[test] refactor serialization suite

[line-o]

Description:

Refactor serialization tests for maintainability and readability.

• combine item-separator with method tests into one
  ser:item-separator-with-method
• simplify test cases that used eq in combination with assertXPath
• make direct calls to function under test (serialize) in each test
• extract fixtures to module variables
• remove namespace from children of output:serialization-parameters
• convert QNames keys to strings for exist-db specific options
• convert string constructors to CDATA sections
• make HTML5 test data real world examples
• use unprefixed form of built-ins
• fix formatting

Reference:

has to wait for

• [test] fix and add adaptive serialization tests #4773
• [enhancement] allow more types in XQSuite #4816

Type of tests:

XQSuite

[line-o]

If this PR gets merged first, then #4773 can be closed.

[line-o]

This PR was rebased on current develop HEAD and does not include #4773 anymore

[line-o]

we have a problem with the docker images it seems (although this is completely unrelated to my changes):

DOCKER> Unable to build image [existdb/existdb:7.0.0-SNAPSHOT] :
"failed to export image: 
  failed to create image:
  failed to get layer sha256:89dbfd2aac2fd2efd0e12805e343c3a1139557ea1930ad204bdb7387b89fe3c4:
  layer does not exist"

@duncdrum do you have an idea what this is about or what might be the reason? Is this just a temporary hiccup and I should try again in a moment?

[adamretter]

Please consider that the following change is not equivalent:

Before this PR we had:

map { xs:QName("exist:output-doctype"): fn:false() }

But now you have:

map { "exist:output-doctype": false() })

Whilst the end result may be the same, we added the functionality and designed the tests to specifically test for the QName, this should not be removed please.

[duncdrum]

@line-o no clue could be a hickup. I would recommend a local build of the image, but that only works on intel so …

[joewiz]

I see that not all such QName constructions were removed - e.g., https://github.com/eXist-db/exist/pull/4777/files#diff-41536ff061981a585744759c566780f801e29b78405dd79d158d466f3272905dR374. I agree that it would be best to restore the others to their original form with the xs:QName construction. Perhaps the versions without the xs:QName construction could be preserved too as variations to ensure that both forms continue to work (as they do today).

[line-o]

I would be in favour of adding one test that uses constructed QNames to ensure this form will continue to work.
I believe that using QNames as keys is the less common form to provide options to fn:serialize since all other options
do not use them but plain strings.

[sonarcloud]

Kudos, SonarCloud Quality Gate passed!   

0 Bugs
0 Vulnerabilities
0 Security Hotspots
0 Code Smells

No Coverage information
No Duplication information

[adamretter]

I believe that using QNames as keys is the less common form

Not for us. Please keep our QName tests, we worked hard to add this and ensure that it works for several groups of users. If you wish to add additional tests for strings as well, I have no objection, but please do not remove our tests!

[line-o]

@adamretter all exist-specific serialization option keys are now tested as xs:string and xs:QName. Please re-review if that addresses your concerns.

[line-o]

@adamretter out of curiosity: is it possible to use QNames for the other keys in the map as well?

[line-o]

@dizzzz could you remove the build cache for this PR so that we can re-run the tests?

[sonarcloud]

Kudos, SonarCloud Quality Gate passed!   

0 Bugs
0 Vulnerabilities
0 Security Hotspots
0 Code Smells

No Coverage information
No Duplication information

[line-o]

@adamretter Your concerns were addressed 2 months ago, would you please re-review this PR

[line-o]

I will bring this PR up on the community call today.