From: Phlip <phlip_cpp@...>
Date: 2004-09-01T12:15:23+09:00
Subject: Re: widget modifs after calls to Tk.mainloop

Nicolas Decoster wrote:

> I guess this is not possible, since for Tcl/Tk the event loop is started
> when the shell is started while for irb it is started only after _full_
> widgets settings.

Following up HN's answer...

The most majorly awesome way to do this is out of a unit test. If you write
GUI code thru unit tests, you can display a window in a tested state, by
temporarily operating mainloop().

class  TestGrapher < Test::Unit::TestCase

  def setup()
   @canvas = TkCanvas.new()
   @canvas.configure('width', 600)
   @canvas.configure('height', 600)
   @canvas.grid()
   return @canvas
  end

        def teardown()
            @canvas.destroy()  if @canvas
        end

  def maybeMainloop(reveal = false)
   if reveal then
    Tk.mainloop()
                Tk.restart()
                @canvas = nil
   end
  end
....
  def test_twoNodes()
   svg = generateSvg(
    'digraph aGraph { a [label=""]; b [label=""]; }')

   putSvgIntoCanvas(svg, @canvas)
   all = @canvas.find_all()
   assert_equal(all.size(), 2)
   assert_equal(all[0].class.name(), 'TkcOval')
   assert_equal(all[1].class.name(), 'TkcOval')
   maybeMainloop # true
  end
....
end

That's not complete - it illustrates a test that creates a canvas,
optionally displays it, exits mainloop, restarts, and optionally destroys
the canvas.

Decomment maybeMainloop # true to see a canvas. (Also take out the stuff
that doesn't work - it's too complex to copy in here.)

Using this technique, you can put one of HN's or your experiments into a
test case.

Writing unit tests to write features is just like running IRB, except much
better. All your features leave a trail of tests, and running them over and
over again keeps your bug rate low.

-- 
  Phlip
  http://industrialxp.org/community/bin/view/Main/TestFirstUserInterfaces