Improve an error message on loaddata/dumpdata when PyYAML is not installed.

[Brad]

While the docs at ​https://docs.djangoproject.com/en/3.2/topics/serialization/#serialization-formats do mention that PyYAML is required for the yaml serializer, ./manage.py dumpdata|loaddata --format yaml, provides quite a sparse error message, "Unable to serialize database: module yaml not found" (from here: ​https://github.com/django/django/blob/fbb1984046ae00bdf0b894a6b63294395da1cce8/django/core/management/commands/dumpdata.py#L245) if PyYAML is not installed.

This could be especially confusing for new Django developers who are not aware that yaml is pointing to the PyYAML project on PyPI and is not a part of Django. Since yaml is the only one of the four serializers requiring an add-on third party package, it may be helpful to provide a more verbose error in this case.

For instance, something like. "Serializing to and from YAML requires the PyYAML package. Install PyYAML to use the YAML format."

[Mariusz Felisiak]

Agreed, we can raise an ImproperlyConfigured error, e.g.

try:
    import yaml
except ImportError as e:
    raise ImproperlyConfigured(
        'Error loading yaml module. Did you install PyYAML?'
    ) from e

[Brad]

Interestingly, I found a comment in django/core/serializers/pyyaml.py:

# Requires PyYaml (https://pyyaml.org/), but that's checked for in __init__."

However, I'm not seeing any such check in django/core/serializers/__init__.py. (I expected a HAS_PYYAML or something like that.)

There are also some tests in tests/serializers/test_yaml.py that test specifically for

YAML_IMPORT_ERROR_MESSAGE = r'No module named yaml'

So those will need to be revised.

[Brad]

Some interesting background:

• ​https://github.com/django/django/pull/1582/files
• https://code.djangoproject.com/ticket/12756

[Brad]

Having read the background from Ticket 12756, the comment from core/serializers/pyyaml.py about the checking being done in __init__.py makes more sense now. The check that's referring to is here:

​https://github.com/django/django/blob/main/django/core/serializers/__init__.py#L70

So again, the import error is caught correctly, but in this ticket I'm proposing adding a friendlier message in the special case of yaml/PyYAML.

[Brad]

PR: ​https://github.com/django/django/pull/14729

Old:

$ ./manage.py dumpdata --format yaml --pks 1 users.User
CommandError: Unable to serialize database: No module named 'yaml'

New:

$ ./manage.py dumpdata --format yaml --pks 1 users.User
CommandError: Unable to serialize database: Error loading yaml module. Did you install PyYAML?

[Brad]

I do realize this is arguably an addition of kludgy code, especially the test_yaml.py changes, in exchange for a small improvement in the error message, so my feelings won't be hurt either way on this one.

[Mariusz Felisiak]

Replying to Brad:

I do realize this is arguably an addition of kludgy code, especially the test_yaml.py changes, in exchange for a small improvement in the error message, so my feelings won't be hurt either way on this one.

Thanks for all your efforts! I agree that it's not worth changing.