Hi Hodac,
dao <dao.ho...@gmail.com> wrote on 03/15/2010 06:35:39 AM:
> I am a little bit confused with those coord conversions, etc... I
> suppose this is a piece of cake for you guys. Could you tell me what's
wrong?
There were three significant errors. The first is that your code
assumed that the same local translate could be applied for each of
the children. This doesn't work you need to calculate the local
translate matrix for each child separately. I would suggest remembering
the start point in beginMove, then in 'move' you need to current
point so each "rect" can calculate it's transform.
Second you need to be careful because as you change the translate
on the element the elements 'screenCTM' changes as well. To get
it to work you will either need to get the screenCTM of the parent
of the element you want to move (an approach I often use) and then
'sandwich' the translate above the local transform (see next issue)
or store the screenCTM at the start of the move for use during each
move update.
Third you used translate.concatenate but you needed to use
translate.preConcatenate. The way your code currently is you
want the resultant matrix to be the same as translating a point
by 'translate' and then by the pre-existing transform (remember
transforms are from child coordiate system to parent). This
is a little tricky however because if you use the parent's
getScreenCTM to calculate the translate then translate.concatenate
would be the correct method to use (you want the point to be
mapped by the elements local transform, then translated in
the parent's coordinate system).
I should also comment that what you are attempting is the most
complex version of this. Typically you have a fair amount of
control over the documents you need to interactively manipulate
(especially complex manipulations like this). So I will typically
simplify things so I don't need to deal with all the complexities
you are dealing with. For example I'll ensure that the element
I will be manipulating the transform on has no transform initially.
If I need to move multiple objects I'll move them into a
group and move the group (which means that all the objects I'll
need to move can't have parents with different transforms).
> I can see that my solution does not work (as Thomas said) but the
> code of Jonathan does not work neither[...].
That code was good but it's not the complete answer for really
complex use cases.
