uber / h3-go / 40544249803

100%

master: 100%

Pull #178

github

h3: clamp the haversine term so antipodal distances are not NaN

For some antipodal pairs rounding pushes the haversine term just above
its mathematical maximum of 1, and sqrt(1 - h) is then NaN, so
GreatCircleDistanceRads and the kilometre and metre forms returned NaN
instead of pi times the radius. Clamping the term at 1 affects only that
overshoot. This is upstream uber/h3#1267. The conformance suite is
unaffected: its distance sampler already redraws far targets within 1e-3
radians of antipodal, where the haversine is ill-conditioned.

The test helper assertRelClose compared with >, and every comparison
against NaN is false, so a NaN result passed; it now rejects NaN, and the
new antipodal case fails without the clamp.

Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>

4 of 4 new or added lines in 1 file covered. (100.0%)

3673 of 3673 relevant lines covered (100.0%)

5166667.48 hits per line