Sorry, no good website really. There are still some quirks with the test finder (and name matcher) as well. But there are some new features. From the README:
In addition to running doctest tests, DataTest modifies doctest to add
a few small changes to avoid false negatives when comparing output
with the expected output.
Because doctest doesn't allow empty lines in a test (it interprets
the empty line as a separation between tests) you can instead use a
single "." in place of a blank line.
Tracebacks can be identified with just Traceback:, you do not
have to give the full Traceback (most recent call last):.
If the expected output begins with <, then the output and
expected output are compared using htmldiff, which allows for
softer comparisons. Whitespace is normalized before comparison,
attributes are compared without respect to order, and ... can be
used as a wildcard in many places. The full details are in the
docstrings for htmldiff
Floats are compared for equality, not just equality of string
representation, so you can use 1.4 instead of
1.3999999999999999.
These all make the doctest examples a little less accurate, but they
make them easier to write without going into every detail.
Utilities
DataTest includes a couple of utilities to make debugging you code
easier. These are contained in the datatest.util package.
datatest.util.dictdiff compares two dictionaries and if they don't
match it displays a per-key comparison. You can provide various
wildcards in one of the dictionaries (the "pattern", i.e., the
expected output). See the doc strings for more.
datatest.util.htmldiff compares two HTML fragments. It ignores
whitespace and the order of tags, in addition to allowing various
wildcards. See the doc strings for more.