Kiwi is an implementation of the algorithm based on the seminal Cassowary paper.
It is not a refactoring of the original C++ solver. Kiwi has been designed from the ground up to be lightweight and fast. Kiwi ranges from 10x to 500x faster than the original Cassowary solver with typical use cases gaining a 40x improvement. Memory savings are consistently > 5x.
Edit: Anyone interested in modern implementations and participating in the community of library implementers and users can check this website: http://overconstrained.io/ :-)
Re:Cassowary – A Constraint Solving Toolkit
by saosebastiao on Tuesday July 07, 2015 @09:04AM
(#9845123)
I'm a huge fan of constraint programming, but everything I've read about layout problems makes me think that LP or MIP is a better fit for the domain. Solving systems of linear equalities and inequalities is what LP solvers do, and they've got decades of research into how to do it extremely fast. Most commercial solvers can solve systems with tens of millions of constraints in mere seconds.
Is there a reason why cassowary took the constraint solver approach other than the fact that it is called constraint-based layout?
Re:Cassowary – A Constraint Solving Toolkit
by ridiculous_fish on Tuesday July 07, 2015 @01:54PM
(#9846992)
(I worked on autolayout at Apple)
Cassowary is LP - it's just the Simplex method. Constraints become equalities and inequalities in the tableau, and the objective function is constructed from non-required ("best effort") constraints.
The novel contribution from Cassowary is its incrementalism technique. It can add, remove, and edit constraints without having to re-solve the whole LP problem over again. So it's well suited for interactive applications.
An example is dragging. You establish a constraint that sets the position of the dragged object, and then modify that constraint during the drag. Cassowary can re-use the old computation to solve the new one very efficiently.
Re:Cassowary – A Constraint Solving Toolkit
by junkor on Wednesday July 08, 2015 @03:10AM
(#9850222)
@ridiculous_fish:
Hi, I just found that you are worked on autolayout, I have a little question about autolayout.
There is a simple view as follow, a subView with some Constraints, they are all UILayoutPriorityRequired,when superview frame change, system will break the subView's height Constraint(300), but the top||bottom Constraint(30). they are all UILayoutPriorityRequired, why the system did things like this, and how does AutoLayout work (what's the calculate order? superView->subView->subviews'subview?),and where can I find the docs about this? :)
|-----------------------------|
| | |
| 30 |
| | |
|-60-[subView(height:300)]-60-|
| | |
| 30 |
| | |
|-----------------------------|
(maybe this case is not much suitable,because I should remove or change some constraints, I just want to use this case to explain the question)
Re:Cassowary – A Constraint Solving Toolkit
by junkor on Wednesday July 08, 2015 @03:14AM
(#9850236)
shit, I use characters draw the constraints, but the Comment removed the \n characters, Add the "\n" by yourself:\n
|-----------------------------|\n
| | |\n
| 30 |\n
| | |\n
|-60-[subView(height:300)]-60-|\n
| | |\n
| 30 |\n
| | |\n
|-----------------------------|\n
Re:Cassowary – A Constraint Solving Toolkit
by saosebastiao on Tuesday July 07, 2015 @02:23PM
(#9847179)
That's interesting! Is it using Lagrangian duality to do any sort of shadow pricing for incremental changes? I guess I should probably read the paper by now.
Re:Cassowary – A Constraint Solving Toolkit
by Quequau on Tuesday July 07, 2015 @04:19AM
(#9844183)
I'm not really sure why we're talking about this now, as it's been some time since it's seen a major update and there are other more active projects now.
In any case, I think the title is a bit vague and should more complete stated; such as:
Cassowary - an incremental constraint solving toolkit for linear equation systems.
Kiwi is an implementation of the algorithm based on the seminal Cassowary paper. It is not a refactoring of the original C++ solver. Kiwi has been designed from the ground up to be lightweight and fast. Kiwi ranges from 10x to 500x faster than the original Cassowary solver with typical use cases gaining a 40x improvement. Memory savings are consistently > 5x.
https://github.com/AbletonAG/aqt-cassowary
Edit: Anyone interested in modern implementations and participating in the community of library implementers and users can check this website: http://overconstrained.io/ :-)
Is there a reason why cassowary took the constraint solver approach other than the fact that it is called constraint-based layout?
Cassowary is LP - it's just the Simplex method. Constraints become equalities and inequalities in the tableau, and the objective function is constructed from non-required ("best effort") constraints.
The novel contribution from Cassowary is its incrementalism technique. It can add, remove, and edit constraints without having to re-solve the whole LP problem over again. So it's well suited for interactive applications.
An example is dragging. You establish a constraint that sets the position of the dragged object, and then modify that constraint during the drag. Cassowary can re-use the old computation to solve the new one very efficiently.
|-----------------------------| | | | | 30 | | | | |-60-[subView(height:300)]-60-| | | | | 30 | | | | |-----------------------------| (maybe this case is not much suitable,because I should remove or change some constraints, I just want to use this case to explain the question)
In any case, I think the title is a bit vague and should more complete stated; such as:
Cassowary - an incremental constraint solving toolkit for linear equation systems.