precision to steel decimal crate implemented

2025-07-07 00:31:13 +02:00
parent 314a957922
commit 4d5d22d0c2
4 changed files with 566 additions and 37 deletions
--- a/steel_decimal/docs/rust_decimal.txt
+++ b/steel_decimal/docs/rust_decimal.txt
@@ -0,0 +1,366 @@
+# rust_decimal for Financial Applications: Complete Guide
+
+rust_decimal provides a 128-bit fixed-precision decimal implementation designed specifically for financial calculations, eliminating floating-point rounding errors that plague traditional financial software. With a 96-bit mantissa and support for up to 28 decimal places, it offers the exact precision required for accounting and monetary calculations while maintaining performance suitable for high-throughput financial systems.
+
+## Input handling best practices
+
+### String parsing and validation patterns
+
+rust_decimal provides multiple parsing methods optimized for different input scenarios. The **most robust approach for financial applications** uses `from_str_exact()` for strict validation combined with comprehensive error handling:
+
+```rust
+use rust_decimal::{Decimal, Error};
+
+fn parse_financial_amount(input: &str) -> Result<Decimal, AmountError> {
+    let trimmed = input.trim();
+    
+    // Pre-validation checks
+    if trimmed.is_empty() {
+        return Err(AmountError::EmptyInput);
+    }
+    
+    if trimmed.len() > 50 {
+        return Err(AmountError::InputTooLong);
+    }
+    
+    // Use from_str_exact for strict parsing
+    Decimal::from_str_exact(trimmed)
+        .map_err(|e| match e {
+            Error::InvalidOperation => AmountError::InvalidFormat,
+            Error::Underflow => AmountError::Underflow,
+            Error::Overflow => AmountError::Overflow,
+            _ => AmountError::ParseError,
+        })
+}
+```
+
+The library supports **multiple input formats automatically**: standard decimal notation (`"123.45"`), scientific notation via `from_scientific("2.512e1")`, and different radix bases through `from_str_radix()`. For **compile-time optimization**, use the `dec!()` macro which parses literals at compile time with zero runtime cost.
+
+### Precision and currency considerations
+
+Different financial contexts require specific precision strategies. **Standard recommendations** include 2-4 decimal places for retail/e-commerce, 4-6 for forex trading, 8-18 for cryptocurrency, and 4-6 for GAAP accounting compliance. The library's maximum scale of 28 decimal places accommodates even the most demanding financial calculations.
+
+**Currency-specific validation patterns** should enforce appropriate ranges and scales:
+
+```rust
+pub struct ValidatedAmount(Decimal);
+
+impl ValidatedAmount {
+    pub fn new(value: Decimal) -> Result<Self, FinancialError> {
+        // Range validation
+        if value < Decimal::MIN || value > Decimal::MAX {
+            return Err(FinancialError::OutOfRange);
+        }
+        
+        // Precision validation for currency context
+        if value.scale() > 28 {
+            return Err(FinancialError::ScaleExceeded);
+        }
+        
+        Ok(ValidatedAmount(value))
+    }
+}
+```
+
+### Handling different input formats
+
+For **production systems processing various input formats**, implement a unified parsing strategy that handles integers, decimals, and scientific notation:
+
+```rust
+// Automatic conversion from integers
+let amount = Decimal::from(12345_i64);  // 12345
+
+// Float conversion with precision control
+let price = Decimal::from_f64(123.45).unwrap();
+
+// Scientific notation parsing
+let large_amount = Decimal::from_scientific("1.23e6").unwrap();  // 1230000
+
+// String parsing with validation
+let user_input = "99.99";
+let parsed = Decimal::from_str(user_input)?;
+```
+
+## Output formatting best practices
+
+### Display and precision control
+
+rust_decimal provides multiple formatting approaches optimized for different financial contexts. The **standard approach** uses the `Display` trait for human-readable output, while **precision-controlled formatting** uses `round_dp()` for specific decimal places:
+
+```rust
+let amount = dec!(123.456789);
+
+// Standard string representation
+let display = amount.to_string();  // "123.456789"
+
+// Precision-controlled output
+let currency_format = amount.round_dp(2).to_string();  // "123.46"
+
+// Scientific notation for large numbers
+let scientific = format!("{:e}", amount);  // "1.23456789e2"
+```
+
+### Rounding strategies for accounting
+
+The library implements **comprehensive rounding strategies** including banker's rounding (IEEE 754 compliant) which eliminates systematic bias in large datasets:
+
+```rust
+use rust_decimal::RoundingStrategy;
+
+let tax = dec!(3.4395);
+
