Pydantic: Support for parsing from YAML file

[drpoggi]

Hello, and thanks for creating this wonderful library. I was wondering if you would merge a PR that added YAML as a possible parsing protocol? Seems fairly trivial to add in parser.py, but I wanted to check with you before doing the work. (I would implement using pyyaml unless you think another one would be better)

[samuelcolvin]

Thanks for the question, and offering to submit a PR.

The answer is probably no, sorry.

The reason is that on reflection I think that parse.py was a mistake in the first place, and I'm intending to remove it. In turn, my reasoning for removing parsing from pydantic is as follows:

• deserialising is completely separate from data validation: once you've deserialised some data pydantic really doesn't care where it came from
• it's generally very simple, just Model.parse_obj(**xxx.loads(data))
• there are numerous choices and sub choices, many of them requiring external libraries: json (std.json, ujson, simplejson, load vs. loads), msgpack (msgpack-python, msgpack, u-msgpack-python), yaml (PyYAML, yaml, python-yaml), parsing environment variables, maybe even xml
• even with these options I often end up using other methods, for example with aiohttp I generally use Model.parse_obj(await request.json())
• the exceptions are all different and you might well want to deal with them differently, eg. returning 400 or 415

So instead of this, I think it would be better to add a section to the help docs demonstrating how to load data from numerous different formats.

Does this make sense?

[drpoggi]

That makes perfect sense. I hadn't seen any discussion on adding YAML in the issue tracker or other PR's so I wanted to raise it here, but I definitely see your reasoning for not wanting to add it. Thanks for taking the time to explain.

[SaschaSchlemmer]

If you remove parsing from pydantic, what will happen with BaseSettings which currently parses strings as json?