Trait Decimal128Encode

Source

pub trait Decimal128Encode {
    // Required method
    fn try_to_i128_mantissa(&self, target_scale: u32) -> Option<i128>;
}

Expand description

Plug-in trait for converting a decimal value into its i128 mantissa rescaled to a target scale.

Implementers MUST use round-half-to-even (banker’s rounding) on scale-down so the bytes the derive emits match polars’s own str_to_dec128 path. A None return surfaces as a polars ComputeError from the generated code.

The codegen invokes this method through UFCS on the selected trait path, so inherent methods with the same name cannot bypass this contract. Custom backends (bigdecimal::BigDecimal, arbitrary-precision types, …) provide their own impls; this crate ships a rust_decimal::Decimal impl below.

Required Methods§

Source

fn try_to_i128_mantissa(&self, target_scale: u32) -> Option<i128>

Returns the mantissa as i128 after rescaling self to target_scale, or None if the conversion would overflow or otherwise violate the schema. Implementations MUST round half-to-even on scale-down.

Implementations on Foreign Types§

Source §

impl Decimal128Encode for Decimal

Available on crate feature rust_decimal only.

Reference Decimal128Encode impl for rust_decimal::Decimal.

Banker’s-rounding contract: round-half-to-even on scale-down, checked_mul overflow-to-None on scale-up. This impl is verified against polars’s str_to_dec128 on a battery of inputs covering half-tie boundaries (positive and negative), large magnitudes, and scale-up overflow by this repository’s df-derive-core integration tests.

Source §