Implements the Constrained Majorization graph layout algorithm (deprecated). More...

#include <cola.h>

Collaboration diagram for cola::ConstrainedMajorizationLayout:

Public Member Functions
	ConstrainedMajorizationLayout (vpsc::Rectangles &rs, std::vector< Edge > const &es, RootCluster clusterHierarchy, const double idealLength, EdgeLengths eLengths=StandardEdgeLengths, TestConvergence doneTest=nullptr, PreIteration *preIteration=nullptr, bool useNeighbourStress=false)
	Constructs a constrained majorization layout instance. More...

void	setConstraints (cola::CompoundConstraints *ccs)
	Specify a set of compound constraints to apply to the layout. More...

void	setUnsatisfiableConstraintInfo (UnsatisfiableConstraintInfos unsatisfiableX, UnsatisfiableConstraintInfos unsatisfiableY)
	Register to receive information about unsatisfiable constraints. More...

void	setStickyNodes (const double stickyWeight, std::valarray< double > const &startX, std::valarray< double > const &startY)

void	setScaling (bool scaling)

void	setExternalSolver (bool externalSolver)

void	setAvoidOverlaps (bool horizontal=false)

void	setNonOverlappingClusters ()

void	setStraightenEdges (std::vector< straightener::Edge > straightenEdges, double bendWeight=0.01, double potBendWeight=0.1, bool xSkipping=true)

void	moveBoundingBoxes ()

void	run (bool x=true, bool y=true)
	Implements the main layout loop, taking descent steps until stress is no-longer significantly reduced. More...

void	runOnce (bool x=true, bool y=true)
	Same as run(), but only applies one iteration. More...

Detailed Description

Implements the Constrained Majorization graph layout algorithm (deprecated).

The optimisation method used is "stress majorization", where a sequence of quadratic functions which strictly bound the stress from above, is solved to monotonically reduce the stress (by iteratively changing the configuration of nodes).

Once the problem has been set up, call run() or runOnce() to run the layout algorithm.

Note: We recommend the use of ConstrainedFDLayout instead of this class. ConstrainedFDLayout tends to produce better results and has more features. We are no longer working on ConstrainedMajorizationLayout.

Constructor & Destructor Documentation

◆ ConstrainedMajorizationLayout()

cola::ConstrainedMajorizationLayout::ConstrainedMajorizationLayout	(	vpsc::Rectangles &	rs,
		std::vector< Edge > const &	es,
		RootCluster *	clusterHierarchy,
		const double	idealLength,
		EdgeLengths	eLengths = `StandardEdgeLengths`,
		TestConvergence *	doneTest = `nullptr`,
		PreIteration *	preIteration = `nullptr`,
		bool	useNeighbourStress = `false`
	)

Constructs a constrained majorization layout instance.

Parameters

[in]	rs	Bounding boxes of nodes at their initial positions.
[in]	es	Simple pair edges, giving indices of the start and end nodes in rs.
[in]	clusterHierarchy	A pointer to a RootCluster object defining the cluster hierarchy (optional).
[in]	idealLength	Aa scalar modifier of ideal edge lengths in eLengths.
[in]	eLengths	Individual ideal lengths for edges. The actual ideal length used for the ith edge is idealLength*eLengths[i], or if eLengths is empty then just idealLength is used (i.e., eLengths[i] is assumed to be 1).
[in]	done	A test of convergence operation called at the end of each iteration (optional).
[in]	preIteration	An operation called before each iteration (optional).

Member Function Documentation

◆ moveBoundingBoxes()

void cola::ConstrainedMajorizationLayout::moveBoundingBoxes ( )

inline

Update position of bounding boxes.

◆ run()

void cola::ConstrainedMajorizationLayout::run	(	bool	x = `true`,
		bool	y = `true`
	)

Implements the main layout loop, taking descent steps until stress is no-longer significantly reduced.

Parameters

[in]	x	If true, layout will be performed in x-dimension (default: true).
[in]	y	If true, layout will be performed in y-dimension (default: true).

