@zzamboni, Your implementation is pretty good but doesn't correctly handle prerelease and build metadata. See the spec and this solution which appears to correctly implement the spec.

@zzamboni, Also, your version doesn't correctly handle invalid semver strings. In particular, it is overly permissive by treating strings like "1.0.0" and "1.0" as equal. While at the same time not handling a very common, but incorrect, inclusion of "v" or "V" as a semver prefix. A mistake made by the current elvish implementation:

I still think your module is awesome but it points out why a solution for the semver issue I opened is so important. Specifically, the semver project should include a file of test cases so that every implementation can be validated against that implementation agnostic set of test cases.

@krader1961 I think my implementation correctly handles prerelease strings, e.g. using the example from the SemVer spec:

'semver:<' 1.0.0-alpha  1.0.0-alpha.1  1.0.0-alpha.beta  1.0.0-beta 1.0.0-beta.2 1.0.0-beta.11 1.0.0-rc.1 1.0.0
▶ $true

See also a whole bunch of other tests at https://github.com/zzamboni/elvish-modules/blob/master/semver.org.

You are correct that a missing trailing numbers (MINOR or PATCH) are considered as zeros. In my defense, this is even documented 😉: https://github.com/zzamboni/elvish-modules/blob/master/semver.org#main-comparison-function. The spec says that all three components MUST be there, so I guess this is a bug. I've opened zzamboni/elvish-modules#17

Build metadata: this is also a bug, it should be ignored. I've created zzamboni/elvish-modules#16

Invalid things like the v at the beginning - what do you mean by "doesn't correctly handle"? You mean an error should be produced? At the moment this is allowed because the module uses string comparison when needed, and numeric otherwise. I suggest in zzamboni/elvish-modules#17 that it might make sense to have a &non-strict flag or variable which allows/disallows parsing things like that.

Thanks for the feedback. PR's welcome! 😉

Invalid things like the v at the beginning - what do you mean by "doesn't correctly handle"?

@zzamboni, A "v" (or "V") prefix is technically invalid in a semver string. However, I would argue it should be silently ignored since it is all too common for an otherwise valid semver string to include that prefix. I didn't actually test that your implementation does so. I simply inferred it didn't from the documentation. If (or when) your implementation silently ignores a "v" prefix that is notable enough to warrant a mention in the documentation and why it is different from how every other invalid semver string is treated.

@krader1961 regarding your original issue, please note that as part of my semver module I implemented a compare function like the one you want. I agree this might be useful to have in Elvish:

The -signed-compare function compares two values using a function which takes two values and returns 1, 0 or -1 to represent the order of the two values (1: v1<v2, 0: v1==v2, -1: v1>v2)

fn -signed-compare [ltfn v1 v2]{
  util:cond [
    { $ltfn $v1 $v2 }  1
    { $ltfn $v2 $v1 } -1
    :else              0
  ]
}

About SemVer comparison, please check out zzamboni/elvish-modules#18, in which I address all the bugs mentioned above.