test: mock pypandoc.convert_text when testing export views

[afuetterer]

Description

Resolves #1092 and #1113

Tasks

• add mocked_convert_text fixture, which should be used in combination with export_format fixture, when testing export views
• add mocked_convert_text to all export view test functions
• remove special treatment of XML export (e.g. assert root.tag == 'rdmo')
• add a new parametrized "unit test", that tests all export formats with one input, see test: mock pypandoc.convert_text when testing export views #1117 (comment)
• test for invalid export format
• assert status codes
• assert Content-Type header
• assert Content-Disposition header
• rebase branch 2.3.0
• squash commits

Types of Changes

• Refactoring (no functional changes, no api changes)

Checklist

• I have read the contributor guide.
• My code follows the code style of this project.

[afuetterer]

How should I test render_to_format? With a single project from the database and once for every export format, meaning number of tests = number of export formats.

It could also be number of rdmo models * number of export formats.

Both is better than running pypandoc for each user in multiple parametrized tests.

[jochenklar]

I think it should suffice to run pandoc with one html output for each format. For some reason render_to_format takes a template and a context (maybe it would be better to have function to render the template and one to convert with pandoc, but this can be done in the future.) I would use the Test project 1 and the project_answers view. So basically this view: https://github.com/rdmorganiser/rdmo/blob/main/rdmo/projects/views/project_answers.py#L60.

[jochenklar]

I think of a dedicated test like this:

    project = Project.object.get(1)
    project.catalog.prefetch_elements()

    context = {
        'project': project,
        'project_wrapper': ProjectWrapper(project),
        'title': project.title,
        'format': format,
        'resource_path': get_value_path(project)
    }

    render_to_format(request, format, project.title,
                     'projects/project_answers_export.html', context)

I wonder if we ever would ever look at the outputs. I think just checking if the render_to_format runs without error is enough.

[afuetterer]

I still need to add the "one" test, but the mocking shaved off a couple of minutes of ci runtime. It is now under 30 minutes.

[afuetterer]

render_to_format, export_format xml and project don't play along. xml was not tested previously. This kind of data can not be exportes as xml?

Is the test still alright?

[jochenklar]

xml is not using render_to_format so it does not need to be tested here. It should also stay in the regular tests and not be mocked.

[jochenklar]

Looks fine otherwise, thanks a lot!

[afuetterer]

🤦‍♂️

Okay, thanks for the hint. Only pypandoc.convert_text is mocked.

I removed multiple checks of the generated XML. Should these checks stay then?

[jochenklar]

yes, the xml serializers/renderers stuff is pretty unique and also fast, please keep it in

[jochenklar]

xml is not using render_to_format so it does not need to be tested here. It should also stay in the regular tests and not be mocked.

[jochenklar]

Looks fine otherwise, thanks a lot!

[afuetterer]

I am confused. What is the next target branch for PRs?

release-2.3.0 or 2.3.0?

[jochenklar]

Hi @afuetterer , the new development branch is 2.3.0, I switch branches a lot and I don't want to type release all the time. After the release the tag willbe 2.3.0 I hope this does not cause problems.