Fix bazel tests

[TG1999]

Reference:

• Fix Bazel tests #729

Related:

• Expected output in testcases do not follow spec encoding #644

[johnmhoran]

@TG1999 One preliminary question I have: if the input PURL has an unencoded slash in the qualifiers value, doesn't that raise an exception, halt the evaluation process, and thus require a null expected_output and a true expected_failure?

@mjherzog @pombredanne @matt-phylum @jkowalleck What do you think?

[matt-phylum]

If the input PURL has an unencoded slash, an exception should not be thrown. It is not part of the spec, and it breaks existing PURLs where slash is not encoded, including examples from previous versions of the spec and output from code that implements an older version of the spec.

When I last updated the purl-survey output, 10 of 15 implementations passed tests that required slash to be unencoded in qualifier values. PURLs canonicalized by those implementations would be broken by turning unencoded slashes in qualifiers into errors.

[johnmhoran]

Thanks @matt-phylum . Putting aside slashes for a moment (which the submitted standard makes clear MUST be percent-encoded in a qualifiers value however we're meant to distinguish between MUST and SHOULD and MAY), if a roundtrip input has no pkg:, expected_output will be null and expected_failure will be true -- right? Surely a tool will not "fix" that deficiency by supplying the missing pkg:.

[petergardfjall]

Anything preventing this from being merged?

It is similar in spirit to #713, #714, #715, #716, #717, #718, #719, #720, #721

[pombredanne]

Looking good now... I added a few touch ups based on @mjherzog reviews.

[pombredanne]

All clear now. Merging.