+// Banker's rounding (default) - preferred for financial compliance
+let rounded = tax.round_dp(2);  // Uses MidpointNearestEven
+
+// Explicit rounding strategies
+let away_from_zero = tax.round_dp_with_strategy(2, RoundingStrategy::MidpointAwayFromZero);
+let truncated = tax.round_dp_with_strategy(2, RoundingStrategy::ToZero);
+```
+
+### Currency formatting and localization
+
+For **multi-currency applications**, implement currency-aware formatting that maintains precision requirements:
+
+```rust
+#[derive(Debug, Clone)]
+pub struct Money {
+    amount: Decimal,
+    currency: Currency,
+}
+
+impl Money {
+    pub fn format_for_display(&self, precision: u32) -> String {
+        match self.currency {
+            Currency::USD => format!("${}", self.amount.round_dp(precision)),
+            Currency::EUR => format!("€{}", self.amount.round_dp(precision)),
+            Currency::BTC => format!("₿{}", self.amount.round_dp(8)),
+        }
+    }
+}
+```
+
+## Robust conversion patterns
+
+### String-to-Decimal-to-String pipeline
+
+The **most efficient conversion pipeline** for financial applications uses compile-time optimization where possible and validated parsing for runtime inputs:
+
+```rust
+// Compile-time optimization for known values
+const COMMISSION_RATE: Decimal = dec!(0.0025);
+const TAX_RATE: Decimal = dec!(0.15);
+
+// Runtime parsing with validation
+fn process_transaction(amount_str: &str) -> Result<String, ProcessingError> {
+    let amount = parse_financial_amount(amount_str)?;
+    let commission = amount * COMMISSION_RATE;
+    let total = amount + commission;
+    
+    Ok(total.round_dp(2).to_string())
+}
+```
+
+### Edge case handling
+
+**Production-ready edge case handling** requires comprehensive validation and error recovery:
+
+```rust
+pub trait SafeDecimalOps {
+    fn safe_add(&self, other: Self) -> Result<Self, FinancialError>;
+    fn safe_multiply(&self, other: Self) -> Result<Self, FinancialError>;
+    fn safe_divide(&self, other: Self) -> Result<Self, FinancialError>;
+}
+
+impl SafeDecimalOps for Decimal {
+    fn safe_add(&self, other: Self) -> Result<Self, FinancialError> {
+        self.checked_add(other).ok_or(FinancialError::Overflow)
+    }
+    
+    fn safe_divide(&self, other: Self) -> Result<Self, FinancialError> {
+        if other.is_zero() {
+            return Err(FinancialError::DivisionByZero);
+        }
+        self.checked_div(other).ok_or(FinancialError::Overflow)
+    }
+}
+```
+
+### Performance considerations
+
+For **high-frequency financial calculations**, rust_decimal offers significant advantages over floating-point arithmetic despite being 2-6x slower. **Key performance characteristics** include 10-20ns for addition/subtraction, 50-100ns for multiplication, and 100-200ns for division. The library uses **stack allocation** (16 bytes per Decimal) and provides **zero-cost abstractions** through compile-time macros.
+
+**Memory optimization strategies** include using the `Copy` trait for efficient stack-based operations, implementing batch processing patterns, and pre-allocating constants:
+
+```rust
+// Efficient batch processing
+fn calculate_portfolio_value(positions: &[Position]) -> Decimal {
+    positions.iter()
+        .map(|pos| pos.quantity * pos.average_price)
+        .sum() // Decimal implements Sum trait
+}
+```
+
+## Financial-specific features
+
+### Scale and precision handling
+
+rust_decimal's **128-bit architecture** provides optimal balance between precision and performance for financial applications. The **96-bit mantissa** supports approximately 28-29 significant digits, while the **32-bit metadata** handles scale (0-28) and sign information.
+
+**Database integration patterns** vary by storage backend:
+- **PostgreSQL**: Use `NUMERIC(19,4)` for standard applications, `NUMERIC(28,8)` for high precision
+- **MySQL**: Use `DECIMAL(13,4)` for general use, `DECIMAL(19,4)` for large amounts
+- **GAAP compliance**: Often requires 4-6 decimal places with specific rounding rules
+
+### Monetary arithmetic best practices
+
+**Core principles for financial calculations** include always using rust_decimal instead of floating-point, maintaining consistent scale throughout calculations, using explicit rounding strategies, and implementing overflow protection:
+
+```rust
+// Safe financial calculation with tax
+let subtotal = dec!(199.99);
+let tax_rate = dec!(0.0875); // 8.75%
+let tax = (subtotal * tax_rate).round_dp(2);
+let total = subtotal.checked_add(tax).expect("Calculation overflow");
+
+// Interest calculation with proper rounding
+let principal = dec!(10000.00);
+let rate = dec!(0.045); // 4.5% annual
+let compound_interest = principal * (dec!(1) + rate / dec!(12)).powi(12);
