Any reason not to use an f-string here?

This file was based on Lib/test/test_json/test_encode_basestring_ascii.py, so I'd left it the same.

But I've now updated the Lib/test/test_json/ files to use f-strings.

You can simply use constant.upper() instead of expected.

I generally prefer to have explicit expected results and avoid computing them:

https://testing.googleblog.com/2014/07/testing-on-toilet-dont-put-logic-in.html

But I can change it if you prefer.

That's a good reason to do keep expected.

To be consistent in avoiding computations in tests we should write this more explicitly:

        self.assertEqual(self.loads("Infinity", parse_constant=str.upper),
                         "INFINITY")
        self.assertEqual(self.loads("-Infinity", parse_constant=str.upper),
                         "-INFINITY")
        self.assertEqual(self.loads("NaN", parse_constant=str.upper), "NAN")

If you want to allow some computations, you can write it as:

        for name in "Infinity", "-Infinity", "NaN":
            self.assertEqual(self.loads(name, parse_constant=str.upper),
                             name.upper())

which is also more breaf and simple that the current code.

I prefer the existing version:

• it shows the explicit inputs and outputs
• avoids extra computations
• we can immediately see we're asserting the function under test in the exact same way each time

That's a good reason to do keep expected.