Closed johanley closed 2 years ago
Expanding the list of stars, to filter only by mag and not geometry, has no effect.
Could it be bad data, in the arrays that define the constellation lines? Reducing the ChartUtils.EDGE_OVERLAP_DEGREES setting from 60 to 58 removes the problem!
Reducing ChartUtils.EDGE_OVERLAP_DEGREES works for the present case, but it's a workaround that doesn't address the core bug.
The behavior seems to be consistent with the south celestial pole moving between a pair of stars connected by a constellation-line. In that case, the pair will have wildly different right ascensions. The data in that case is correct, but the algo is defective.
starchart-south-pole-on-a-line.pdf Look at the south pole for -2796. It's exactly on a line joining two stars in Hydrus. That's the explanation.
The code now abandons any lines that would cross the whole chart if drawn. (Constellations.java)
Latitude 46.0, longitude -63.0. Visible in this range of years: -2755..-2835 (approx). Straight line moves across the map from 4 o'clock to 10 o'clock.