scifloat

Alternative to conventional floats used for scientific measurements.

pip install scifloat

Quick start

from scifloat.scifloat import scifloat

s = round(scifloat(1.2421, 0.0513))
print(s)          # 1.24 ± 0.06

a = round(scifloat(1.36551315, [0.0302, 0.00891]))
print(a)  # 1.37 +0.04 | -0.01

Nothing is rounded until you call round() — scifloat(x, dx) just stores the raw value and error(s). x and dx can be ints, floats, or strings.

The two rounding rules

1. Errors round up, to 2 significant digits if the leading digit is 1 or 2, otherwise to 1 (0.0111 → 0.012, 0.0513 → 0.06).
2. The value rounds to the error's precision, using classic round-half-to-even (1.35 → 1.4, 1.25 → 1.2).

For an asymmetric error, the coarser (fewer decimal places) side sets the shared precision — pass match_precision=False to keep each side at its own precision instead.

Arithmetic

+ - * / between two scifloats propagate the errors in quadrature:

x = scifloat(10.5, 0.3)
y = scifloat(2.0, 0.1)
print(round(x + y))  # 12.5 ± 0.4
print(round(x / y))  # 5.2 ± 0.4

LaTeX export

s.to_latex()
# '$\\num[parse-numbers=false]{1.24 \\pm 0.06}$'

s.to_latex(symbol="x_obs", unit=r"\meter")
# '$x_{\\mathrm{obs}} = \\SI[parse-numbers=false]{1.24 \\pm 0.06}{\\meter}$'

More

See the tutorial notebook for a full walkthrough of both rounding rules, asymmetric errors, and every to_latex() option.