Increase test coverage

[tobixen]

There are still code sections that aren't exercised through the tests

[tobixen]

A coverage check was done prior to the 0.7-release, the stuff below remains, and a new check should be done prior to the 1.0-release.

Things that I won't fix - at least not before 0.7:

• hacks for python2-support is of course not covered when running tests with py3
• some weird corner cases reported in patch to support ctags and etags, and to allow response.reason to be optional to support Debian 9 requests library #81 and Necessary for iCloud.. #56 is not exercised by the test (will maybe help to fix Caldav test servers #45)
• other obscure error situations where an exception will be raised
• elements.ical
• DAVClient.mkcol
• CalendarObjectResource._create isn't 100% covered, but I've moved this to a separate issue Clean up CalendarObjectResource._create #96.

Things that should be looked into:

• DAVObject.canonical_url is not used anywhere, and can probably be removed, abandoned code? Should probably wait until 1.0 to preserve backward compatibility.
• _handle_prop_response: some weird corner cases not exercised

[tobixen]

For the CalendarObjectResource classes, I find this in the coverage report:

• Some stuff in the header varies dependent on python version and wasn't covered when I ran tests on one specific version. Ignorable.
• set_dtend has a line missing coverage. Ignorable, as I haven't decided what to do with this method.
• scheduling needs better coverage
  • add_organizer, not tested ... raised in event.add_organizer needs some TLC #524
  • add_attendee, not tested lots of corner cases not tested
  • change_attendee_status, some exception handling
  • decline_invite and tentatively_accept_invite
  • _reply_to_invite_request - some corner cases missing coverage
• error loggings and lines raising unexpected Exceptions ... I'm ignoring those as for now.
• split_expanded have a timezone line that is not tested - but perhaps it will go away when running tests towards all servers?
• occurrence.add("RECURRENCE-ID", (...)) - is this line needed? I've thrown in an error._assert(RECURRENCE-ID" in occurrence) - if there are no reports of mussing recurrence-id within some years, those lines may be deleted.
• set_relation(reverse=True) is not tested
• get_end should be tested for objects that has neither duration nor end. Is it appropriate to return None? Such an object should be defined to have zero duration? decided to procrastinate this.
• event.load(), return self.load_by_multiget() is not tested - as the test server that has this issue only offers RO-access. Perhaps make a unit test? decided to ignore this as for now
• if "Schedule-Tag" in r.headers: in load()
• load_by_multiget
• _find_id_path - all corner cases should be checked - but first it should be introduced to be run on load()
• get_self in self.save - some of it is executed when running tests towards all test servers, but there are corner cases that aren't covered.
• event.data=icalendar_object should be tested
• Exceptions for dealing with empty/unloaded events are ignored
• get_duration for one-day-events, ingore as for now
• No DTSTART in Todo._next ... ingored
• No STATUS in task to be completed ... can be ignored
• Counting logic in task._complete_recurring_thisandfuture for when a recurrence marked with "thisandfuture" has a RRULE set. Is that even valid according to the RFC? It should be checked up and eventually this code should be exercised by tests.
• task._is_pending should be tested better. May be elevated to a public method. Should be moved to icalendar library.
• task.uncomplete needs test
• task.set_duration is not well-enough covered
• task.set_due is not well enough covered

I think checking this list suffices for 2.0.

[tobixen]

There are quite some missing checkboxes above, I procrastinate until 2.1 to deal with the rest.

After 2.1, we should take out a complete coverage report once more and deal with the file with most not-covered statements.