math is scalar math on libm, with ValueError: math domain error for domain errors. Import it by bare name, both runtimes resolve it with no manifest. To pin a different version, import it by URL or through a packages.json alias, see Modules.

Module constants are pi, e, tau, inf, nan (values, not calls). Integer helpers are factorial, gcd, lcm, isqrt, comb, perm, bounded by the VM’s 128-bit integers. hypot and gcd are variadic. modf and frexp return tuples, and floor, ceil, and trunc return int. The rest of the scalar surface, by group:

• Power and log: sqrt, cbrt, exp, exp2, expm1, pow, log (optional base), log2, log10, log1p.
• Trigonometric: sin, cos, tan, asin, acos, atan, atan2, hypot, dist.
• Hyperbolic: sinh, cosh, tanh, asinh, acosh, atanh.
• Angular and special: degrees, radians, erf, erfc, gamma, lgamma.
• Float ops and classification: fabs, fmod, remainder, copysign, ldexp, modf, frexp, floor, ceil, trunc, isnan, isinf, isfinite.
• Reductions: fsum, prod (returns float, even for integer inputs).

There is no complex-number surface. A batch path processes a whole bytes buffer of little-endian f64 values in one call, and pairs with struct for numeric work. Element-wise: sqrt_all, abs_all, exp_all, log_all, sin_all, cos_all. Same-length pairs: add_all, sub_all, mul_all. Scalar broadcast: scale_all(buf, k). Reductions to a float: fsum_all, prod_all, dot_all. Row-major matrix times vector: matvec(m, x, cols). A buffer length that is not a multiple of 8, or mismatched operand lengths, raises ValueError.

1.4142135623730951
3.141592653589793
13.0
120

The integer helpers return int:

6
12
9
10 20

modf and frexp decompose a float into tuples, and floor, ceil, and trunc return int:

(0.75, 3.0)
(0.75, 4)
2 3 -2

The batch functions work on a bytes buffer of little-endian f64 values. Pack it with struct:

6.0
[10.0, 20.0, 30.0]
14.0