mparray 0.2.0

MPArray is an Array API Standard compatible array library that features arbitrary precision arithmetic.

Install with pip:

pip install mparray

Import to access the array namespace:

import mparray as xp

Use any feature defined by the 2025.12 version of the standard:

x = xp.arange(5)
# MPArray([0, 1, 2, 3, 4], dtype=int64)

Integer arrays are filled with Python ints; consequently, elements never overflow.

x = x[:2] + 10**50
# MPArray([100000000000000000000000000000000000000000000000000,
#          100000000000000000000000000000000000000000000000001],
#         dtype=int64)

Dtypes follow array API promotion rules for compatibility, but do not limit the underlying representation. For instance, int32 arrays also use Python ints.

xp.astype(x, xp.int32)
# MPArray([100000000000000000000000000000000000000000000000000,
#          100000000000000000000000000000000000000000000000001],
#         dtype=int32)

Real and complex floating point arrays are backed by mpmath.

from mpmath import mp
mp.dps = 55  # set the desired precision
y = xp.astype(x, xp.float64)
# MPArray([mpf('100000000000000000000000000000000000000000000000000.0'),
#          mpf('100000000000000000000000000000000000000000000000001.0')],
#         dtype=float64)
xp.exp(y)
# MPArray([mpf('4.535356657536074105363661926352788592352627616652483210606e+43429448190325182765112891891660508229439700580366'),
#          mpf('1.232837758776106335937289338644047193491770865081040599056e+43429448190325182765112891891660508229439700580367')],
#         dtype=float64)

Arbitrary-precision equivalents of some SciPy special functions are available:

xp.special.ndtr(-y)
# MPArray([mpf('8.998129248551223738202088675072137970394711995443364311445e-2171472409516259138255644594583025411471985029018332830572230087781144236110599295949020894164292471'),
#          mpf('6.461085845597172042837958170074610008600868697345677551181e-2171472409516259138255644594583025411471985029018376260020420755455996799425617236333573042128793971')],
#         dtype=float64)