FIX some percent-encoding and qualifier order issue

[giterlizzi]

I recently started implementing the new tests in URI-PackageURL, but after #514 I encountered some problems.

• "cocoapods", percent encoding issue
• "conan", qualifiers order issue (see Fix tests that didn't have qualifier keys sorted #551)
• "generic", qualifiers order issue (see Fix tests that didn't have qualifier keys sorted #551)
• "maven", qualifiers order issue (see Fix tests that didn't have qualifier keys sorted #551)
• "mlflow", qualifiers order issue (see Fix tests that didn't have qualifier keys sorted #551)
• "npm", percent encoding issue
• "oci", percent encoding and qualifiers order issues
• "rpm", qualifiers order issue (see Fix tests that didn't have qualifier keys sorted #551)
• "swid", percent encoding issue

This PR is companion of #551 and fix some percent-encoding except "swid" because I have a doubts about whether + should be encoded as a space (%20) or as a plus sign (%2B) in namespace and name components.

In addition, I think some Conan tests should be removed or revised because they are contradictory (e.g. 'invalid conan purl only namespace’ and 'invalid conan purl only channel qualifier’) -- see last comment of #168

(edited)

[matt-phylum]

Can you add a test to specification-test.json to ensure that + still decodes as +? Currently we're getting that testing through the cocoapods test and I agree that the cocoapods test shouldn't be doing that.

I'm pretty sure in the SWID tests the + is supposed to have been %20. Adobe has some documentation about their SWID tags and they don't use plus signs there. https://www.adobe.com/devnet-docs/acrobatetk/tools/AdminGuide/identify.html?highlight=swid

[giterlizzi]

To keep the repo tidy in terms of duplicate issues and PRs, should this PR be closed in favour of #587 ?

[jkowalleck]

To keep the repo tidy in terms of duplicate issues and PRs, should this PR be closed in favour of #587 ?

#587 was merged.
are there any cases left?

[giterlizzi]

I synced the JSON tests and no longer have sorting or percent-encoding errors in the qualifiers. However, there are still issues with some tests in other PURL types, but I will respond directly in the respective issues.

Thanks ;)