Documentation tests

rustdoc supports executing your documentation examples as tests. This makes sure that examples within your documentation are up to date and working.

The basic idea is this:

#![allow(unused)]
fn main() {
/// # Examples
///
/// ```
/// let x = 5;
/// ```
fn f() {}
}

The triple backticks start and end code blocks. If this were in a file named foo.rs, running rustdoc --test foo.rs will extract this example, and then run it as a test.

Please note that by default, if no language is set for the block code, rustdoc assumes it is Rust code. So the following:

```rust
let x = 5;
```

is strictly equivalent to:

```
let x = 5;
```

There's some subtlety though! Read on for more details.

Passing or failing a doctest

Like regular unit tests, regular doctests are considered to "pass" if they compile and run without panicking. So if you want to demonstrate that some computation gives a certain result, the assert! family of macros works the same as other Rust code:

#![allow(unused)]
fn main() {
let foo = "foo";
assert_eq!(foo, "foo");
}

This way, if the computation ever returns something different, the code panics and the doctest fails.

Pre-processing examples

In the example above, you'll note something strange: there's no main function! Forcing you to write main for every example, no matter how small, adds friction and clutters the output. So rustdoc processes your examples slightly before running them. Here's the full algorithm rustdoc uses to preprocess examples:

Some common allow attributes are inserted, including unused_variables, unused_assignments, unused_mut, unused_attributes, and dead_code. Small examples often trigger these lints.
Any attributes specified with #![doc(test(attr(...)))] are added.
Any leading #![foo] attributes are left intact as crate attributes.
If the example does not contain extern crate, and #![doc(test(no_crate_inject))] was not specified, then extern crate <mycrate>; is inserted (note the lack of #[macro_use]).
Finally, if the example does not contain fn main, the remainder of the text is wrapped in fn main() { your_code }.

For more about that caveat in rule 4, see "Documenting Macros" below.

Hiding portions of the example

Sometimes, you need some setup code, or other things that would distract from your example, but are important to make the tests work. Consider an example block that looks like this:

#![allow(unused)]
fn main() {
/// ```
/// /// Some documentation.
/// # fn foo() {} // this function will be hidden
/// println!("Hello, World!");
/// ```
fn f() {}
}

It will render like this:

#![allow(unused)]
fn main() {
/// Some documentation.
fn foo() {}
println!("Hello, World!");
}

Yes, that's right: you can add lines that start with # , and they will be hidden from the output, but will be used when compiling your code. You can use this to your advantage. In this case, documentation comments need to apply to some kind of function, so if I want to show you just a documentation comment, I need to add a little function definition below it. At the same time, it's only there to satisfy the compiler, so hiding it makes the example more clear. You can use this technique to explain longer examples in detail, while still preserving the testability of your documentation.

For example, imagine that we wanted to document this code:

#![allow(unused)]
fn main() {
let x = 5;
let y = 6;
println!("{}", x + y);
}

We might want the documentation to end up looking like this:

First, we set x to five:

#![allow(unused)]
fn main() {
let x = 5;
let y = 6;
println!("{}", x + y);
}

Next, we set y to six:

#![allow(unused)]
fn main() {
let x = 5;
let y = 6;
println!("{}", x + y);
}

Finally, we print the sum of x and y:

#![allow(unused)]
fn main() {
let x = 5;
let y = 6;
println!("{}", x + y);
}

To keep each code block testable, we want the whole program in each block, but we don't want the reader to see every line every time. Here's what we put in our source code:

First, we set `x` to five:

```
let x = 5;
# let y = 6;
# println!("{}", x + y);
```

Next, we set `y` to six:

```
# let x = 5;
let y = 6;
# println!("{}", x + y);
```

Finally, we print the sum of `x` and `y`:

```
# let x = 5;
# let y = 6;
println!("{}", x + y);
```

By repeating all parts of the example, you can ensure that your example still compiles, while only showing the parts that are relevant to that part of your explanation.

The #-hiding of lines can be prevented by using two consecutive hashes ##. This only needs to be done with the first # which would've otherwise caused hiding. If we have a string literal like the following, which has a line that starts with a #:

#![allow(unused)]
fn main() {
let s = "foo
# bar # baz";
}

We can document it by escaping the initial #:

/// let s = "foo
/// ## bar # baz";

Using `?` in doc tests

When writing an example, it is rarely useful to include a complete error handling, as it would add significant amounts of boilerplate code. Instead, you may want the following:

#![allow(unused)]
fn main() {
/// ```
/// use std::io;
/// let mut input = String::new();
/// io::stdin().read_line(&mut input)?;
/// ```
fn f() {}
}

The problem is that ? returns a Result<T, E> and test functions don't return anything, so this will give a mismatched types error.

