Treating leading zeros consistently and restricting to SemVer versions only

[adamralph]

Currently, leading zeros are removed from the major, minor and patch parts, but they are preserved in pre-release identifiers.

For example, the tag 01.02.03-beta.0004 is parsed as version 1.2.3-beta.0004. This doesn't feel right. It seems logical to either parse as version 01.02.03-beta.0004 (preserving all leading zeros) or as version 1.2.3-beta.4 (preserving no leading zeros).

But that leads into a wider question: both convention and SemVer 2.0 put restrictions on which characters can and can't be in a specific version part. I cannot recall seeing anyone using leading zeros in major, minor, or patch parts, and SemVer disallows it, which generally makes it "OK" that MinVer strips out those leading zeros. So effectively, MinVer is restricting the version to some parts of SemVer (and general convention), and transforming some parts of the version to comply on the consumer's behalf. This also feels off. It should either be completely restrictive, or completely unrestrictive (possibly with an option to switch modes). And if it is restrictive, what better standard to use than SemVer? Also, what does "completely unrestrictive" mean? There would probably still have to be some restrictions, and not just free text.

I'm capturing this in a discussion for now because I'm not sure whether this discussion is going to reveal bugs or features, nor how many of each. There may also be two (or more) separate concerns here but I can't disentangle them yet.

Thanks to @bording for bringing this to my attention.

[adamralph]

I think #720 addresses all of this.

[teneko]

Hi @adamralph. I've written a SemVer parser I am about to release soon. It is 100% restrictive and customizable when one of these restrictions are hit. Would you be interested in it?

Here the parser and tests:
https://github.com/Teneko-NET-Tools/Teneko.VersionableMessages/blob/main/src/Teneko.VersionableMessages.SemVer.Parser/SemanticVersionParser.cs
https://github.com/Teneko-NET-Tools/Teneko.VersionableMessages/blob/main/test/Teneko.VersionableMessages.SemVer.Parser.Test/SemanticVersionParserTest.cs

You can easily use like I did here in my version class: https://github.com/Teneko-NET-Tools/Teneko.VersionableMessages/blob/main/src/Teneko.VersionableMessages.SemVer/SemanticVersion.cs

[adamralph]

I just wanted you let to know that there is something to parse that is not bound to any SemanticVersion class. :)

Bear in mind that NuGet.Versioning is an internal dependency for MinVer (from 4.0.0 onwards), so using NuGet.Versioning.SemanticVersion directly doesn't bind consumers of the MinVer package to that type. At any time, I could swap out NuGet.Versioning for something else without affecting consumers of the MinVer package. And in fact, NuGet.Versioning.SemanticVersion is useful because it takes care of other things in the SemVer spec like comparisons (albeit with the minor off-spec behaviour mentioned above).

Thanks by the way for this great piece of software, I mean MinVer I love the simple approach.

I'm glad you find it useful 🙂

I like the way you think having MinVer strictly scoped to MSBuild and make it usable over MSBuild properties

I'm not sure if you are already aware, but the minver-cli package provides a .NET tool which may be used independently of project files and SDK targets.

[teneko]

Bear in mind that NuGet.Versioning is an internal [..]

Oh that's totally fine. :)

I'm not sure if you are already aware [..]

Yes I am aware. :) Thanks for the offer. I tried it out and it works not out of the box. After specifying the prefix (-t v) minver-cli finds my tags but unfortunately cuts of my prefix in the result, so it does not get preserved unless told being discarded.

Then when I would just stick to minver-cli I would like to do all MSBuild stuff manually. Specifying NuGet version, assembly version and so on. I would like to use github workflow, a shell script what else to represent some handy branch versioning strategies. My vision is to reduce that kind of boilerplate as much as possible while being as extensible as possible.

But nevertheless I appreciate what MinVer is at the current state. Please keep it up. :)

[adamralph]

I tried it out and it works not out of the box. After specifying the prefix (-t v) minver-cli finds my tags but unfortunately cuts of my prefix in the result, so it does not get preserved unless told being discarded.

That is by design. minver-cli produces a SemVer 2.0 version. A tag prefix cannot be part of the version, since that would not be a valid SemVer 2.0 version.

[teneko]

That's ok. I think I am sticking to the word convention because at the end of the day a prefix found in a valid semantic version is just as it is a convention I can assume. But in terms of normalization you would like to discard the prefix or even replace it. So we both have a different understanding what a versioning tool should be capable of and that's ok! :)

[adamralph]

I don't understand, but never mind.