How should correct test answers be defined?

[AshlinHarris]

Generally, function outputs for PreprocessMD.jlare DataFrames.

For doctests, the expected output is simply the text displayed on the REPL when a DataFrame is printed. So, for this type of test, two strings are compared. In the unlikely event that DataFrames.jl changes how DataFrames are printed, the tests break, which might be the bestapproach.

In contrast, a few integration tests in test/runtests.jl compare outputs to the result of a DataFrame definition. For this type of test, a DataFrame is compared to a DataFrame. But most tests in this file compare strings, not DataFrames.

I prefer to compare DataFrames directly for integration tests, but most would need to be rewritten for this.