You can get around this limitation by manually adding a main that returns Result<T, E>, because Result<T, E> implements the Termination trait:

/// A doc test using ?
///
/// ```
/// use std::io;
///
/// fn main() -> io::Result<()> {
///     let mut input = String::new();
///     io::stdin().read_line(&mut input)?;
///     Ok(())
/// }
/// ```
fn f() {}

Together with the # from the section above, you arrive at a solution that appears to the reader as the initial idea but works with doc tests:

/// ```
/// use std::io;
/// # fn main() -> io::Result<()> {
/// let mut input = String::new();
/// io::stdin().read_line(&mut input)?;
/// # Ok(())
/// # }
/// ```
fn f() {}

As of version 1.34.0, one can also omit the fn main(), but you will have to disambiguate the error type:

#![allow(unused)]
fn main() {
/// ```
/// use std::io;
/// let mut input = String::new();
/// io::stdin().read_line(&mut input)?;
/// # Ok::<(), io::Error>(())
/// ```
fn f() {}
}

This is an unfortunate consequence of the ? operator adding an implicit conversion, so type inference fails because the type is not unique. Please note that you must write the (()) in one sequence without intermediate whitespace so that rustdoc understands you want an implicit Result-returning function.

Showing warnings in doctests

You can show warnings in doctests by running rustdoc --test --test-args=--show-output (or, if you're using cargo, cargo test --doc -- --show-output). By default, this will still hide unused warnings, since so many examples use private functions; you can add #![warn(unused)] to the top of your example if you want to see unused variables or dead code warnings. You can also use #![doc(test(attr(warn(unused))))] in the crate root to enable warnings globally.

Documenting macros

Here’s an example of documenting a macro:

/// Panic with a given message unless an expression evaluates to true.
///
/// # Examples
///
/// ```
/// # #[macro_use] extern crate foo;
/// # fn main() {
/// panic_unless!(1 + 1 == 2, “Math is broken.”);
/// # }
/// ```
///
/// ```should_panic
/// # #[macro_use] extern crate foo;
/// # fn main() {
/// panic_unless!(true == false, “I’m broken.”);
/// # }
/// ```
#[macro_export]
macro_rules! panic_unless {
    ($condition:expr, $($rest:expr),+) => ({ if ! $condition { panic!($($rest),+); } });
}
fn main() {}

You’ll note three things: we need to add our own extern crate line, so that we can add the #[macro_use] attribute. Second, we’ll need to add our own main() as well (for reasons discussed above). Finally, a judicious use of # to comment out those two things, so they don’t show up in the output.

Attributes

Code blocks can be annotated with attributes that help rustdoc do the right thing when testing your code:

The ignore attribute tells Rust to ignore your code. This is almost never what you want as it's the most generic. Instead, consider annotating it with text if it's not code or using #s to get a working example that only shows the part you care about.

#![allow(unused)]
fn main() {
/// ```ignore
/// fn foo() {
/// ```
fn foo() {}
}

should_panic tells rustdoc that the code should compile correctly but panic during execution. If the code doesn't panic, the test will fail.

#![allow(unused)]
fn main() {
/// ```should_panic
/// assert!(false);
/// ```
fn foo() {}
}

The no_run attribute will compile your code but not run it. This is important for examples such as "Here's how to retrieve a web page," which you would want to ensure compiles, but might be run in a test environment that has no network access. This attribute can also be used to demonstrate code snippets that can cause Undefined Behavior.

#![allow(unused)]
fn main() {
/// ```no_run
/// loop {
///     println!("Hello, world");
/// }
/// ```
fn foo() {}
}

compile_fail tells rustdoc that the compilation should fail. If it compiles, then the test will fail. However, please note that code failing with the current Rust release may work in a future release, as new features are added.

#![allow(unused)]
fn main() {
/// ```compile_fail
/// let x = 5;
/// x += 2; // shouldn't compile!
/// ```
fn foo() {}
}

edition2015, edition2018 and edition2021 tell rustdoc that the code sample should be compiled using the respective edition of Rust.

#![allow(unused)]
fn main() {
/// Only runs on the 2018 edition.
///
/// ```edition2018
/// let result: Result<i32, ParseIntError> = try {
///     "1".parse::<i32>()?
///         + "2".parse::<i32>()?
///         + "3".parse::<i32>()?
/// };
/// ```
fn foo() {}
}

Starting in the 2024 edition¹, compatible doctests are merged as one before being run. We combine doctests for performance reasons: the slowest part of doctests is to compile them. Merging all of them into one file and compiling this new file, then running the doctests is much faster. Whether doctests are merged or not, they are run in their own process.

