Talk submission type is not always localized in API

[twavv]

Current Behavior

Currently, querying talks from the API results in either an object with text for various localizations or a plain string (in my experience, using JuliaCon 2019, only the "talk" type is a plain string, everything else is an object with an "en" key).

Expected Behavior

The type of submission_type should be consistent.

Steps to Reproduce

# Real, publicly available data
curl https://pretalx.com/api/events/juliacon2019/talks?offset=25

{
    "count": 156,
    "next": "https://pretalx.com/api/events/juliacon2019/talks/?limit=25&offset=50",
    "previous": "https://pretalx.com/api/events/juliacon2019/talks/?limit=25",
    "results": [
        {
            "code": "YZPSDK",
            ...
            "submission_type": {
                "en": "Lightning Talk"
            },
            ...
        },
        ...
        {
            "code": "8AM9JC",
            ...
            "submission_type": "Talk",
            ...
        },
        ...
    ]
}

[rixx]

Thank you for bringing this up! This is definitely unexpected/buggy behaviour.

This is due to the behaviour of the library django-i18nfield that we use for internationalisation. We're looking to get a fix for this issue worked into the library, and will then pull it into pretalx.

[oxinabox]

@travigd this is one of the things that Pretalx.jl does; normalize that field.

[rixx]

@oxinabox If you encounter bugs in pretalx like that, it would be very helpful to the project if we knew about them, so that we can fix them – bug reports are welcome.

[oxinabox]

I didn't really click that it was a bug.
Thought it was a feature I was using wrong