Spatial Projects SQL (spatialSQL)
for MapInfo Professional®

Buffer Issues

MapInfo Professional objects > buffer provides a fundamental user and developer GIS function. There are a number of issues in current versions (all logged with MapInfo, 29 Apr 05). The following applies to version 6.5 unless noted otherwise.

Issue	Comment	Example
1. Numeric errors in computation.	(12 May 05) Issue withdrawn - not reproducible
2. Resolution (smoothness)	Buffer a point with a resolution of 3 and you get a 3 point region. Buffer a region with a resolution of 3 and resulting polygon arcs are smoothed to approximately 20 points per 90° (smoothness of 101?). A minimum smoothness seems like a fair approach to avoid unwanted errors caused by too much approximation in a buffer shape, however there appears to be inconsistancy in the way objects are treated. If a user is buffering a very detailed object (a coastline), smoothness of 3 may be appropriate and would certainly reduce the size of the buffer significantly. Allowing the user to choose smoothness with warnings in documentation might be a better approach.
3. Buffer Width Distance - spherical / cartesian	MapInfo insists that if your current map projection is unprojected (lat/long), a spherical computation is the only valid one. It would seem that the choice of spherical / cartesian should be based on the size of the buffer. If the difference between the buffer distance and the equivalent spherical distance is small , a cartesian calculation would be valid regardless of the chosen projection.
4.Buffer generation errors.	Certain object shapes will result in erroneous buffers in the form of missing polygons. These have been tested with MI 6.5 and 7.8 with the same results. The shapes involve multiple adjacent islands or holes. The input data is available here. Smoothness has no impact on the result.	Buffer = -4 km.  Buffer = 4 km MapInfo buffer shown in red "Correct" buffer shown in green
5. Spherical distance buffers for large areas	Because buffer distance is calculated at a single point for use at all points on the target feature, the value can be wrong for large areas.	Reference 1: Spherical distance buffers   http://lists.directionsmag.com/discussion/read.php?f=7&i=44451&t=44451 Reference 2: Spheroid formulae. Errors expected in using cartesian assumptions on the sphere (not quite the issue here but the commentary and formulae are useful). http://lists.directionsmag.com/discussion/read.php?f=7&i=40416&t=40389 The approximation is valid for small areas but at what size area would you start to expect problems?