Releases: jprochazk/garde
v0.22.1
v0.22.0
What's Changed
- Allow an optional context when using dive by @lasantosr in #143
Full Changelog: v0.21.1...v0.22.0
v0.21.1
What's Changed
- fix: exclude
cfgofjs-sysfrom macro expansion by @ChieloNewctle in #141
New Contributors
- @ChieloNewctle made their first contribution in #141
Full Changelog: v0.21.0...v0.21.1
v0.21.0
What's Changed
- feat: support
equalfor length and range rules by @Rolv-Apneseth in #122 - chore: update axum-test dependency to 15.5 by @JosephLenton in #126
- Update to Rust 1.81 by @jprochazk in #139
- Add rust_decimal support by @lasantosr in #135
- Implement
Matcherforstd::sync::LazyLock<T>by @Fogapod in #131
Breaking changes
The MSRV is now 1.81.
⚠️ The axum_garde crate has been deprecated. ⚠️
The last release for axum_garde is 0.21. I recommend switching to axum-valid as soon as possible, as axum_garde is no longer maintained and will not receive updates of any kind.
The reason is that I personally have not been using the crate, and it has never felt like quite the right solution for integrating it into axum. I can't invest time into maintaining it anymore. axum-valid supports garde, among other validation crates, and moving to it should not be too much effort. A find-and-replace of axum_garde::WithValidation for axum_valid::Garde should get you most of the way there.
The source code for it has been removed from this repository as of 0c8823f.
New Contributors
- @Rolv-Apneseth made their first contribution in #122
- @Fogapod made their first contribution in #131
Full Changelog: v0.20.0...v0.21.0
v0.20.0
v0.19.2
What's changed
- Fixed a bug introduced in
v0.19.1that would causegardeto not compile when compiled with the--no-default-features -F email.
Full Changelog: v0.19.1...v0.19.2
v0.19.1
v0.19.0
What's Changed
Validate without context
Previously, validate always required a context parameter, even if it was not actually used. In #102, @lasantosr split the validate function into validate and validate_with, where the former does not require a Context parameter if Context implements Default.
#[derive(garde::Validate)]
struct Foo<'a> {
#[garde(length(min = 1))]
bar: &'a str,
}
Foo { bar: "test" }.validate() // no more `&()`!Add field matching rule
@darkenmay added a new matches rule in #110, which adds support for the basic case of matching one field against another in the same struct:
#[derive(garde::Validate)]
struct User<'a> {
#[garde(length(min = 1))]
password: &'a str,
#[garde(matches(password))] // checks if this field equals `password`.
password2: &'a str,
}
User { password: "ferris123", password2: "ferris123" }.validate() Note that this currently does not work for enums.
Other improvements
- Add Report::into_inner by @lasantosr in #107
- feat: implement
ValidateforCowby @bonofiglio in #108
New Contributors
- @lasantosr made their first contribution in #102
- @darkenmay made their first contribution in #110
Full Changelog: v0.18.0...v0.19.0
v0.18.0
There are quite a few changes in this release:
- Length modes in #88
- Adapters in #93
- Remove default features, replace with
fullin #95 - Serialize/deserialize
Reportin #90 - Remove empty colon from display impl of
Reportin #91
Some of these are breaking changes. If you are only using the derive functionality and calling validate, updating should still be as simple as a version bump:
cargo add garde@0.18 -F full
If it isn't, please open an issue!
Length modes
You can now specify what kind of length you wish to validate by adding a new mode argument to your length rules. The argument is optional.
#[derive(garde::Validate)]
#[garde(transparent)]
struct Username(
#[garde(
length(bytes, min = 1, max = 100),
length(graphemes, min = 1, max = 25)
)]
String
);The above validates both bytes (via v.len()) and graphemes (via v.graphemes().count() using the unicode-segmentation crate. To understand why the distinction is important, consider the x̧̡̬̘͓̖̲̻̻̲̠̪̻͓͙̜̂̓̊̔̀̀͗̑̀̅̀̂̚͘̕̚͘͢͜͠ character, which occupies 73 bytes in memory, but is considered a single grapheme.
The available modes are:
simple, which is the default, and its behavior depends on the type. For strings, it measures the number of bytes, and for collections (such asVec), it meaures the number of items.bytes, which usess.len(), measuring the number of bytes,graphemes, which usess.graphemes(true).count(), measuring the number of graphemeschars, which usess.chars().count(), measuring the number of unicode scalar valuesutf16, which usess.encode_utf16().count(), measuring the number of UTF-16 code units
The original Length and HasLength traits have been removed, and replaced by one trait per length mode.
Adapters
Using adapters, it is possible to implement validation rules for 3rd-party types directly in your own crates.
mod my_str_adapter {
#![allow(unused_imports)]
pub use garde::rules::*; // re-export garde's rules
pub mod length {
pub use garde::rules::length::*; // re-export `length` rules
pub mod simple {
// re-implement `simple`, but _only_ for the concrete type &str!
pub fn apply(v: &str, (min, max): (usize, usize)) -> garde::Result {
if !(min..=max).contains(&v.len()) {
Err(garde::Error::new("my custom error message"))
} else {
Ok(())
}
}
}
}
}Adapters are applied using the adapt field-level attribute:
#[derive(garde::Validate)]
struct Stuff<'a> {
#[garde(
adapt(my_str_adapter),
length(min = 1),
ascii,
)]
v: &'a str,
}Now my_str_adapter::length::simple will be used instead of garde::rules::length::simple when validating the above type, removing the need for a newtype over str!
full feature
Having default features for a crate such as garde means that downstream users are more likely to end up with those default features in their lockfile with no way to turn them off, resulting in increased compile times in exchange for no benefit. Due to this, the default feature from garde has been removed, and we now have a new full feature which contains everything from the removed default feature.
To retain all the features previously in default, enable the full feature when you update:
cargo add garde@0.18 -F full
I urge you to consider if you really need all of full, though! There may be parts you're not using, and taking a few minutes to see what is the minimum feature set with which your project compiles is definitely worth it.
Miscellaneous
-
The
Reporttype now has implementations ofserde::Deserializeandserde::Serializewhen theserdefeature is enabled. Note that the implementation ofSerializehas been updated to make it possible to implement a losslessserialize -> deserializeroundtrip, which is a breaking change if you were doing something with the serialized reports. -
Validation errors on newtypes previously looked like this:
: length is lower than 1This is because newtypes report their errors with an empty path, resulting in a lone
:in the error message. This has now been fixed, and will report as:length is lower than 1