+let final_amount = compound_interest.round_dp(2);
+```
+
+### Integration with accounting systems
+
+**Database integration** requires appropriate feature flags and schema design:
+
+```rust
+// PostgreSQL integration
+[dependencies]
+rust_decimal = { version = "1.37", features = ["db-postgres"] }
+
+// Usage with tokio-postgres
+let amount: Decimal = dec!(1234.56);
+client.execute(
+    "INSERT INTO transactions (amount) VALUES ($1)",
+    &[&amount]
+)?;
+```
+
+**JSON serialization** maintains precision through string-based representation:
+
+```rust
+use serde::{Deserialize, Serialize};
+
+#[derive(Serialize, Deserialize)]
+struct Invoice {
+    #[serde(with = "rust_decimal::serde::str")]
+    total: Decimal,
+    #[serde(with = "rust_decimal::serde::arbitrary_precision")]
+    tax: Decimal,
+}
+```
+
+## Integration patterns
+
+### Codebase integration strategies
+
+**Domain-driven design patterns** provide robust abstractions for financial applications:
+
+```rust
+// Value object pattern for type safety
+#[derive(Debug, Clone, PartialEq)]
+pub struct Balance(Decimal);
+
+impl Balance {
+    pub fn new(amount: Decimal) -> Result<Self, Error> {
+        if amount < Decimal::ZERO {
+            return Err(Error::NegativeBalance);
+        }
+        Ok(Balance(amount))
+    }
+    
+    pub fn add(&self, other: &Balance) -> Result<Balance, Error> {
+        let new_amount = self.0.checked_add(other.0)
+            .ok_or(Error::Overflow)?;
+        Ok(Balance(new_amount))
+    }
+}
+```
+
+**Aggregate root patterns** encapsulate business logic and maintain consistency:
+
+```rust
+pub struct Account {
+    id: AccountId,
+    balance: Balance,
+    transactions: Vec<Transaction>,
+}
+
+impl Account {
+    pub fn debit(&mut self, amount: Decimal) -> Result<(), DomainError> {
+        if self.balance.value() < amount {
+            return Err(DomainError::InsufficientFunds);
+        }
+        
+        self.balance = Balance::new(self.balance.value() - amount)?;
+        self.transactions.push(Transaction::debit(amount));
+        Ok(())
+    }
+}
+```
+
+### Testing approaches
+
+**Property-based testing** with proptest ensures mathematical correctness:
+
+```rust
+use proptest::prelude::*;
+
+proptest! {
+    #[test]
+    fn test_addition_commutative(a in any::<Decimal>(), b in any::<Decimal>()) {
+        prop_assert_eq!(a + b, b + a);
+    }
+    
+    #[test]
+    fn test_compound_interest_monotonic(
+        principal in 0.01f64..1000000.0,
+        rate in 0.001f64..0.5,
+        periods in 1u32..100
+    ) {
+        let p = Decimal::from_f64(principal).unwrap();
+        let r = Decimal::from_f64(rate).unwrap();
+        let result = p * (Decimal::ONE + r).powi(periods as i64);
+        
+        // Property: compound interest should always be >= principal
+        prop_assert!(result >= p);
+    }
+}
+```
+
+### Error handling strategies
+
+**Comprehensive error handling** uses the `thiserror` crate for production systems:
+
+```rust
+#[derive(Debug, thiserror::Error)]
+pub enum FinancialError {
+    #[error("Insufficient funds: available {available}, required {required}")]
+    InsufficientFunds { available: Decimal, required: Decimal },
+    
+    #[error("Currency mismatch: expected {expected}, got {actual}")]
+    CurrencyMismatch { expected: String, actual: String },
+    
+    #[error("Precision overflow in calculation")]
+    PrecisionOverflow,
+    
+    #[error("Division by zero")]
+    DivisionByZero,
+}
+```
+
+## Conclusion
+
+rust_decimal provides a mature, production-ready foundation for financial applications requiring exact precision. Its 128-bit fixed-precision architecture, comprehensive rounding strategies, and extensive ecosystem integration make it ideal for everything from simple e-commerce transactions to complex multi-currency trading platforms. The library's emphasis on correctness over raw performance, combined with Rust's memory safety guarantees, creates a robust platform for mission-critical financial systems where precision errors can have significant monetary consequences.
+
+The key to successful implementation lies in proper domain modeling, comprehensive error handling, appropriate precision management, and thorough testing including property-based testing for financial invariants. With these practices, rust_decimal serves as a reliable foundation for financial software systems handling high-throughput transaction processing while maintaining the exact precision required for regulatory compliance and accounting accuracy.
--- a/steel_decimal/src/functions.rs
+++ b/steel_decimal/src/functions.rs
@@ -3,6 +3,29 @@ use rust_decimal::prelude::*;
 use rust_decimal::MathematicalOps;
 use std::str::FromStr;

