Cassowary is an incremental constraint solving toolkit that efficiently solves systems of linear equalities and inequalities. Constraints may be either requirements or preferences. Client code specifies the constraints to be maintained, and the solver updates the constrained variables to have values that satisfy the constraints.
Cassowary was developed by Greg Badros, Alan Borning, and Peter J. Stuckey, and was optimized for user interface applications.  Badros used Cassowary among others for implementing Constraint Cascading Style Sheets (CCSS), an extension to Cascading Style Sheets (CSS). CCSS adds support for layout constraints. These allow designers to describe the layout of a web page in a more flexible manner. Cassowary is used to solve these problems and calculate the final layout.
Smalltalk , C ++, and Java versions are available in the main distribution. Furthermore, there are bindings for GNU Guile , Python , and STk.
- As of September 2014, GSS, a layout language and layout engine for the web, is being developed by The Grid.  It is a modern extension of CCSS. 
- Scwm , the Scheme Constraints Window Manager.
- As of 2011, Cassowary is being used in the layout engine for Mac OS X (Lion and greater) and iOS (6 and greater).
- Enaml’s layout engine is built on top of the Cassowary linear constraint optimizer.    
- Jump up^ The Cassowary linear arithmetic constraint solving algorithm, Authors: Greg J. Badros, Alan Borning & Peter J. Stuckey
- Jump up^ https://github.com/slightlyoff/cassowary.js/
- Jump up^ “Grid Style Sheets development history” .
- Jump up^ “About CCSS” .
- Jump up^ Casuarius provides Python bindings to a C ++ implementation of the Cassowary algorithm.
- Jump up^ casuariusonGitHub
- Jump up^ Kiwi, a replacement for Casuarius / Cassowary. The new solver removes that bottleneck. It still uses the same Cassowary algorithm, but it’s a scratch implementation of the algorithm based on the Cassowary paper, not the existing Cassowary source code. While the Cassowary algorithm is good, the existing C ++ implementation has lots of inefficiencies. The new solver is anywhere between 12x and 500x faster depending on the problem (40x typical) and uses around 5x less memory. And as I typically do, I thing to write the Python bindings by hand in C ++ rather than use Cython. The resulting code is faster, and we now have less external dependency.
- Jump up^ kiwionGitHub