Lenient Semantic Version Parser

Lenient parser for Semantic Version numbers.

Motivation

This crate aims to provide an alternative parser for semver Versions.

Instead of adhering to the semver specification, this parser is more lenient in what it allows. The differenc include:

• Minor and Path are optional an default to 0 (e.g. "1" parses as "1.0.0")
• Pre-release identifier may be separated by . as well (e.g. "1.2.3.rc1" parses as "1.2.3-rc1")
• Some pre-release identifiers are parsed as build identifier (e.g. "1.2.3.Final" parses as "1.2.3+Final")
• Additional numeric identifiers are parsed as build identifier (e.g "1.2.3.4.5" parses as "1.2.3+4.5")
• A leading v or V is allowed (e.g. "v1.2.3" parses as "1.2.3")
• Numbers that overflow an u64 are treated as strings (e.g. "1.2.3-9876543210987654321098765432109876543210" parses without error)

This diagram shows lenient parsing grammar

Examples

```rust use semver::Version;

let version = lenientsemver::parse("1.2.3"); asserteq!(version, Ok(Version::new(1, 2, 3)));

// examples of a version that would not be accepted by semverparser asserteq!( lenientsemver::parse("1.2.M1").unwrap(), Version::parse("1.2.0-M1").unwrap() ); assert!(Version::parse("1.2.M1").iserr());

asserteq!( lenientsemver::parse("1").unwrap(), Version::parse("1.0.0").unwrap() ); assert!(Version::parse("1").is_err());

asserteq!( lenientsemver::parse("1.2.3.Final").unwrap(), Version::parse("1.2.3+Final").unwrap() ); assert!(Version::parse("1.2.3.Final").is_err());

asserteq!( lenientsemver::parse("1.2.3.4.5").unwrap(), Version::parse("1.2.3+4.5").unwrap() ); assert!(Version::parse("1.2.3.4.5").is_err());

asserteq!( lenientsemver::parse("v1.2.3").unwrap(), Version::parse("1.2.3").unwrap() ); assert!(Version::parse("v1.2.3").is_err());

asserteq!( lenientsemver::parse("1.2.9876543210987654321098765432109876543210").unwrap(), Version::parse("1.2.0-9876543210987654321098765432109876543210").unwrap() ); assert!(Version::parse("1.2.9876543210987654321098765432109876543210").is_err()); ```

Parsing into custom versions

The parser is not fixed on returning a semver::Version, it instead parses into a lenient_semver::VersionBuilder. The default features for this crate contain a VersionBuilder implementation for semver::Version, but any implementation can be used with parse_into.

Examples

```rust use lenient_semver::VersionBuilder;

/// Simpler version struct that lives only on the stack

[derive(Debug, Default)]

struct MyVersion { numbers: [u64; 3], isprerelease: bool, }

/// The VersionBuilder trait is generic over the lifetime of the input string. /// We don't store references to those strings, so we don't care about the specific lifetime. impl VersionBuilder<'_> for MyVersion { /// We will modify the target struct directly type Out = Self;

/// Construct a new builder instance.
/// One can only expect `set_major` to be called before `build`, all other methods are optional.
fn new() -> Self {
    Self::default()
}

/// Construct the final result. In this case, we can just return ourselves.
fn build(self) -> Self::Out {
    self
}

/// Called when the major component was found.
fn set_major(&mut self, major: u64) {
    self.numbers[0] = major;
}

/// Called when the minor component was found.
fn set_minor(&mut self, minor: u64) {
    self.numbers[1] = minor;
}

/// Called when the patch component was found.
fn set_patch(&mut self, patch: u64) {
    self.numbers[2] = patch;
}

/// Called when any pre-relase metadata identifier was found.
/// This identifier can just numeric, no attempts at parsing it into a number have been made.
/// For this implementation, we don't care about the value, just it's presence.
fn add_pre_release(&mut self, _pre_release: &str) {
    self.is_pre_release = true
}