An example of time spent when running doctests:

sysinfo crate:

wall-time duration: 4.59s
total compile time: 27.067s
total runtime: 3.969s

Rust core library:

wall-time duration: 102s
total compile time: 775.204s
total runtime: 15.487s

This is based on the edition of the whole crate, not the edition of the individual test case that may be specified in its code attribute.

In some cases, doctests cannot be merged. For example, if you have:

#![allow(unused)]
fn main() {
//! ```
//! let location = std::panic::Location::caller();
//! assert_eq!(location.line(), 4);
//! ```
}

The problem with this code is that, if you change any other doctests, it'll likely break when runing rustdoc --test, making it tricky to maintain.

This is where the standalone attribute comes in: it tells rustdoc that a doctest should not be merged with the others. So the previous code should use it:

#![allow(unused)]
fn main() {
//! ```standalone
//! let location = std::panic::Location::caller();
//! assert_eq!(location.line(), 4);
//! ```
}

In this case, it means that the line information will not change if you add/remove other doctests.

Custom CSS classes for code blocks

#![allow(unused)]
fn main() {
/// ```custom,{class=language-c}
/// int main(void) { return 0; }
/// ```
pub struct Bar;
}

The text int main(void) { return 0; } is rendered without highlighting in a code block with the class language-c. This can be used to highlight other languages through JavaScript libraries for example.

Without the custom attribute, it would be generated as a Rust code example with an additional language-C CSS class. Therefore, if you specifically don't want it to be a Rust code example, don't forget to add the custom attribute.

To be noted that you can replace class= with . to achieve the same result:

#![allow(unused)]
fn main() {
/// ```custom,{.language-c}
/// int main(void) { return 0; }
/// ```
pub struct Bar;
}

To be noted, rust and .rust/class=rust have different effects: rust indicates that this is a Rust code block whereas the two others add a "rust" CSS class on the code block.

You can also use double quotes:

#![allow(unused)]
fn main() {
/// ```"not rust" {."hello everyone"}
/// int main(void) { return 0; }
/// ```
pub struct Bar;
}

Syntax reference

The exact syntax for code blocks, including the edge cases, can be found in the Fenced Code Blocks section of the CommonMark specification.

Rustdoc also accepts indented code blocks as an alternative to fenced code blocks: instead of surrounding your code with three backticks, you can indent each line by four or more spaces.

    let foo = "foo";
    assert_eq!(foo, "foo");

These, too, are documented in the CommonMark specification, in the Indented Code Blocks section.

However, it's preferable to use fenced code blocks over indented code blocks. Not only are fenced code blocks considered more idiomatic for Rust code, but there is no way to use attributes such as ignore or should_panic with indented code blocks.

Include items only when collecting doctests

Rustdoc's documentation tests can do some things that regular unit tests can't, so it can sometimes be useful to extend your doctests with samples that wouldn't otherwise need to be in documentation. To this end, Rustdoc allows you to have certain items only appear when it's collecting doctests, so you can utilize doctest functionality without forcing the test to appear in docs, or to find an arbitrary private item to include it on.

When compiling a crate for use in doctests (with --test option), rustdoc will set #[cfg(doctest)]. Note that they will still link against only the public items of your crate; if you need to test private items, you need to write a unit test.

In this example, we're adding doctests that we know won't compile, to verify that our struct can only take in valid data:

#![allow(unused)]
fn main() {
/// We have a struct here. Remember it doesn't accept negative numbers!
pub struct MyStruct(pub usize);

/// ```compile_fail
/// let x = my_crate::MyStruct(-5);
/// ```
#[cfg(doctest)]
pub struct MyStructOnlyTakesUsize;
}

Note that the struct MyStructOnlyTakesUsize here isn't actually part of your public crate API. The use of #[cfg(doctest)] makes sure that this struct only exists while rustdoc is collecting doctests. This means that its doctest is executed when --test is passed to rustdoc, but is hidden from the public documentation.

Another possible use of #[cfg(doctest)] is to test doctests that are included in your README file without including it in your main documentation. For example, you could write this into your lib.rs to test your README as part of your doctests:

#![allow(unused)]
fn main() {
#[doc = include_str!("../README.md")]
#[cfg(doctest)]
pub struct ReadmeDoctests;
}

This will include your README as documentation on the hidden struct ReadmeDoctests, which will then be tested alongside the rest of your doctests.

Controlling the compilation and run directories

By default, rustdoc --test will compile and run documentation test examples from the same working directory. The compilation directory is being used for compiler diagnostics, the file!() macro and the output of rustdoc test runner itself, whereas the run directory has an influence on file-system operations within documentation test examples, such as std::fs::read_to_string.

The --test-run-directory flag allows controlling the run directory separately from the compilation directory. This is particularly useful in workspaces, where compiler invocations and thus diagnostics should be relative to the workspace directory, but documentation test examples should run relative to the crate directory.

The rustdoc book