Sarracenia __init__ unit test

[gcglinton]

This test is pretty exhaustive, and probably overkill for the needs, but it's got 90% code coverage so that's pretty good.

There's some odd things that weren't able to provide coverage for though:

• Message.dumps():
  • Couldn't generate anything that wasn't printable, and thus fail the try on line 451
• Message.fromFileInfo():
  • It generates a new msg at every run, so couldn't get it to do the rename bits on lines 532-536. How sure how those are supposed to work
  • Messages never have the new_retrievePath key, so the if o.rename block on lines 543-547 throws a key error if it runs (there's a test case for that though)
  • Couldn't get the try on line 555 to fail no matter how many directories were striped, so the except never runs
• Message.updatePaths():
  • The new_dir and new_file parameters are optional, but their concatenation on line 686 throws an error if they're not set (there's a test case for this too)
  • With new_dir effectively being mandatory, the ifs on lines 719 and 735 don't really work properly because they can never evaluate to True
• "Extras" checking
  • It doesn't work because no way was found to emulate the packages being in a different state than they actually are on the system

Interesting bits/lessons learned when writing these tests:

• Mocked some internal modules/methods/attributes in order to "trick" the code to trigger a certain behaviour
  • Lines 150 and 162 are replacing random.choice's return with None
  • Line 409 is faking being Windows for the following lines
  • Line 471 is faking urllib.request.urlopen's return as a static string, bypassing the need to actually do an HTTP request
• Some method calls throw errors, and Pytest catches them properly (lines 317, 351, 443)
  • Could be good for doing doing input validation and such

[petersilva]

more tests always better than no tests... this is great! if there is something wrong, we can just fix it later...