}

let input = "1.3.3.7-alpha21+build.42"; let myversion = lenientsemver::parse_into::(input).unwrap();

asserteq!([1, 3, 3], myversion.numbers); assert!(myversion.ispre_release); ```

The VersionBuilder has empty default implementation for the various methods, making it easy to use it for use-cases beyond just parsing. The following example implements a function that checks if a given string represents any form of pre-release version.

```rust use lenient_semver::VersionBuilder;

/// newtype around bool, so we can implement the VersionBuilder trait for it

[derive(Debug, Default)]

struct IsPreRelease(bool);

impl VersionBuilder<'_> for IsPreRelease { /// Here we parse into a different value than Self type Out = bool;

fn new() -> Self {
    Self::default()
}

/// Return the wrapped bool
fn build(self) -> Self::Out {
    self.0
}

/// We only care about this method and can ignore all the other ones
fn add_pre_release(&mut self, _pre_release: &str) {
    self.0 = true;
}

}

/// This method also return false for invalid version strings, /// which is technically true, as those are not pre-release versions. /// Usually you would want to have a better error handling. fn isprerelease(v: &str) -> bool { lenientsemver::parseinto::(v).unwrapordefault() }

assert!(isprerelease("1.2.3-pre") == true); assert!(isprerelease("1.2.3") == false); assert!(isprerelease("1.2.3+build") == false); ```

Features

lenient_semver comes with a number of features:

| feature name | default enabled | transitive dependencies | purpose | -------------: | --------------- | ----------------------- | -------- | semver11 | yes | semver = "0.11.0" | Provides VersionBuilder implementation for semver = "0.11.0". | semver10 | no | semver = "0.10.0" | Provides VersionBuilder implementation for semver = "0.10.0". | versionlite | no | lenient_version = "*" | A custom Version as alternative to semver::Version that complements some leneient features, such as additional numbers beyond patch. | versionsemver | no | lenient_version = "*" | Add conversions From lenient_version Into semver::Version. | version_serde | no | serde = "1" | Serde Deserializer and Serializer implementation for lenient_version.

Examples

semver11

toml lenient_semver = { version = "*", features = [ "semver11" ] }

```rust use semver::Version as Version11;

// This features is enabled by default and is usable through parse directly, // but can also be used with parse_into. let version = lenientsemver::parseinto::("v1.2.3.Final").unwrap(); assert_eq!(version, Version11::parse("1.2.3+Final").unwrap()); ```

semver10

toml lenient_semver = { version = "*", features = [ "semver10" ] }

``rust // We have both version of semver available, the older one // is renamed tosemver010`. use semver010::Version as Version10;

// The default parse is fixed to the latest semver::Version, // so we need to use parse_into. let version = lenientsemver::parseinto::("v1.2.3.Final").unwrap(); assert_eq!(version, Version10::parse("1.2.3+Final").unwrap()); ```

version_lite

toml lenient_semver = { version = "*", features = [ "version_lite" ] }

With this features, lenientsemver now comes with it's own version. That particular implementation supports numbers beyond patch directly. Note that lenientsemver still parses those additional number without complaining, but they are added as build attribute to semver Versions.

```rust use lenient_semver::Version;

let version = lenientsemver::parseinto::("1.3.3.7").unwrap(); assert_eq!(version, Version::parse("1.3.3.7").unwrap()); // Version::parse delegates to this parser ```

The native support allows such version to be compared properly, which does not work with semver.

```rust use lenient_semver::Version;

let versiona = Version::parse("1.3.3.7").unwrap(); let versionb = Version::parse("1.3.3.8").unwrap(); assert!(versiona < versionb);

// with semver, that fails: let versiona = lenientsemver::parse("1.3.3.7").unwrap(); let versionb = lenientsemver::parse("1.3.3.8").unwrap(); asserteq!(versiona < versionb, false); asserteq!(versiona, versionb); ```

Furthermore, Version does not own the data for the metadata identifiers. The metadata can be disassociated, so the version can reference a different owner.

```rust use lenient_semver::{Version, VersionBuilder};

let input = "1.3.3.7-beta.21+build.42"; // make an owned copy, so we don't cheat by using the 'static lifetime. let input = String::from(input);

// This version references slices from the input String let version = lenientsemver::parseinto::(input.as_ref()).unwrap();

// Which prevents us from dropping the input // drop(input);

// We can disassociate the metadata, which allows the new version to reference something else let (mut version, pre, build) = version.disassociate_metadata();

// We still get the referenced input slices, so we create owned copies let pre: Vec = pre.intoiter().map(ToOwned::toowned).collect(); let build: Vec = build.intoiter().map(ToOwned::toowned).collect();

// now we can safely drop the input drop(input);

// We can also re-add the cloned identifiers. // The version would now be bound to the lifetime of this method. // Just for fun, we swap pre-release and build for pre in &pre { version.addbuild(pre.asref()); } for build in &build { version.addprerelease(build.as_ref()); }

asserteq!("1.3.3.7-build.42+beta.21".tostring(), version.to_string()); ```

version_semver

toml lenient_semver = { version = "*", features = [ "version_semver" ] }

If you need to store an owned copy of the version information, you should copy into semver::Version or your custom version type instead. If you only ever intend to store the version information, it might make more sense to parse directly into semver::Version instead.