+/// Global precision setting for the current Steel execution context
+thread_local! {
+    static PRECISION_CONTEXT: std::cell::RefCell<Option<u32>> = std::cell::RefCell::new(None);
+}
+
+/// Set execution precision for all decimal operations in current thread
+pub fn set_execution_precision(precision: Option<u32>) {
+    PRECISION_CONTEXT.with(|p| *p.borrow_mut() = precision);
+}
+
+/// Get current execution precision
+pub fn get_execution_precision() -> Option<u32> {
+    PRECISION_CONTEXT.with(|p| *p.borrow())
+}
+
+/// Format decimal according to current execution context
+fn format_result(decimal: Decimal) -> String {
+    match get_execution_precision() {
+        Some(precision) => decimal.round_dp(precision).to_string(),
+        None => decimal.to_string(), // Full precision (default behavior)
+    }
+}
+
 /// Helper function to parse decimals with strict accounting precision
 /// Supports both standard decimal notation AND scientific notation
 fn parse_decimal(s: &str) -> Result<Decimal, String> {
@@ -21,31 +44,26 @@ fn parse_decimal(s: &str) -> Result<Decimal, String> {

 /// Parse scientific notation (e.g., "1e2", "1.5e-3") using decimal arithmetic
 fn parse_scientific_notation(s: &str) -> Result<Decimal, String> {
-    // Split on 'e' or 'E' (case insensitive)
    let lower_s = s.to_lowercase();
    let parts: Vec<&str> = lower_s.split('e').collect();
    if parts.len() != 2 {
        return Err(format!("Invalid scientific notation '{}': malformed", s));
    }

