strategy to avoid failure is to put comments in complete lines or after *complete* R expressions. Below are some known cases in which `tidy_source()` fails. ## Inline comments after an incomplete expression or ; ``` r 1 + 2 + ## comments after an incomplete line 3 + 4 x <- ## this is not a complete expression 5 x <- 1; # you should not use ; here! ``` Code with comments after incomplete R expression cannot be reformatted by **formatR**. By the way, `tidy_source()` will move comments after `{` to the next line, e.g., ```{r comment-brace, tidy=FALSE, eval=FALSE} if (TRUE) {## comments } ``` will become ```{r comment-brace, eval=FALSE} ``` ## Inappropriate blank lines Blank lines are often used to separate complete chunks of R code, and arbitrary blank lines may cause failures in `tidy_source()` as well when the argument `blank = TRUE`, e.g. ``` r if (TRUE) {'this is a BAD style of R programming!'} else 'failure!' ``` There should not be a blank line after the `if` statement. Of course `blank = FALSE` will not fail in this case. ## `?` with comments We can use the question mark (`?`) to view the help page, but **formatR** package is unable to correctly format the code using `?` with comments, e.g. ``` r ?sd # help on sd() ``` In this case, it is recommended to use the function `help()` instead of the short-hand version `?`. # 7. How does `tidy_source()` actually work? In a nutshell, `tidy_source(text = code)` is basically `deparse(parse(text = code))`, but actually it is more complicated only because of one thing: `deparse()` drops comments, e.g., ```{r} deparse(parse(text = '1+2-3*4/5 # a comment')) ``` The method to preserve comments is to protect them as strings in R expressions. For example, there is a single line of comments in the source code: ``` r # asdf ``` It will be first masked as ``` r invisible(".IDENTIFIER1 # asdf.IDENTIFIER2") ``` which is a legal R expression, so `base::parse()` can deal with it and will no longer remove the disguised comments. In the end the identifiers will be removed to restore the original comments, i.e. the strings `invisible(".IDENTIFIER1` and `.IDENTIFIER2")` are substituted with empty strings. Inline comments are handled differently: two spaces will be added before the hash symbol `#`, e.g. ``` r 1+1# comments ``` will become ``` r 1+1 # comments ``` Inline comments are first disguised as a weird operation with its preceding R code, which is essentially meaningless but syntactically correct! For example, ``` r 1+1 %\b% "# comments" ``` then `base::parse()` will deal with this expression; again, the disguised comments will not be removed. In the end, inline comments will be freed as well (remove the operator `%\b%` and surrounding double quotes). All these special treatments to comments are due to the fact that `base::parse()` and `base::deparse()` can tidy the R code at the price of dropping all the comments. # 8. Global options There are global options which can override some arguments in `tidy_source()`: | argument | global option | default | |-----------------|------------------------------------|--------------------| | `comment` | `options('formatR.comment')` | `TRUE` | | `blank` | `options('formatR.blank')` | `TRUE` | | `arrow` | `options('formatR.arrow')` | `FALSE` | | `pipe` | `options('formatR.pipe')` | `FALSE` | | `indent` | `options('formatR.indent')` | `4` | | `wrap` | `options('formatR.wrap')` | `TRUE` | | `width.cutoff` | `options('formatR.width')` | `options('width')` | | `brace.newline` | `options('formatR.brace.newline')` | `FALSE` | | `args.newline` | `options('formatR.args.newline')` | `FALSE` | Also note that single lines of long comments will be wrapped into shorter ones automatically when `wrap = TRUE`, but roxygen comments will not be wrapped (i.e., comments that begin with `#'`).