References vpsc::HORIZONTAL, setScaling(), and vpsc::VERTICAL.

Here is the call graph for this function:

◆ runOnce()

void cola::ConstrainedMajorizationLayout::runOnce	(	bool	x = `true`,
		bool	y = `true`
	)

Same as run(), but only applies one iteration.

This may be useful here it's too hard to implement a call-back (e.g., in java apps).

Parameters

[in]	x	If true, layout will be performed in x-dimension (default: true).
[in]	y	If true, layout will be performed in y-dimension (default: true).

References vpsc::HORIZONTAL, setScaling(), and vpsc::VERTICAL.

Here is the call graph for this function:

◆ setAvoidOverlaps()

void cola::ConstrainedMajorizationLayout::setAvoidOverlaps ( bool horizontal = false )

inline

At each iteration of layout, generate constraints to avoid overlaps. If bool horizontal is true, all overlaps will be resolved horizontally, otherwise some overlaps will be left to be resolved vertically where doing so leads to less displacement

◆ setConstraints()

void cola::ConstrainedMajorizationLayout::setConstraints ( cola::CompoundConstraints * ccs )

inline

Specify a set of compound constraints to apply to the layout.

Parameters

[in] ccs The compound constraints.

◆ setExternalSolver()

void cola::ConstrainedMajorizationLayout::setExternalSolver ( bool externalSolver )

inline

Says that the Mosek optimisation library should be used to solve the quadratic programs rather than the libvpsc solver.

◆ setNonOverlappingClusters()

void cola::ConstrainedMajorizationLayout::setNonOverlappingClusters ( )

inline

Add constraints to prevent clusters overlapping.

◆ setScaling()

void cola::ConstrainedMajorizationLayout::setScaling ( bool scaling )

inline

Scaling speeds up the solver by conditioning the quadratic terms matrix.

Referenced by run(), and runOnce().

Here is the caller graph for this function:

◆ setStickyNodes()

void cola::ConstrainedMajorizationLayout::setStickyNodes	(	const double	stickyWeight,
		std::valarray< double > const &	startX,
		std::valarray< double > const &	startY
	)

Sticky nodes causes nodes to spring back to (startX,startY) when unconstrained.

◆ setStraightenEdges()

void cola::ConstrainedMajorizationLayout::setStraightenEdges	(	std::vector< straightener::Edge >	straightenEdges,
		double	bendWeight = `0.01`,
		double	potBendWeight = `0.1`,
		bool	xSkipping = `true`
	)

inline

For the specified edges (with routings), generate dummy vars and constraints to try and straighten them. bendWeight controls how hard we try to straighten existing bends potBendWeight controls how much we try to keep straight edges straight

◆ setUnsatisfiableConstraintInfo()

void cola::ConstrainedMajorizationLayout::setUnsatisfiableConstraintInfo	(	UnsatisfiableConstraintInfos *	unsatisfiableX,
		UnsatisfiableConstraintInfos *	unsatisfiableY
	)

inline

Register to receive information about unsatisfiable constraints.

In the case of unsatisifiable constraints, the solver will drop unsatisfiable constraints of lowest priority. Information about these will be written to these lists after each iteration of constrained layout.

param[out] unsatisfiableX A pointer to an UnsatisfiableConstraintInfos object that will be used to record unsatisfiable constraints in the x-dimension. param[out] unsatisfiableY A pointer to an UnsatisfiableConstraintInfos object that will be used to record unsatisfiable constraints in the y-dimension.

The documentation for this class was generated from the following files:

libcola/cola.h
libcola/cola.cpp

Public Member Functions

Detailed Description

Constructor & Destructor Documentation

◆ ConstrainedMajorizationLayout()

Member Function Documentation

◆ moveBoundingBoxes()

◆ run()

◆ runOnce()

◆ setAvoidOverlaps()

◆ setConstraints()

◆ setExternalSolver()

◆ setNonOverlappingClusters()

◆ setScaling()

◆ setStickyNodes()

◆ setStraightenEdges()

◆ setUnsatisfiableConstraintInfo()