Precision issues

[grimsa]

As floating point numbers are used to represent coordinates, there are precision issues (having cumulative effects).

For example, if we are calculating the location of a point that divides an edge by a previously calculated ratio, sometimes it might produce a point that lies slightly off the edge. In once case around 4% difference in expected and actual areas of subpolygon was observed.

Ideas for improvement:

• review order of arithmetic actions. Maybe some calculations could also be simplified to improve precision.
• for each point produced, track the edge it belongs to and use that in further calculations.
• Alternatively, snap produced points to the edge it belonged to.
• Or make small adjustments once each subpolygon is produced.