pintrs 0.1.5

Fast Rust-based Python extension for physical unit manipulation

pintrs

A fast, Rust-powered drop-in replacement for pint -- the Python physical units library.

pintrs reimplements pint's core unit registry and quantity system in Rust via PyO3, giving you the same Python API with significantly better performance.

Installation

pip install pintrs

Quick start

from pintrs import UnitRegistry

ureg = UnitRegistry()

# Create quantities
distance = ureg.Quantity(5.0, "kilometer")
time = ureg.Quantity(2.0, "hour")

# Arithmetic with automatic unit tracking
speed = distance / time
print(speed)           # 2.5 kilometer / hour
print(speed.to("m/s")) # 0.6944... meter / second

# Attribute-style access
print(ureg.meter)      # 1 meter
print(ureg.speed_of_light)

Performance

pintrs is 10-90x faster than pint on common operations. Benchmarks on Python 3.13 (lower is better):

Operation	pintrs	pint	Speedup
Quantity creation	0.35 us	3.64 us	10x
Parse string ("9.81 m/s**2")	0.74 us	70.61 us	96x
Conversion (km -> m)	1.19 us	8.08 us	7x
Conversion (km/h -> m/s)	1.68 us	14.11 us	8x
Addition (compatible units)	0.94 us	12.66 us	13x
Multiply by scalar	0.13 us	5.59 us	41x
Multiply quantities	0.17 us	5.38 us	31x
Parse units ("kg * m / s ** 2")	0.95 us	23.66 us	25x
String formatting	0.29 us	8.58 us	29x

Run python examples/benchmark.py to reproduce (install pint for comparison).

Features

• Drop-in replacement for pint's UnitRegistry, Quantity, Unit, and common operations
• NumPy support via ArrayQuantity with full ufunc integration
• Type-safe with full .pyi stubs for mypy and pyright in strict mode
• Measurement support for quantities with uncertainty propagation
• Compatibility stubs for Context, Group, System so existing code doesn't break

NumPy integration

import numpy as np
from pintrs import UnitRegistry
from pintrs.numpy_support import ArrayQuantity

ureg = UnitRegistry()

distances = ArrayQuantity(np.array([1.0, 2.0, 3.0]), "kilometer", ureg)
result = distances.to("meter")
print(result.magnitude)  # [1000. 2000. 3000.]

# NumPy ufuncs work transparently
print(np.sqrt(ArrayQuantity(np.array([4.0, 9.0]), "m**2", ureg)))

Measurements (uncertainty)

from pintrs import Measurement, Quantity

m = Measurement(Quantity(100.0, "meter"), 0.5)
print(m)       # 100.0 +/- 0.5 meter
print(m.rel)   # 0.005

# Error propagation (adds in quadrature)
m2 = Measurement(Quantity(50.0, "meter"), 0.3)
print(m + m2)

Decorators

from pintrs import UnitRegistry, wraps, check

ureg = UnitRegistry()

@wraps(ureg, ret="meter/second", args=("meter", "second"))
def speed(distance, time):
    return distance / time

result = speed(ureg.Quantity(100, "km"), ureg.Quantity(2, "hour"))
print(result)  # in m/s

@check(ureg, "[length]", "[time]")
def velocity(d, t):
    return d / t

Custom units

ureg = UnitRegistry()
ureg.define("smoot = 1.7018 * meter")
print(ureg.Quantity(1, "smoot").to("meter"))  # 1.7018 meter

Compatibility with pint

pintrs targets API compatibility with pint's most-used features:

• UnitRegistry, Quantity, Unit with full arithmetic
• Unit parsing, conversion, base/root/compact/reduced/preferred units
• __getattr__ on registry (ureg.meter, ureg.speed_of_light)
• Serialization via __reduce__ (pickle) and to_tuple/from_tuple
• wraps and check decorators
• Measurement with uncertainty propagation
• Context/Group/System stubs (API-compatible, no-op)

Not yet implemented: full context-based conversions (spectroscopy, etc.), Babel/locale formatting, logarithmic units, pandas ExtensionArray integration.

Development

# Build (requires Rust toolchain and maturin)
maturin develop --release

# Lint and format
ruff check --fix . && ruff format .

# Type check
mypy python/pintrs/
pyright python/pintrs/

# Test
pytest

License

Apache-2.0