always-decimal

A tiny Python package to safely convert floats, strings, and numbers into Decimal objects.
It solves common headaches when comparing PostgreSQL numeric values with Python floats by ensuring consistent Decimal values.

✨ Features

• 🔒 Exact conversion from float → Decimal (no hidden rounding).
• ⚖️ Configurable coercion: set scale, rounding mode, normalize trailing zeros.
• 🧩 Works with float, int, str, or Decimal.
• 🛡️ Raises clear DecimalCoercionError on invalid inputs.
• 🐍 Compatible with Python 3.10+.
• 📦 Easy install via uv or pip.

📦 Installation

Using uv (recommended)

uv add always-decimal

Using pip

pip install always-decimal

For local development:

uv pip install -e ".[dev]"
# or
pip install -e ".[dev]"

🚀 Usage

from always_decimal import ensure_decimal, to_decimal_exact

# 1. Exact conversion: no rounding, no scale changes
d1 = to_decimal_exact(0.1)
print(d1)
# Decimal('0.1000000000000000055511151231257827021181583404541015625')

# 2. Safe coercion with fixed scale
price = ensure_decimal(19.995, scale=2)
print(price)
# Decimal('20.00')  (ROUND_HALF_EVEN default)

# 3. String input with quantization
val = ensure_decimal("1.2345", scale=3)
print(val)
# Decimal('1.235')

# 4. Normalizing (remove trailing zeros)
norm = ensure_decimal("1.2300", scale=4, normalize=True)
print(norm)
# Decimal('1.23')

⚙️ API

to_decimal_exact(value: float | Decimal) -> Decimal

Convert a float or Decimal to a Decimal exactly:

• float → Decimal.from_float(value)
• Decimal → returned as-is Raises TypeError for other types.

ensure_decimal(value, *, scale=None, rounding=ROUND_HALF_EVEN, clamp_exp=True, normalize=False) -> Decimal

Convert and coerce input to Decimal.

• scale: number of fractional digits to quantize (e.g., 2 → cents).
• rounding: rounding mode (default: bankers rounding).
• clamp_exp: clamp exponents to context limits.
• normalize: trim trailing zeros.

🧪 Running Tests

make test

or directly:

pytest

📄 License

MIT

🙌 Contributing

Issues and PRs are welcome! Please format with ruff, type-check with mypy, and run pytest before submitting.