--- title: "Reusing documentation" description: > Tools for reusing documentation across topics, and between documentation and vignettes. output: rmarkdown::html_vignette vignette: > %\VignetteIndexEntry{Reusing documentation} %\VignetteEngine{knitr::rmarkdown} %\VignetteEncoding{UTF-8} --- ```{r, include = FALSE} knitr::opts_chunk$set( collapse = TRUE, comment = "#>" ) ``` roxygen2 provides several ways to avoid repeating yourself in code documentation, while assembling information from multiple places in one documentation file: - Combine documentation for closely related functions into a single file with `@describeIn` or `@rdname`. - Automatically copy tags with `@inheritParams`, `@inheritSection`, or `@inherit`. - Use functions to generate repeated text with inline R code. - Share text between documentation and vignettes with child documents. The chapter concludes by showing you how to update superseded reuse mechanisms that we no longer recommend: `@includeRmd`, `@eval`/`@evalRd`, and `@template`. ## Multiple functions in the same topic You can document multiple functions in the same file by using either `@rdname` or `@describeIn` tag. It's a technique best used with care: documenting too many functions in one place leads to confusion. Use it when all functions have the same (or very similar) arguments. ### `@rdname` Use `@rdname `[^1] to include multiple functions in the same page. Tags (e.g. `@title`, `@description`, `@examples`) will be combined, across blocks but often this yields text that is hard to understand. So we recommend that you make one block that contains the title, description, common parameters, examples and so on, and then only document individual parameters in the other blocks. [^1]: The destination is a topic name. There's a one-to-one correspondence between topic names and `.Rd` files where a topic called `foo` will produce a file called `man/foo.Rd`. There are two basic ways to do that. You can create a standalone documentation block and then add the functions to it. This typically works best when you have a family of related functions and you want to link to that family from other functions (i.e. `trig` in the examples below). ```{r} #' Trigonometric approximations #' @param x Input, in radians. #' @name trig NULL #' @rdname trig #' @export sin_ish <- function(x) x - x^3 / 6 #' @rdname trig #' @export cos_ish <- function(x) 1 - x^2 / 2 #' @rdname trig #' @export tan_ish <- function(x) x + x^3 / 3 ``` Alternatively, you can add docs to an existing function. This tends to work better if you have a "primary" function with some variants or some helpers. ```{r} #' Logarithms #' #' @param x A numeric vector #' @export log <- function(x, base) ... #' @rdname log #' @export log2 <- function(x) log(x, 2) #' @rdname log #' @export ln <- function(x) log(x, exp(1)) ``` ### `@describeIn` An alternative to `@rdname` is `@describeIn`. It has a slightly different syntax because as well as a topic name, you also provide a brief description for the function, like `@describein topic one sentence description`. The primary difference between `@rdname` and `@describeIn`, is that `@describeIn` creates a new section containing a bulleted list all of each function, along with its description. It uses a number of heuristics to determine the heading of this section, depending on you're documenting related functions, methods for a generic, or methods for a class. In general, I no longer recommend `@describeIn` because I don't think the heuristics it uses are as good as a thoughtful hand-crafted summary. If you're currently using `@describeIn`, you can general replace it with `@rdname`, as long as you give some thought to the multiple-function `@description`. ### Order of includes By default, roxygen blocks are processed in the order in which they appear in the file. When you're combining multiple files, this can sometimes cause the function usage to appear in a suboptimal order. You can override the default ordering with `@order`. For example, the following the block would place `times` first in `arith.Rd` because 1 comes before 2. ```{r} #' @rdname arith #' @order 2 add <- function(x, y) x + y #' @rdname arith #' @order 1 times <- function(x, y) x * y ``` ## Automatically copy tags If two or more functions share have similarities but are different or complex enough that you don't want to document them in a single file, you can use one of the four `@inherit` tags to automatically copy various components from another topic: - `@inheritParams foo` will copy `@param` contents from `foo`. - `@inherit foo` will copy all supported components from `foo`. - `@inheritSection foo {Section title}` will copy the `@section {Section title}` section from `foo`. - `@inheritDotParams foo` will generate documentation for `…` by copying the documentation for `foo()`'s arguments. We think of this as "inheritance" rather than just copying, because anything you inherit can be overridden