New directory walking should ignore hidden folders

[peterjc]

I was looking forward to the new directory parsing functionality added in v1.2 to close #38, however it does not work well if you have hidden folders like .tox present. I would expect .venv would be another common problem case.

$ rst-lint --version
1.2.1

$ rst-lint --level warning .
SEVERE ./.tox/style/lib/python2.7/site-packages/restructuredtext_lint/test/test_files/invalid_line_mismatch.rst:1 Title overline & underline mismatch.
WARNING ./.tox/style/lib/python2.7/site-packages/restructuredtext_lint/test/test_files/invalid.rst:2 Title underline too short.
WARNING ./.tox/style/lib/python2.7/site-packages/restructuredtext_lint/test/test_files/second_short_heading.rst:6 Title underline too short.
WARNING ./.tox/style/lib/python2.7/site-packages/restructuredtext_lint/test/test_files/second_short_heading.rst:6 Title underline too short.
ERROR ./.tox/style/lib/python2.7/site-packages/restructuredtext_lint/test/test_files/invalid_link.rst:None Anonymous hyperlink mismatch: 1 references but 0 targets.
See "backrefs" attribute for IDs.
ERROR ./.tox/style/lib/python2.7/site-packages/restructuredtext_lint/test/test_files/invalid_target.rst:1 Unknown target name: "fork me on github <https://github.com/evvers/git-pre-commit-hook".

I would suggest that the recursion ignore hidden folders by default (leading .), and possibly also any hidden RST files too?

[twolfson]

flake8 seems to recurse into . folders. I think the expected usage is to use a glob for files and specify folders to recurse (e.g. flake8 *.rst restructuredtext_lint/) =/

$ flake8 .
./.test/foo.py:1:10: E999 SyntaxError: invalid syntax

There is an option for --exclude which does ignore .tox. I'm happy to accept another PR for this behavior:

  --exclude=patterns    Comma-separated list of files or directories to
                        exclude. (Default:
                        .svn,CVS,.bzr,.hg,.git,__pycache__,.tox,.eggs,*.egg)

[peterjc]

I was starting to explore how flake8 did this, the --exclude=patterns approach seems worth using here too 👍