Display

fmt::Debug hardly looks compact and clean, so it is often advantageous to customize the output appearance. This is done by manually implementing fmt::Display, which uses the {} print marker. Implementing it looks like this:

1. // Import (via `use`) the `fmt` module to make it available.
2. use std::fmt;

4. // Define a structure which `fmt::Display` will be implemented for. This is simply
5. // a tuple struct containing an `i32` bound to the name `Structure`.
6. struct Structure(i32);

8. // In order to use the `{}` marker, the trait `fmt::Display` must be implemented
9. // manually for the type.
10. impl fmt::Display for Structure {
11. // This trait requires `fmt` with this exact signature.
12. fn fmt(&self, f: &mut fmt::Formatter) -> fmt::Result {
13. // Write strictly the first element into the supplied output
14. // stream: `f`. Returns `fmt::Result` which indicates whether the
15. // operation succeeded or failed. Note that `write!` uses syntax which
16. // is very similar to `println!`.
17. write!(f, "{}", self.0)
18. }
19. }

fmt::Display may be cleaner than fmt::Debug but this presents a problem for the std library. How should ambiguous types be displayed? For example, if the std library implemented a single style for all Vec<T>, what style should it be? Either of these two?

• Vec<path>: /:/etc:/home/username:/bin (split on :)
• Vec<number>: 1,2,3 (split on ,)

No, because there is no ideal style for all types and the std library doesn’t presume to dictate one. fmt::Display is not implemented for Vec<T> or for any other generic containers. fmt::Debug must then be used for these generic cases.

This is not a problem though because for any new container type which is not generic,fmt::Display can be implemented.

1. use std::fmt; // Import `fmt`

3. // A structure holding two numbers. `Debug` will be derived so the results can
4. // be contrasted with `Display`.
5. #[derive(Debug)]
6. struct MinMax(i64, i64);

8. // Implement `Display` for `MinMax`.
9. impl fmt::Display for MinMax {
10. fn fmt(&self, f: &mut fmt::Formatter) -> fmt::Result {
11. // Use `self.number` to refer to each positional data point.
12. write!(f, "({}, {})", self.0, self.1)
13. }
14. }

16. // Define a structure where the fields are nameable for comparison.
17. #[derive(Debug)]
18. struct Point2D {
19. x: f64,
20. y: f64,
21. }

23. // Similarly, implement for Point2D
24. impl fmt::Display for Point2D {
25. fn fmt(&self, f: &mut fmt::Formatter) -> fmt::Result {
26. // Customize so only `x` and `y` are denoted.
27. write!(f, "x: {}, y: {}", self.x, self.y)
28. }
29. }

31. fn main() {
32. let minmax = MinMax(0, 14);

34. println!("Compare structures:");
35. println!("Display: {}", minmax);
36. println!("Debug: {:?}", minmax);

38. let big_range =   MinMax(-300, 300);
39. let small_range = MinMax(-3, 3);

41. println!("The big range is {big} and the small is {small}",
42. small = small_range,
43. big = big_range);

45. let point = Point2D { x: 3.3, y: 7.2 };

47. println!("Compare points:");
48. println!("Display: {}", point);
49. println!("Debug: {:?}", point);

51. // Error. Both `Debug` and `Display` were implemented but `{:b}`
52. // requires `fmt::Binary` to be implemented. This will not work.
53. // println!("What does Point2D look like in binary: {:b}?", point);
54. }

So, fmt::Display has been implemented but fmt::Binary has not, and therefore cannot be used. std::fmt has many such traits and each requires its own implementation. This is detailed further in std::fmt.

Activity

After checking the output of the above example, use the Point2D struct as guide to add a Complex struct to the example. When printed in the same way, the output should be:

1. Display: 3.3 +7.2i
2. Debug: Complex { real: 3.3, imag: 7.2 }

See also

derive, std::fmt, macros, struct, trait, and use