Add support for <BLANKLINE> keyword

[skycaptain]

Python doctest is using the keyword <BLANKLINE> for declaring empty lines within a doctest result. Using such a keyword helps the doctest parser finding connected blocks of tests.

See

• https://docs.python.org/3.7/library/doctest.html#how-are-docstring-examples-recognized
• Allow >>> Python style doctests #236 (comment)

[cbm755]

This would only matter with -NORMALIZE_WHITESPACE---by default, we match any whitespace.

But I don't see any problem with some minimal support for these, e.g., replacing them with empty strings in want before comparing to got.

[skycaptain]

Right, I thought of this to be a helper for the regex searching and seperating the doctests. In the end <BLANKLINE> gets replaced by a newline character.

[catch22]

I don't think we've needed this one so far (and we default to normalizing whitespace since there are differences between Octave and Matlab IIRC). Is this solving any problem that arises when using Sphinx?

[skycaptain]

Right. Rethought this. Matlab prints out lots of blank lines, so legability would definitly suffer from such a keyword. But, without having looked at the code; when do you decide what line is still part of the doctest output and what part is of the helptext. Like in this test:
https://github.com/catch22/octave-doctest/blob/master/test/test_whitespace.m
Is it the two empty lines?

[catch22]

That's right ("When the m-file contains plaintext documentation, doctest finds tests by searching for lines that begin with >>. It then finds the expected output by searching for the next >> or two blank lines." in https://octave.sourceforge.io/doctest/function/doctest.html).