Ugh... Qt4 porting

Mosfet dan.duley at
Mon May 16 14:40:18 BST 2005

Lars wrote:
>> I hate to say it, but now I am beginning to feel that TrollTech is changing
>> things just to be difficult :P
>Sure. That's all we ever wanted to achieve with Qt 4.

Well, that does seem to be the general consensus, (sorry, that was low ;-).

>> Trivial changes include no longer using convertDepth() but a format enum.
> How is it harder to write "image.convertToFormat(QImage::Format_ARGB32) 
> than "image.convertDepth(32)"? 

Note I said "trivial". That's not a big deal, I'm just listing the changes 

>> There is no longer a create method either, the only way I can see to
>> allocate an empty image is to assign it from another one. Not ideal, but I
>> can deal with it.
>QImage(width, height, QImage::Format_ARGB32) gives you an empty image with
>the dimensions and format specified. What more do you need?

Uhm, that is not an empty image. An empty image is a null one. You supply 
constructors to create NULL images but no way to actually allocate the data 
except for assignment.

Excuse me but this seems rather silly. Doing a create method is only a couple 
lines of code to allocate the private data.

NULL images are used when you don't know what format the image is going to be 
in, create it in it's own scope, then modify or return it outside the scope.

>First of all it's still there as a Qt3Support method. We removed it because 
>the jump table is a rather strange concept for people who can't multiply and  
>add (so you can write scanline = jumpTable()[i] instead of scanline = data()  
>+ i*bytePerLine()). It doesn't give you a whole lot and wastes some memory  
>for every image. 

Yes, you can do pointer arithmetic instead but it's very nice to also be able 
to randomly access pixels w/ x and y array indexes. Like a lot of stuff that 
has been removed it's not strictly necessary, but useful. I'll ignore the 
"can't multiply or add" part :P

A jumptable shouldn't waste memory unless if you create one.

More information about the kde-core-devel mailing list