Merge lp://qastaging/~haggai-eran/nux/rtl-rebased into lp://qastaging/nux/2.0

Haggai

Selecting right to left has to be done per layout. Lets say you have an horizontal layout that has a checkbox followed by a button, and the button is followed by a combobox. In LTR order you get this:
 checkbox | button | combobox

If you make the HLayout to be RTL then the content will look like this:

 combobox| button | checkbox

This should work out fine because only a specific HLayout has been been changed to RTL.

So you have to select only the layouts of the interface that need to be inverted. Rather than having SetDefaultDirection in Nux.cpp, it should be a function member of LinearLayout. Then for each layout (HLayout or VLayout), you will be able to decide if it goes from LTR or RTL. Individual views like the Checkbox, will remain un-affected.

How about that?

Jay,

I agree that we should be able to set directionality on per-layout basis. I even implemented it as a member of Area. However, I think that the default when using an RTL locale should be right to left, for all widgets, and not just specific HLayouts.

In the example you described, you could show the detailed widgets like this (in LTR):

[ combobox ↓ | button | [x] checkbox ]

Then, I expect the RTL rendering to be like this:

[ checkbox [x] | button | ↓ combobox ]

The internal layout of comboboxes and checkboxes should also be mirrored.

Please compare how GTK looks in LTR and RTL in the following two screenshots of the widget factory:
http://i44.tinypic.com/2lks0e8.jpg
http://i43.tinypic.com/25ri0ds.png

Haggai

Ok, I see how it is. I think many widget will have to have support for a RTL design.
But first we can get this branch in since it has global support for RTL at the layout level. Then the next step would be to go to each individual widgets that requires it and review the design.

It is up to me now to review and adapt this branch and see that we can merge it. We now have the requirement of proposing tests along side major features. Can you think of some way we can test the features of this branch?

Hi,

I think that perhaps rewriting the code to pack in the same order as in LTR might still have some advantages. I've tried to think on how that would work. Instead of walking the list of HLayout children in reverse, and packing them to the left, I'll walk the list in the usual order, and pack from right to left.

As far as I can see, I would only need to change HLayout::ComputeContentSize, and HLayout::ComputeContentPosition, which would make the changes simpler. I'm sure in both cases other widget will need to be modified, but if any widget's layout depends on the order of children's placement, then that order wouldn't change.

Regarding tests, I suppose that a tests for HLayout can be modified to test the RTL case as well, but I cannot find such a test (perhaps my tree isn't updated). I imagine a test that creates an HLayout, with several children that are using different settings for scaling, vertical placement, etc. and asserting there placement. What do you think?