-    // Parse mantissa and exponent
    let mantissa = Decimal::from_str(parts[0])
        .map_err(|_| format!("Invalid mantissa in '{}': {}", s, parts[0]))?;
    let exponent: i32 = parts[1].parse()
        .map_err(|_| format!("Invalid exponent in '{}': {}", s, parts[1]))?;

-    // Handle exponent using decimal arithmetic to maintain precision
    let result = if exponent == 0 {
        mantissa
    } else if exponent > 0 {
-        // Multiply by 10^exponent
        let ten = Decimal::from(10);
        let power_of_ten = ten.checked_powi(exponent as i64)
            .ok_or_else(|| format!("Exponent too large in '{}': {}", s, exponent))?;
        mantissa.checked_mul(power_of_ten)
            .ok_or_else(|| format!("Scientific notation result overflow in '{}'", s))?
    } else {
-        // Divide by 10^|exponent| for negative exponents
        let ten = Decimal::from(10);
        let positive_exp = (-exponent) as i64;
        let divisor = ten.checked_powi(positive_exp)
@@ -57,23 +75,23 @@ fn parse_scientific_notation(s: &str) -> Result<Decimal, String> {
    Ok(result)
 }

-// Basic arithmetic operations
+// Basic arithmetic operations (now precision-aware)
 pub fn decimal_add(a: String, b: String) -> Result<String, String> {
    let a_dec = parse_decimal(&a)?;
    let b_dec = parse_decimal(&b)?;
-    Ok((a_dec + b_dec).to_string())
+    Ok(format_result(a_dec + b_dec))
 }

 pub fn decimal_sub(a: String, b: String) -> Result<String, String> {
    let a_dec = parse_decimal(&a)?;
    let b_dec = parse_decimal(&b)?;
-    Ok((a_dec - b_dec).to_string())
+    Ok(format_result(a_dec - b_dec))
 }

 pub fn decimal_mul(a: String, b: String) -> Result<String, String> {
    let a_dec = parse_decimal(&a)?;
    let b_dec = parse_decimal(&b)?;
-    Ok((a_dec * b_dec).to_string())
+    Ok(format_result(a_dec * b_dec))
 }

 pub fn decimal_div(a: String, b: String) -> Result<String, String> {
@@ -84,16 +102,87 @@ pub fn decimal_div(a: String, b: String) -> Result<String, String> {
        return Err("Division by zero".to_string());
    }

-    Ok((a_dec / b_dec).to_string())
+    Ok(format_result(a_dec / b_dec))
 }

-// Advanced mathematical functions
+// Precision-specific operations (explicit precision override)
+pub fn decimal_add_p(a: String, b: String, precision: u32) -> Result<String, String> {
+    let a_dec = parse_decimal(&a)?;
+    let b_dec = parse_decimal(&b)?;
+    let result = a_dec + b_dec;
+    
+    Ok(result.round_dp(precision).to_string())
+}
+
+pub fn decimal_sub_p(a: String, b: String, precision: u32) -> Result<String, String> {
+    let a_dec = parse_decimal(&a)?;
+    let b_dec = parse_decimal(&b)?;
+    let result = a_dec - b_dec;
+    
+    Ok(result.round_dp(precision).to_string())
+}
+
+pub fn decimal_mul_p(a: String, b: String, precision: u32) -> Result<String, String> {
+    let a_dec = parse_decimal(&a)?;
+    let b_dec = parse_decimal(&b)?;
+    let result = a_dec * b_dec;
+    
+    Ok(result.round_dp(precision).to_string())
+}
+
+pub fn decimal_div_p(a: String, b: String, precision: u32) -> Result<String, String> {
+    let a_dec = parse_decimal(&a)?;
+    let b_dec = parse_decimal(&b)?;
+
+    if b_dec.is_zero() {
+        return Err("Division by zero".to_string());
+    }
+
+    let result = a_dec / b_dec;
+    
+    Ok(result.round_dp(precision).to_string())
+}
+
+// Precision control functions
+pub fn set_precision(precision: u32) -> String {
+    if precision > 28 {
+        "Error: Maximum precision is 28 decimal places".to_string()
+    } else {
+        set_execution_precision(Some(precision as u32));
+        format!("Precision set to {} decimal places", precision)
+    }
+}
+
+pub fn get_precision() -> String {
+    match get_execution_precision() {
+        Some(p) => p.to_string(),
+        None => "full".to_string(),
+    }
+}
+
+pub fn clear_precision() -> String {
+    set_execution_precision(None);
+    "Precision cleared - using full precision".to_string()
+}
+
+// Format functions with explicit precision
+pub fn decimal_format(value: String, precision: u32) -> Result<String, String> {
+    let decimal = parse_decimal(&value)?;
+    
+    if precision > 28 {
+        Err("Maximum precision is 28 decimal places".to_string())
+    } else {
+        Ok(decimal.round_dp(precision as u32).to_string())
+    }
+}
+
+// Advanced mathematical functions (updated to use format_result)
 pub fn decimal_pow(base: String, exp: String) -> Result<String, String> {
    let base_dec = parse_decimal(&base)?;
    let exp_dec = parse_decimal(&exp)?;

