Emver

This module was designed to address the problem of releasing updates to Embassy Packages where the upstream project was either unaware of or apathetic towards supporting their application on the Embassy platform. In most cases, the original package will support semver2. This leaves us with the problem where we would like to preserve the original package's version, since one of the goals of the Embassy platform is transparency. However, on occasion, we have screwed up and published a version of a package that needed to have its metadata updated. In this scenario we were left with the conundrum of either unilaterally claiming a version number of a package we did not author or let the issue persist until the next update. Neither of these promote good user experiences, for different reasons. This module extends the semver standard linked above with a 4th digit, which is given PATCH semantics.

Usage

Add this to your Cargo.toml

[dependencies]
emver = "0.1.0"

Operations

A Version contains 4 components: major, minor, patch, and revision. The meaning of first three can be found in the Semver2 specification. The fourth is also given patch semantics but is intended to be incremented by package distributors when they are not themselves authors.

A Version can also be parsed from a dot separated string like 0.1.2.3. They can also be serialized to strings, but in cases where the last digit is zero, the last dot and the zero are omitted. Parsing will still work over a triple. The relevant parse function for Version is parse_version and it is generated by the nom parser library. It can be applied to a string and will produce a Result<Version, ParseError>.

The other half of this library deals with the type VersionRange. A VersionRange is a set that is either anchored at a particular Version with some sort of comparison operator: = >= <= > < or it is described as a conjunction or disjunction of other VersionRanges. For convenience we also provide two constructors (Any, None) to serve as identity elements on the Conj and Disj constructors respectively. As a result, to gain maximum performance, you should use the conj and disj smart constructors as opposed to their dumb counterparts Conj and Disj. This will immediately evaluate the identities and annihilators as opposed to building up the AST further, saving peak memory.

For convenience, there are two Monoid wrappers exposed: (AnyRange, AllRange). This allows you to fold an Iterable with the combine operation seeded with the empty value. The semantic differences are whether or not combine uses disj or conj respectively.

Most of the time you will want to parse these values from strings, but the internals are exposed for the rarer cases. Some of the grammar from node-semver is supported (^1.2.3, ~2.3.4, 1.2.x, 0.0.0-1.1.1) as well.

Finally, the most useful operation in this package is the satisfies operation on Version with the argument of a VersionRange. This is simply a predicate that tells you whether the Version falls inside the VersionRange.

Laws

All laws listed below are equality of observation, not a literal Eq instance giving representational Equality. The only observer that this library has is the satisfies operation. When you read "a === b", you should interpret that as obs.satisfies(a) === obs.satisfies(b). These laws simply mean that it is always safe to do a substitution of a term on the LHS for a term on the RHS without changing the meaning of your program.

• Conj is commutative: conj(a,b) === conj(b,a)
• Disj is commutative: conj(a,b) === disj(b,a)
• Conj is associative: conj(conj(a,b),c) === conj(a,conj(b,c))
• Disj is associative: disj(disj(a,b),c) === disj(a,disj(b,c))
• Any is identity of Conj: conj(a, Any) === a
• None is identity of Disj: disj(a, None) === a
• Any annihilates Disj: disj(a, Any) === Any
• None annihilates Conj: conj(a, None) === None
• Conj distributes over Disj: conj(a,disj(b,c)) === disj(conj(a,b),conj(a,c))
• Disj distributes over Conj: disj(a,conj(b,c)) === conj(disj(a,b),disj(a,c))