-
Notifications
You must be signed in to change notification settings - Fork 163
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Go suggestions for PURL-TYPES.rst #196
base: master
Are you sure you want to change the base?
Conversation
Full Go Module names must be case-sensitive because git (one of the supported VCSs) is case-sensitive. For instance, if you change the casing a module in your go.mod file, go will complain on "go build": module declares its path as: github.com/foo/bar but was required as: github.com/FOO/BAR GitHub treats them the same, but only because it was built with case-insensitive urls. Other git repositories may not be the same.
Go Modules has a strict definition of versioning: https://go.dev/ref/mod
@pombredanne @stevespringett are there any objections to this PR? Would be good to get these updates included in the spec, especially for go package versioning. |
Hi @pombredanne @stevespringett, would it be possible to merge this PR? Since go modules are case sensitive, package name lowercase normalization definitely seems like a bug, and it's causing some issues in our system. |
- Examples:: | ||
|
||
pkg:golang/github.com/gorilla/context@234fd47e07d1004f0aed9c | ||
pkg:golang/github.com/gorilla/[email protected] | ||
pkg:golang/google.golang.org/genproto#googleapis/api/annotations |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
☝️ This still leaves an empty version example without while removing the mention from the version
description
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
mentioned this above, but PURL-SPECIFICATION
indicates that versions can be empty
PURL-TYPES.rst
Outdated
- The ``version`` is often empty when a commit is not specified and should be | ||
the commit in most cases when available. | ||
- The ``subpath`` is used to point to a package inside a module. | ||
- The ``version`` for modules should be semver with a v prefix, or a pseudo-version that points to an untagged commit. |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
pseudo-version
is a bit wiggly language.
- The ``version`` for modules should be semver with a v prefix, or a pseudo-version that points to an untagged commit. | |
- The ``version`` may be empty. Otherwise, it may start with a lowercased "v" followed by a valid semver version or consist of a sha-1 or short sha-1 git commit hash. |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
pseudo-version
is an explicitly defined thing in the ecosystem. However, pseudo-versions also start with v so it's not clear to people who aren't familiar with the idiosyncrasies of Go that you may get something like v0.0.0-20170915032832-14c0d48ead0c
which is not semver and has weak ordering with other v0.0.0 versions and cannot be compared with non-v0.0.0 versions. It'd probably be a good idea to link to the Go documentation or give more details here.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
@matt-phylum
snipping off the "v"
0.0.0-20170915032832-14c0d48ead0c
And then running it through the regexp provided here:
https://semver.org/#is-there-a-suggested-regular-expression-regex-to-check-a-semver-string
const regexSemverNumberedGroups =
/^(0|[1-9]\d*)\.(0|[1-9]\d*)\.(0|[1-9]\d*)(?:-((?:0|[1-9]\d*|\d*[a-zA-Z-][0-9a-zA-Z-]*)(?:\.(?:0|[1-9]\d*|\d*[a-zA-Z-][0-9a-zA-Z-]*))*))?(?:\+([0-9a-zA-Z-]+(?:\.[0-9a-zA-Z-]+)*))?$/
Results in...
regexSemverNumberedGroups.exec('0.0.0-20170915032832-14c0d48ead0c')
// =>['0.0.0-20170915032832-14c0d48ead0c', '0', '0', '0', '20170915032832-14c0d48ead0c', undefined, index: 0, input: '0.0.0-20170915032832-14c0d48ead0c', groups: undefined]
a valid parse. Semver can have a trailing -stuff
, like -20170915032832-14c0d48ead0c
. It is considered metadata.
which is not semver and has weak ordering with other v0.0.0 versions and cannot be compared with non-v0.0.0 versions.
There are handy utils for parsing, coercing, and comparing those too:
https://www.npmjs.com/package/semver
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Edit: this is wrong and you can skip reading it.
It looks like semver, but it's not semver. Even if it can be parsed using a semver parser, it has none of the semantics of semantic versioning.
