Sorry for a late reply - but thanks for a very useful answer. Got one more question though.

How would you technically moc QFileDialog.exec_ to avoid that the unit test is halted by the Dialog popping up in its own event loop. Anyway, this is how I do it:

class Proxy:
    def __init__(self):
        self._called = False

    def caller(self, *args):
        self._called = args

    def called(self):
        return self._called

proxy = Proxy()
new_exec = proxy.exec_
old_exec = QtGui.QFileDialog.exec_
QtGui.QFileDialog.exec_ = new_exec

... Then do testing and query proxy.called() to test that the QFileDialog was called

Finally clean up and reinsert the old exec_ method.

QtGui.QFileDialog.exec_ = new.instancemethod(old_exec, None, QtGui.QFileDialog)

Makes sense?

On 07/03/2012 18:13, Andreas Pakulat wrote:
On 07.03.12 15:42:34, Mads Ipsen wrote:
OK, I agree, the snippets I provided were not nice, and I apologize
for that.

But suppose that the constructor in your class starts a thread that
generates say an icon for a QStandardItem. Then you want to make
sure that the thread has finished before you test that the icon has
been created.

* What would be the proper way of unit testing this?
I'd try to put in some moc-objects and split the testing. So one
unit-test would just test the loading of the icon from whatever source
it comes, without a thread in a blocking way.

Another unit-test would put in a moc-object that immediately returns the
icon so you can be sure that the amount of time the thread needs to run
is relatively small. Then in the unit-test, create the widget and then
let the main thread sleep for a bit so the background thread finishes.
Afterwards verify the moc-icon is properly retrieved. You don't need an
event loop here yet, unless you want the widget to be actually drawn.

* How should starting/stopping the event loop be handled?
Hmm, the Qt unit-test framework works by starting the event-loop and
triggering the unit-test functions via a timer. I guess this could also
be done in python with the unittest framework somehow.

Personally I'd probably avoid testing the actual event-loop-requiring
bits of the application via unit-tests completely. Instead I'd go for a
gui-test-tool to test those parts and use unit-tests for all the
business-logic and non-gui stuff.

Andreas

_______________________________________________
PyQt mailing list    PyQt@riverbankcomputing.com
http://www.riverbankcomputing.com/mailman/listinfo/pyqt


--
+-----------------------------------------------------+
| Mads Ipsen                                          |
+----------------------+------------------------------+
| Gåsebæksvej 7, 4. tv |                              |
| DK-2500 Valby        | phone:          +45-29716388 |
| Denmark              | email:  mads.ip...@gmail.com |
+----------------------+------------------------------+


_______________________________________________
PyQt mailing list    PyQt@riverbankcomputing.com
http://www.riverbankcomputing.com/mailman/listinfo/pyqt

Reply via email to