I'm saying that the first window in tab order is the lowest window in
the Z order among all its sibling windows. A window later in the tab
order is drawn above one earlier in the tab order.
In general, though, Z order is irrelevant to child windows and tab order
is irrelevant to overlapped or popup windows, so the equivalence is
mostly just a curiosity.
I'm not sure I understand what you're saying you've proven. If I put the
mouse over the OK button in the Run dialog and print
mouse.window.control, it says "OK". (Text is the default property for a
Control object, so I left it off.) If I print mouse.window.control.next,
it says "Cancel" which is indeed the window you get to if you tab from
OK. Control.Next returns the next control in the tab order, as I'd
expect it to.
On 12/15/2010 9:25 AM, Doug Lee wrote:
On your last point, are you saying that tab order goes in reverse
window tree order by default, i.e., up Z order as you tab? I'm sure
it does not go the same direction - in other words, I have proven that
if window A has window B as next window in the winAPI, by default
window B may be reached by Shift+Tab from window A but not by Tab.
On Wed, Dec 15, 2010 at 08:57:08AM -0500, Ron Parker wrote:
On 12/15/2010 8:43 AM, Doug Lee wrote:
Subject pretty much says it all: What order are the windows in the
Children and DirectChildren returned in? Possible orders I can think
of for DirectChildren would be Z order (same as with the WinAPI calls
for get next/previous window) and tab order. Children could be in
some sort of screen-based logical order, tab order, recursive Z order,
or depth-first or breadth-first tree order variants of those. Of
course, either could also be officially undefined, but I bet they're
not.
Officially, yes, the order is undefined.
Unofficially - which is to say, this is subject to change in any
version, so if you count on it you should verify it for each new version
of Window-Eyes - it's in reading order, left to right, top to bottom, by
the upper left corners of the windows.
As if that weren't useless enough, the "Children" tree is breadth-first,
mainly because that turns out to be marginally easier to implement
without recursion. (For breadth-first you just push the elements onto
the back of a queue. For depth-first you have to push them onto a stack,
but in reverse order. Whoever wrote this code - I think it was me -
apparently didn't want to mess with enumerating them in reverse order
for something that wasn't officially going to be documented anyway.)
By the way, Z order and tab order are the same thing. Strange but true.