```rust use semver::Version;

let input = String::from("v1.3.3.7-beta-21+build-42"); let version = lenientsemver::Version::parse(&input).unwrap(); let version = Version::from(version); asserteq!("1.3.3-beta.21+7.build.42", &version.to_string()); ```

version_serde

toml lenient_semver = { version = "*", features = [ "version_serde" ] }

This feature also enabled version_lite and brings serde support for the own Version type. Since lenient_semver::Version does not take ownership of the metadata identifiers, the lifetime of the deserialization result is bound to the input.

```rust use lenient_semver::{Version, VersionBuilder}; use serde::Deserialize;

[derive(Debug, Deserialize)]

struct DependencySpec<'input> { /// Refer to name as owned value name: String, /// Borrows from the input string #[serde(borrow)] version: Version<'input>, }

let input = " { \"name\": \"lenient_semver\", \"version\": \"1.3.3.7+build.42\" }"; // make an owned copy, so we don't cheat by using the 'static lifetime. let input = String::from(input);

// use serde as one would normally do let dep: DependencySpec = serdejson::fromstr(input.as_ref()).unwrap(); println!("{:?}", dep);

// cannot move out of input because it is borrowed // drop(input);

let mut expected = Version::new(1, 3, 3); expected.addadditional(7); expected.addbuild("build"); expected.add_build("42");

assert_eq!(dep.version, expected);

// now we can drop the input drop(input); ```

parse_partial

toml lenient_semver = { version = "*", features = [ "parse_partial" ] }

This feature enables the partial feature of the parser. The partial parser will not try to consume all input. Instead it parses the version as far as possible and will return the unconsumed input alongside the parsed version.

```rust use lenient_semver::{Version, VersionBuilder, parser}; let input = "1.2.3 42+build 1.3.3.7 // end";

// parse first version let (version, remainder) = parser::parsepartial::(input).unwrap(); let expected = Version::new(1, 2, 3); asserteq!(version, expected); // trailing whitespace is considered part of a version and consumed as well assert_eq!("42+build 1.3.3.7 // end", remainder);

// parse second version let (version, remainder) = parser::parsepartial::(remainder).unwrap(); let mut expected = Version::new(42, 0, 0); expected.addbuild("build"); asserteq!(version, expected); asserteq!("1.3.3.7 // end", remainder);

// parse last version let (version, remainder) = parser::parsepartial::(remainder).unwrap(); let mut expected = Version::new(1, 3, 3); expected.addadditional(7); asserteq!(version, expected); asserteq!("// end", remainder);

// parse partial still expects to parse something. // It will fail with UnexpectedInput or MissingMajorNumber if the input does not match at least a major version. // let's try to parse the remaining input let error = parser::parsepartial::(remainder).unwraperr(); asserteq!(error.errorkind(), parser::ErrorKind::UnexpectedInput); asserteq!(error.errorline(), "Unexpected /");

// or an empty string let error = parser::parsepartial::(" ").unwraperr(); asserteq!(error.errorkind(), parser::ErrorKind::MissingMajorNumber); asserteq!( error.errorline(), "Could not parse the major identifier: No input" );

// The rules of when a certain number will be parsed are even more relaxed let (version, remainder) = parser::parsepartial::("1foobar").unwrap(); let expected = Version::new(1, 0, 0); asserteq!(version, expected); assert_eq!(remainder, "foobar");

// Furthermore, the characters * and ? are allowed to appear everywhere where other alphabetic character are allowed. // This relaxes the rule that only a-z, A-Z, and 0-9 are allowed. // Those characters have no special meaning and will be parsed as pre-release or build segment. let (version, remainder) = parser::parsepartial::("1.2.*+final?").unwrap(); let mut expected = Version::new(1, 2, 0); expected.addprerelease("*"); expected.addbuild("final?"); asserteq!(version, expected); asserteq!(remainder, ""); ```

License: MIT OR Apache-2.0