Pseudo-versions and regular versions can never be compared without access to the repository history. If a repository has a v1.0.0, there is almost certainly a v0.0.0-* pseudo-version which is logically greater than the v1.0.0.
Pseudo-versions can be compared with other pseudo-versions based on date, but dates do not have the same semantics as version numbers because they don't capture branching information. If there's a problem affecting v1.0.0..v1.0.1 and v2.0.0..v2.0.1, that problem is also affecting the pseudo-versions for those commits and any unreleased commits within those version ranges, and date ordering fails for commits that are made on branches parallel to the fix, or for commits that are made on the v1 series after the v2 series has been fixed but before the v1 series has been fixed. If you sorted the pseudo-versions and visualized whether they are affected or not, the result would be more like a noisy gradient than the sharp edge normally implied by those version ranges.
It's an inherent problem with the tooling trying to create versions for arbitrary commits, not a lack of library support.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Here are some docs on pseudo-versions: https://go.dev/ref/mod#pseudo-versions We could link those from the docs to clarify what the language is referring to
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
I was wrong about this. Until today, I had only ever seen pseudo-versions that started with v0.0.0. Even in the documentation about pseudo-versions, every example given starts with v0.0.0. However, there are actually pseudo-versions that start with other prefixes--they are just much less common in the data that I normally deal with. You can treat it as regular semver.
Example of a non-v0.0.0 pseudo-version:
$ go get github.com/spf13/cobra@756ba6dad61458cbbf7abecfc502d230574c57d2
go: downloading github.com/spf13/cobra v1.8.2-0.20240728161807-756ba6dad614
1.8.2-0.20240728161807-756ba6dad614 follows after v1.8.1. There is not yet a tagged 1.8.2 release. This follows the usual semver behavior of being a prerelease greater than the previous release and less than the next release.
It's still not really semver because in this example you can't infer whether one 1.8.2 pseudo-version comes before or after another in commit order, but that's a much more minor difference. In most cases it doesn't matter and it's probably not worth calling attention to beyond just linking to the pseudo-version documentation or just mentioning that it may be a pseudo-version.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
@matt-phylum Cool! I think I covered that on the JS implementation here.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
This has been codified in packageurl-js
's v2.0.0 release 🎉
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
- The
version
may be empty.
I think this is implicit based on PURL-SPECIFICATION
: The version is prefixed by a '@' separator when not empty
. If we added it here, seemed like it should be included in every other type too? (other than OCI and Swift, apparently)
Otherwise, it may start with a lowercased "v" followed by a valid semver version or consist of a sha-1 or short sha-1 git commit hash.
Since Go supports multiple VCS systems, how about something like this?
The `version` may start with a lowercased "v" followed by: a semantic version, or a
Go "pseudo-version", which consists of a semantic version followed by a timestamp
and revision identifier.
pkg:golang/google.golang.org/genproto#googleapis/api/annotations | ||
pkg:golang/github.com/gorilla/context@234fd47e07d1004f0aed9c#api |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Please keep one of the github sha-1 hash examples in
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
At least for modern Go tools, it should be a pseudo-version, not a bare commit ID like it is now: https://go.dev/doc/modules/version-numbers#in-development
@@ -258,15 +258,13 @@ golang | |||
|
|||
- There is no default package repository: this is implied in the namespace | |||
using the ``go get`` command conventions. | |||
- The ``namespace`` and `name` must be lowercased. |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
where did this definition of namespace/name go?
see #308
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
@jkowalleck your Issue adds more nuance to this problem, so I left a link in there to an argument against namespaces for Go : #63 (comment)
to expedite this PR, I'll just revert that removal for now.
Bring back the casing line for now, until that issue is resolved in the future.
More specific "version" language
Hello, Go Modules has been out for a few years and many packages have adopted it now. Here are some suggestions for updating the platform-specific instructions for golang:
remove case-sensitivity rule: this really depends on the VCS host used, but for instance, go.mod is case-sensitive, so you must use the canonical go module's casing when importing it.