    base_dec.checked_powd(exp_dec)
-        .map(|result| result.to_string())
+        .map(|result| format_result(result))
        .ok_or_else(|| "Power operation failed or overflowed".to_string())
 }

@@ -101,7 +190,7 @@ pub fn decimal_sqrt(a: String) -> Result<String, String> {
    let a_dec = parse_decimal(&a)?;

    a_dec.sqrt()
-        .map(|result| result.to_string())
+        .map(|result| format_result(result))
        .ok_or_else(|| "Square root failed (negative number?)".to_string())
 }

@@ -109,7 +198,7 @@ pub fn decimal_ln(a: String) -> Result<String, String> {
    let a_dec = parse_decimal(&a)?;

    a_dec.checked_ln()
-        .map(|result| result.to_string())
+        .map(|result| format_result(result))
        .ok_or_else(|| "Natural log failed (non-positive number?)".to_string())
 }

@@ -117,7 +206,7 @@ pub fn decimal_log10(a: String) -> Result<String, String> {
    let a_dec = parse_decimal(&a)?;

    a_dec.checked_log10()
-        .map(|result| result.to_string())
+        .map(|result| format_result(result))
        .ok_or_else(|| "Log10 failed (non-positive number?)".to_string())
 }

@@ -125,16 +214,16 @@ pub fn decimal_exp(a: String) -> Result<String, String> {
    let a_dec = parse_decimal(&a)?;

    a_dec.checked_exp()
-        .map(|result| result.to_string())
+        .map(|result| format_result(result))
        .ok_or_else(|| "Exponential failed or overflowed".to_string())
 }

-// Trigonometric functions
+// Trigonometric functions (updated to use format_result)
 pub fn decimal_sin(a: String) -> Result<String, String> {
    let a_dec = parse_decimal(&a)?;

    a_dec.checked_sin()
-        .map(|result| result.to_string())
+        .map(|result| format_result(result))
        .ok_or_else(|| "Sine calculation failed or overflowed".to_string())
 }

@@ -142,7 +231,7 @@ pub fn decimal_cos(a: String) -> Result<String, String> {
    let a_dec = parse_decimal(&a)?;

    a_dec.checked_cos()
-        .map(|result| result.to_string())
+        .map(|result| format_result(result))
        .ok_or_else(|| "Cosine calculation failed or overflowed".to_string())
 }

@@ -150,11 +239,11 @@ pub fn decimal_tan(a: String) -> Result<String, String> {
    let a_dec = parse_decimal(&a)?;

    a_dec.checked_tan()
-        .map(|result| result.to_string())
+        .map(|result| format_result(result))
        .ok_or_else(|| "Tangent calculation failed or overflowed".to_string())
 }

