We come across a number of edge case when parsing pieces of RST, so we do have custom logics.
Currently we rerun papyri on all of numpy/scipy and make sure it does not crash.
A better test harness would have a number of example snippets, and expected AST tree, and the test suite should compare that we are not regressing.
We should also have an easy to reach command line when the examples are wrong, and need to be updated.
We come across a number of edge case when parsing pieces of RST, so we do have custom logics. Currently we rerun papyri on all of numpy/scipy and make sure it does not crash.
A better test harness would have a number of example snippets, and expected AST tree, and the test suite should compare that we are not regressing.
We should also have an easy to reach command line when the examples are wrong, and need to be updated.