by a more specific definition in the documentation. This applies particularly to `@inheritParams` which allows you to copy the documentation for some parameters while documenting others directly. We'll focus on this first. ### Parameters The oldest, and most frequently used, inherits tag is `@inheritParams`. It's particularly useful when multiple functions use the same argument name for the same task, as you can document the argument once, then inherit those docs elsewhere. For example, take the dplyr functions `arrange()`, `mutate()`, and `summarise()` which all have an argument called `.data`. `arrange()` is documented like so: ```{r} #' @param .data A data frame, data frame extension (e.g. a tibble), or a #' lazy data frame (e.g. from dbplyr or dtplyr). See *Methods*, below, for #' more details. #' @param ... <[`data-masking`][rlang::args_data_masking]> Variables, or #' functions of variables. Use [desc()] to sort a variable in descending #' order. arrange <- function(.data, ...) {} ``` Then `mutate()` and `summarise()` don't need to provide `@param .data` but can instead inherit the documentation from `arrange()`: ```{r} #' @inheritParams arrange mutate <- function(.data, ...) {} #' @inheritParams arrange summarise <- function(.data, ...) {} ``` If this was all you wrote it wouldn't be quite right because `mutate()` and `summarise()` would also inherit the documentation for `...`, which has a different interpretation in these functions. So, for example, `mutate()` provides its own definition for `...`: ```{r} #' @inheritParams arrange #' @param ... <[`data-masking`][rlang::args_data_masking]> Name-value pairs. #' The name gives the name of the column in the output. #' #' The value can be: #' #' * A vector of length 1, which will be recycled to the correct length. #' * A vector the same length as the current group (or the whole data frame #' if ungrouped). #' * `NULL`, to remove the column. #' * A data frame or tibble, to create multiple columns in the output. mutate <- function(.data, ...) {} ``` Note that only the documentation for arguments with the same names are inherited. For example, `arrange()` also has a `.by_group` argument. Since no other function in dplyr has an argument with this name, its documentation will never be inherited. ### Multiple parameters Sometimes you document two (or more) tightly coupled parameters together. For example, `dplyr::left_join()` has: ```{r} #' @param x,y A pair of data frames, data frame extensions (e.g. a tibble), or #' lazy data frames (e.g. from dbplyr or dtplyr). See *Methods*, below, for #' more details. ``` When joint parameter documentation is inherited, it's all or nothing, i.e. if a function has `@inheritParams left_join` it will only inherit the documentation for `x` and `y` if it has both `x` and `y` arguments and neither is documented by the inheriting function. ### The dot prefix Many tidyverse functions that accept named arguments in `...` also use a `.` prefix for their own arguments. This reduces the risk of an argument going to the wrong place. For example, `dplyr::mutate()` has `.by`, `.keep`, `.before`, and `.after` arguments, because if they didn't have that prefix, you wouldn't be able to create new variables called `by`, `keep`, `before`, or `after`. We call this pattern the [dot prefix](https://design.tidyverse.org/dots-prefix.html). This means that an argument with the same meaning can come in one of two forms: with and without the `.`. `@inheritParams` knows about this common pattern so ignores a `.` prefix when matching argument name. In other words, `.x` will inherit documentation for `x`, and `x` will inherit documentation from `.x`. ### Inheriting other components You can use `@inherits foo` to inherit the documentation for every supported tag from another topic. Currently, `@inherits` supports inheriting the following tags: `r paste0("\x60", roxygen2:::inherit_components, "\x60", collapse = ", ")`. By supplying a space separated list of components after the function name, you can also choose to inherit only selected components. For example, `@inherit foo returns` would just inherit the `@returns` tag, and `@inherit foo seealso source` would inherit the `@seealso` and `@source` tags. `@inherit foo sections` will inherit *every* `@section` tag (which can also be specified in markdown by using the top-level heading spec, `#`). To inherit a *specific* section from another function, use `@inheritSection foo Section title`. For example, all the "adverbs" in purrr use `#' @inheritSection safely Adverbs` to inherit a standard section that provides advice on using an adverb in package code. ### Documenting `...` When your function passes `...` on to another function, it can be useful to inline the documentation for some of the most common arguments. This technique is inspired by the documentation for `plot()`, where `...