uber / h3-go / 26835325248

100%

master: 100%

Pull #121

github

feat: reimplement greatCircle* as pure Go

Eliminate CGo overhead by implementing the haversine formula directly
in Go. Benchstat shows ~66% faster execution and zero allocations.

Before (CGo):

```lang=bash
BenchmarkGreatCircleDistanceRads-16    	19489219	        62.82 ns/op	      32 B/op	       2 allocs/op
BenchmarkGreatCircleDistanceKm-16      	19358064	        62.81 ns/op	      32 B/op	       2 allocs/op
BenchmarkGreatCircleDistanceM-16       	18969169	        63.68 ns/op	      32 B/op	       2 allocs/op
```

After (pure Go):

```lang=bash
BenchmarkGreatCircleDistanceRads-16    	58551271	        20.70 ns/op	       0 B/op	       0 allocs/op
BenchmarkGreatCircleDistanceKm-16      	57732436	        21.12 ns/op	       0 B/op	       0 allocs/op
BenchmarkGreatCircleDistanceM-16       	56447564	        21.64 ns/op	       0 B/op	       0 allocs/op
```

Benchstat:

```lang=bash
                           │   old (CGo)   │         new (pure Go)          │
                           │    sec/op     │   sec/op     vs base           │
GreatCircleDistanceRads-16   62.82n ± 2%   20.70n ± 1%  -67.06% (p=0.000 n=10)
GreatCircleDistanceKm-16     62.81n ± 1%   21.12n ± 2%  -66.37% (p=0.000 n=10)
GreatCircleDistanceM-16      63.68n ± 0%   21.64n ± 2%  -66.01% (p=0.000 n=10)
geomean                      63.10n        21.15n       -66.48%

                           │  old (CGo)  │       new (pure Go)        │
                           │    B/op     │   B/op     vs base         │
GreatCircleDistanceRads-16   32.00 ± 0%   0.00 ± 0%  -100.00% (p=0.000 n=10)
GreatCircleDistanceKm-16     32.00 ± 0%   0.00 ± 0%  -100.00% (p=0.000 n=10)
GreatCircleDistanceM-16      32.00 ± 0%   0.00 ± 0%  -100.00% (p=0.000 n=10)

                           │  old (CGo)  │       new (pure Go)        │
                           │  allocs/op  │ allocs/op   vs base        │
GreatCircleDistanceRads-16   2.000 ± 0%   0.000 ± 0%  -100.00% (p=0.000 n=10)
GreatCircleDistanceKm-16     2.000 ± 0%   ... (continued)

12 of 12 new or added lines in 1 file covered. (100.0%)

871 of 871 relevant lines covered (100.0%)

284.03 hits per line