issue_comments

I've decided not to bother with this, especially having redesigned some hooks to return Datasette standard Response objects.

So the Python 3.7 way to do this would be dataclasses, but I still want to support 3.5.

This excellent essay has made me think attrs might be the way to go: https://www.revsys.com/tidbits/dataclasses-and-attrs-when-and-why/

Also I have a slight preference against (ab)using __slots__ to enforce fields, although I have done it myself in the past. It would be possible to do this with __setattr__ instead, although that's an implementation detail and I'm not too fussed about it.

In this particular case I don't think there's an issue making all those required. However, I suspect we might have to allow optional values at some point - my preferred solution to russss/datasette-geo#2 would need one.

The register_output_renderer() hook currently returns a dictionary with body, content_type and status_code keys but each of these keys are optionaly. I'm tempted to make all three required to better fit this model - @russss any objections?

Alternative would be to support default values for properties of the DataSpec subclass - maybe:

__defaults__ = {"body": "", "content_type": "text/plain", "status_code": 200}

Potential design:

```python from collections import OrderedDict

class DataSpec: slots = []

def __init__(self, **kwargs):
    if list(kwargs.keys()) != self.__slots__:
        raise TypeError(
            "{}() has required arguments {} (got {})".format(
                self.__class__.__name__, self.__slots__, list(kwargs.keys())
            )
        )
    for key in self.__slots__:
        setattr(self, key, kwargs[key])

def __repr__(self):
    return "<{} {}>".format(self.__class__.__name__, dict(self.as_dict()))

def as_dict(self):
    return OrderedDict([(key, getattr(self, key)) for key in self.__slots__])

class Output(DataSpec): slots = ["body", "content_type", "status_code"] ```

I looked at using namedtuples for this but hey have one major constraint: there isn't a clean way to convert them to dictionary-style JSON: https://bugs.python.org/issue30343

So something that uses a class which knows how to be rendered as JSON would be a better fit.