Subject: Re: Regression Testing? [was Re: Fwd: Re: Release Schedule]
From: Mike Nordell (tamlin@algonet.se)
Date: Tue Jun 05 2001 - 08:14:54 CDT
Speaking of testing...
What if we (finally) separated the document from the UI and allowed
automated tests do what they should do: testing?
This (of course) connects to my long standing idea that we should create a
*clean* MVC architecture, where the Views are only subscribers to
update-events. Using that, it would be quite easy to create programs to
"automatically" test all but display and UI interaction.
For the sake of argument, let's assume only keyboard input right now:
Imagine we have an architecture where input events are handled by an input
event handler. That one in turn decides wether this is only a cursor moving
event or an "edit" event. If it is an Edit event it is forwarded to the
Document class which is supposed to take care of modification to its
*internal data* (what it choose to display to the View(s) might very well be
a completely different view of that data).
In that world, it would be a piece of cake to create new test cases to
verify at least the document engine, since no one but the document manager
(the Document class) is indeed allowed to touch its data.
I could write an essay about his, but I think this is enough to make people
thinging about "what is and object" and "why are we allowed to touch other
objects 'private parts'".
It introduces one (IMO good) limitation: Changes are done/redone on the
document. If one view of a doc "Do" a change, another view of that Doc will
be able to "Undo" that change, since the changes are properties of the
Document.
What do you say, is it perhaps time for this overhaul (if not, imagine doing
this "tomorrow")?
/Mike
This archive was generated by hypermail 2b25 : Tue Jun 05 2001 - 08:14:37 CDT