Hilabeling:

Our algorithms

simple[weighted/ stable]( corresponds to the algorithm Greedy in our paper)
mis [weighted/ stable]
kamis [unweighted/unstable] (see http://algo2.iti.kit.edu/kamis/)
maxHS [weighted/ stable]

(add more algorithms read pal/ReadMe_gpl)

to discuss

weight of label l of feature f: InactiveCost(f) - cost(l) [may not proper]
In qgis, point features normally have only one label candidate (8 positions for the regions) [may define our preferences ourselves]

further improvements

reimplementation of pal (directed in qgis with qgsfeature and qgslabling)c implements modification and update in problem (Now modification emits new layout and builds new problem)

Mey help for experiments

For labeling, qgis uses the QgsLabelingEngine class https://qgis.org/api/classQgsLabelingEngine.html#details
The eigine uses function run()

• It builds the labeling problem p in pal (exactProblem).
  
• The problem p then provides a labling using the searchMethod set in its eiginesetting (the seaarchMethod is implemented in * pal/problem.cpp) .
  
• The labeling is drawn in the end