Raise specific error classes instead of a generic NotmuchError with an
magic status value (e.g. NotmuchError(STATUS.NULL_POINTER) ->
NullPointerError()), update the documentation accordingly.
Signed-off-by: Justus Winter <4winter@informatik.uni-hamburg.de>
Put each libnotmuch function reference right in front of the
corresponding python wrapper.
Signed-off-by: Justus Winter <4winter@informatik.uni-hamburg.de>
Rename Database.__del__ to Database.close, move it just below the open
function and call close() in a newly created destructor just below the
constructor.
Signed-off-by: Justus Winter <4winter@informatik.uni-hamburg.de>
Before 3434d1940 the return values of libnotmuch functions were
declared as c_void_p and the code checking for errors compared the
returned value to None, which is the ctypes equivalent of a NULL
pointer.
But said commit wrapped all the data types in python classes and the
semantic changed in a subtle way. If a function returns NULL, the
wrapped python value is falsish, but no longer equal to None.
Backported from master to 0.11.
Before 3434d1940 the return values of libnotmuch functions were
declared as c_void_p and the code checking for errors compared the
returned value to None, which is the ctypes equivalent of a NULL
pointer.
But said commit wrapped all the data types in python classes and the
semantic changed in a subtle way. If a function returns NULL, the
wrapped python value is falsish, but no longer equal to None.
All strings are unicode strings in python 3 and the basestring and
unicode types are removed hence the need for a specialized version.
Signed-off-by: Justus Winter <4winter@informatik.uni-hamburg.de>
Recent changes introduced lots of unicodification of strings, mostly in
the form of .decode('utf-8', errors='ignore'). However, python 2.5 does
not like the errors keyword argument and complains. It does work when
used as a simple arg though, so that's what this patch does.
Signed-off-by: Sebastian Spaeth <Sebastian@SSpaeth.de>
Formerly Message.get_replies() returned an iterator or None forcing
users to check the result before iterating over it leading to strange
looking code at the call site.
Fix this flaw by adding an EmptyMessagesResult class that behaves like
the Messages class but immediatly raises StopIteration if used as an
iterator and returning objects of this type from Message.get_replies()
to indicate that there are no replies.
Since 2b01161191, Message.__str__ doesn't
construct a hash containing the thread data before
constructing the formatstring. This changes the formatstring
to accept positional parameters instead of a hash.
Now that types are checked correctly, we also need to make sure that all the
arguments actually are instances of these types. Otherwise the function calls
will fail and raise an exception similar to this one:
ctypes.ArgumentError: argument 3: <type 'exceptions.TypeError'>: expected
LP_LP_NotmuchMessageS instance instead of pointer to c_void_p
We were not returning anything at all, which does not match the API
documentation. Fixed. Thanks to Patrick Totzke for the heads up.
Signed-off-by: Sebastian Spaeth <Sebastian@SSpaeth.de>
Add type information to the ctypes._FuncPtr wrappers and
use the wrapper classes instead of c_void_p for pointers
to notmuch_*_t.
This enables the ctypes library to type check parameters
being handed to functions from the notmuch library.
Signed-off-by: Justus Winter <4winter@informatik.uni-hamburg.de>
