Coverage increased (+0.06%) to 95.379% when pulling 786186f on dtao:dtao/add-schema-infer-method into 95489bd on alecthomas:master.

I have written code like this for my own projects and found it useful. I figured it might be nice to add it to the library itself.

Forgive me if this functionality already exists. I couldn't find it!

@dtao aside from the possible discussion of whether this logic should be part of this library, I wonder whether it should take multiple samples to infer the schema?

Because it may be that a schema "should be" a list of multiple values (e.g., Optional("gender"): In(["male", "female", "other"]) rather than Optional("gender"): "male").

I guess this functionality can be added to the library. Also, I agree with the point @svisser made.

@dtao Can we have Schema.infer taking multiple values?

I'm open to changing things, but from @svisser's comment I think we might not be on the same page re: what this method does (or should do). This is certainly at least partially my fault for not being more thorough in the PR description (and perhaps in the docstring as well).

The way the method works in the current implementation isn't that "male" gets converted to "male"; rather, it get converted to str. The use case I have in mind is that you have a known-good object (what is sometimes called "golden data")—perhaps already available in a handy JSON file—and you want to create a schema to validate objects that look like that object, i.e. have the same keys, with values of the same types.

So:

s = Schema.infer({
    'int': 42,
    'str': 'foo',
    'bool': True
})

s({'int': 27, 'str': 'bar', 'bool': False})  # valid

I don't personally think it's necessary (or even feasible) for this method to provide a comprehensive way of inferring every feature already supported by the Schema class. It seems the scenario @svisser had in mind was the case of enums (["male", "female", "other"]); I am personally fine with a limitation of this method being that you can't infer those, or ranges, or callables, or any number of other things. But it does provide a very easy way of, for example, quickly scaffolding a suite of API regression tests.

I think it's fine to merge this, but if you could add something like the following to the docstring: "Note: only very basic inference is supported".

@alecthomas I added the note that you suggested. I thought about being a bit more verbose and explaining more, but hopefully the example in the docstring with the note together will do most of the work to manage expectations.

Coverage increased (+0.06%) to 95.379% when pulling b710aa0 on dtao:dtao/add-schema-infer-method into 95489bd on alecthomas:master.

Yeah, that's great, thanks :)