` can take any graphical parameter; `?plot` describes some of the most common arguments so that you don't have to look them up in `?par`. `@inheritDotParams` has two components: the function to inherit from and the arguments to inherit. Since you'll typically only want to document the most important arguments, `@inheritDotParams` comes with a flexible specification for argument selection inspired by `dplyr::select()`: - `@inheritDotParams foo` takes all parameters from `foo()`. - `@inheritDotParams foo a b e:h` takes parameters `a`, `b`, and all parameters between `e` and `h`. - `@inheritDotParams foo -x -y` takes all parameters except for `x` and `y`. ### Inheriting from other packages It's most common to inherit from other documentation topics within the current package, but roxygen2 also supports inheriting documentation from other packages by using `package::function` syntax, e.g.: - `@inheritParams package::function` - `@inherit package::function` - `@inheritSection package::function Section title` - `@inheritDotParams package::function` When inheriting documentation from another package bear in mind that you're now taking a fairly strong dependency on an external package, and to ensure every developer produces the same documentation you'll need to make sure that they all have the same version of the package installed. And if the package changes the name of the topic or section, your documentation will require an update. For those reasons, this technique is best used sparingly. ## Inline code To insert code inline, enclose it in `` `r ` ``. Roxygen will interpret the rest of the text within backticks as R code and evaluate it, and replace the backtick expression with its value. Here's a simple example: ```{r include = FALSE} roxygen2:::markdown_on() simple_inline <- "#' Title `r 1 + 1` #' #' Description `r 2 + 2` foo <- function() NULL " ``` ```{r code=simple_inline} ``` This is equivalent to writing: ```{r code = roxygen2:::markdown(simple_inline)} ``` The resulting text, together with the whole tag is interpreted as markdown, as usual. This means that you can use R to dynamically write markdown. For example if you defined this function in your package: ```{r} alphabet <- function(n) { paste0("`", letters[1:n], "`", collapse = ", ") } ``` You could then write: ```{r echo=FALSE} env <- new.env() env$alphabet <- alphabet roxygen2:::roxy_meta_set("evalenv", env) backtick <- "#' Title #' #' @param x A string. Must be one of `r alphabet(5)` foo <- function(x) NULL " ``` ```{r code = backtick} ``` The result is equivalent to writing the following by hand: ```{r code = roxygen2:::markdown_pass1(backtick)} ``` This is a powerful technique for reducing duplication because you can flexibly parameterise the function however best meets your needs. Note that the evaluation environment is deliberately a child of the package that you're documenting so you can call internal functions. ## Child documents You can use the same `.Rmd` or `.md` document in the documentation, `README.Rmd`, and vignettes by using child documents: ```` ```{r child = "common.Rmd"}`r ''` ``` ```` The included Rmd file can have roxygen Markdown-style links to other help topics. E.g. `[roxygen2::roxygenize()]` will link to the manual page of the `roxygenize` function in roxygen2. See `vignette("rd-formatting")` for details. If the Rmd file contains roxygen (Markdown-style) links to other help topics, then some care is needed, as those links will not work in Rmd files by default. A workaround is to specify external HTML links for them. These external locations will *not* be used for the manual which instead always links to the help topics in the manual. Example: ``` See also the [roxygen2::roxygenize()] function. [roxygen2::roxygenize()]: https://roxygen2.r-lib.org/reference/roxygenize.html ``` This example will link to the supplied URLs in HTML / Markdown files and it will link to the `roxygenize` help topic in the manual. Note that if you add external link targets like these, then roxygen will emit a warning about these link references being defined multiple times (once externally, and once to the help topic). This warning originates in Pandoc, and it is harmless. ## Superseded Over the years, we have experimented with a number of other ways to reduce duplication across documentation files. A number of these are now superseded and we recommend changing them to use the techniques described above: - Instead of `@includeRmd man/rmd/example.Rmd`, use a child document. - Instead of `@eval` or `@evalRd`, use inline R code. - Instead of `@template` and `@templateVars` write your own function and call it from inline R code. Inline R markdown can only generate markdown text within a tag so in principle it is less flexible than `@eval`/`@evalRd`/`@template`. However, our experience has revealed that generating multiple tags at once tends to be rather inflexible, and you often end up refactoring into smaller pieces so we don't believe this reflects a real loss of functionality.