No, try to provoke the test with faulty code in README.md

?? I am sorry, I genuinely have no idea what you are referring to.

Can you please add a description to the PR and explain its purpose?

There should be no >>> at the start of the code in the readme.

Please ignore for now. Please read up on doctest

@tschm, I know how doctest works and have written doctest plugins for pytest myself.

Please don't assume lack of knowledge but provide an explanation or description what you are intending to do here.

Regarding doctest: the >>> are used to indicate code in a docstring. The docstring is then displayed without the >>>. Now you inserted >>> in the readme, which gets displayed including the >>>, so turns perfectly fine code into code with syntax error once users want to copy it.

I am not 100% sure on those >>>. I use them to indicate a doctest executable python command where as rows without them indicate the expected output (tested sign for sign). Yes, they make copy some inconvenient, but better than faulty code in a README...

Yes, they make copy some inconvenient

The entire point of the readme is to provide the most conveniently possible entry point to potential users. So I would strongly oppose any change that reduces user convenience substantially for a little bit of developer convenience.

Speaking about developer convenience, it is easy to get the python blocks. Just grep anything between triple-backtick-python and triple-backtick lines.
No need for a complicated abuse of doctest.

do I understand correctly, you want to test code in the README for correctness? (that it runs)

That would be a very useful CI step!

Though I would do it differently, and that avoids doing the business with the >>>:

• grep code blocks as the lines between triplebacktick-python and triplebacktick
• execute them in the test env

Done

Yes, I want to test the code in the README. So the >>> have an advantage

>>> 2+2
4

A few comments about this block. The "4" is the expected result. Replacing "4" with "5" should break the test.
Also note the empty line below 4 and in front of the close triplebacktick. If you remove it the triplebacktick becomes part of the expectation! Of course, if we remove the 4 and the >>> we test whether 2+2 runs. That's better than no test but not as powerful as testing for expected output