std.math: add lerp

Author: zooster

lib/std/math.zig

@@ -1312,6 +1312,33 @@ test "lossyCast" {
     try testing.expect(lossyCast(u32, @as(f32, maxInt(u32))) == maxInt(u32));
 }
 
+/// Performs linear interpolation between *a* and *b* based on *t*.
+/// *t* must be in range 0.0 to 1.0. T must be a float type.
+///
+/// This does not guarantee returning *b* if *t* is 1 due to floating-point errors.
+/// This is monotonic.
+pub fn lerp(comptime T: type, a: T, b: T, t: T) T {
+    if (@typeInfo(T) != .Float and @typeInfo(T) != .ComptimeFloat)
+        @compileError("T must be a float type");
+
+    assert(t >= 0 and t <= 1);
+    return @mulAdd(T, b - a, t, a);
+}
+
+test "lerp" {
+    try testing.expectEqual(@as(f16, 75), lerp(f16, 50, 100, 0.5));
+    try testing.expectEqual(@as(f32, 43.75), lerp(f32, 50, 25, 0.25));
+    try testing.expectEqual(@as(f64, -31.25), lerp(f64, -50, 25, 0.25));
+    try testing.expectApproxEqRel(@as(f80, -7.16067345e+03), lerp(f80, -10000.12345, -5000.12345, 0.56789), 1e-19);
+    try testing.expectApproxEqRel(@as(f128, 7.010987590521e+62), lerp(f128, 0.123456789e-64, 0.123456789e64, 0.56789), 1e-33);
+
+    // this highlights the precision
+    try testing.expectEqual(@as(f32, 0.0), lerp(f32, 1.0e8, 1.0, 1.0));
+    try testing.expectEqual(@as(f16, 0.0), lerp(f16, 1.0e4, 1.0, 1.0));
+    try testing.expectEqual(@as(f32, 1.0), lerp(f32, 1.0e7, 1.0, 1.0));
+    try testing.expectEqual(@as(f16, 1.0), lerp(f16, 1.0e3, 1.0, 1.0));
+}
+
 /// Returns the maximum value of integer type T.
 pub fn maxInt(comptime T: type) comptime_int {
     const info = @typeInfo(T);