-// Comparison functions
+// Comparison functions (unchanged)
 pub fn decimal_gt(a: String, b: String) -> Result<bool, String> {
    let a_dec = parse_decimal(&a)?;
    let b_dec = parse_decimal(&b)?;
@@ -185,30 +274,34 @@ pub fn decimal_eq(a: String, b: String) -> Result<bool, String> {
    Ok(a_dec == b_dec)
 }

-// Utility functions
+// Utility functions (updated to use format_result)
 pub fn decimal_abs(a: String) -> Result<String, String> {
    let a_dec = parse_decimal(&a)?;
-    Ok(a_dec.abs().to_string())
+    Ok(format_result(a_dec.abs()))
 }

 pub fn decimal_round(a: String, places: i32) -> Result<String, String> {
    let a_dec = parse_decimal(&a)?;
+    if places < 0 {
+        Ok(a_dec.to_string())
+    } else {
        Ok(a_dec.round_dp(places as u32).to_string())
    }
+}

 pub fn decimal_min(a: String, b: String) -> Result<String, String> {
    let a_dec = parse_decimal(&a)?;
    let b_dec = parse_decimal(&b)?;
-    Ok(a_dec.min(b_dec).to_string())
+    Ok(format_result(a_dec.min(b_dec)))
 }

 pub fn decimal_max(a: String, b: String) -> Result<String, String> {
    let a_dec = parse_decimal(&a)?;
    let b_dec = parse_decimal(&b)?;
-    Ok(a_dec.max(b_dec).to_string())
+    Ok(format_result(a_dec.max(b_dec)))
 }

-// Constants
+// Constants (unchanged)
 pub fn decimal_zero() -> String {
    "0".to_string()
 }
@@ -225,13 +318,13 @@ pub fn decimal_e() -> String {
    "2.7182818284590452353602874714".to_string()
 }

-// Financial functions
+// Financial functions (updated to use format_result)
 pub fn decimal_percentage(amount: String, percentage: String) -> Result<String, String> {
    let amount_dec = parse_decimal(&amount)?;
    let percentage_dec = parse_decimal(&percentage)?;
    let hundred = Decimal::from(100);

-    Ok((amount_dec * percentage_dec / hundred).to_string())
+    Ok(format_result(amount_dec * percentage_dec / hundred))
 }

 pub fn decimal_compound(principal: String, rate: String, time: String) -> Result<String, String> {
@@ -243,12 +336,12 @@ pub fn decimal_compound(principal: String, rate: String, time: String) -> Result
    let compound_factor = (one + rate_dec).checked_powd(time_dec)
        .ok_or("Compound calculation overflow")?;

-    Ok((principal_dec * compound_factor).to_string())
+    Ok(format_result(principal_dec * compound_factor))
 }

-// Type conversion helper
+// Type conversion helper (updated to use format_result)
 pub fn to_decimal(s: String) -> Result<String, String> {
    parse_decimal(&s)
-        .map(|d| d.to_string())
+        .map(|d| format_result(d))
        .map_err(|e| format!("Invalid decimal: {}", e))
 }
--- a/steel_decimal/src/registry.rs
+++ b/steel_decimal/src/registry.rs
@@ -10,6 +10,8 @@ impl FunctionRegistry {
    /// Register all decimal math functions with the Steel VM
    pub fn register_all(vm: &mut Engine) {
        Self::register_basic_arithmetic(vm);
+        Self::register_precision_arithmetic(vm);
+        Self::register_precision_control(vm);
        Self::register_advanced_math(vm);
        Self::register_trigonometric(vm);
        Self::register_comparison(vm);
@@ -27,6 +29,22 @@ impl FunctionRegistry {
        vm.register_fn("decimal-div", decimal_div);
    }

+    /// Register precision-specific arithmetic functions
+    pub fn register_precision_arithmetic(vm: &mut Engine) {
+        vm.register_fn("decimal-add-p", decimal_add_p);
+        vm.register_fn("decimal-sub-p", decimal_sub_p);
+        vm.register_fn("decimal-mul-p", decimal_mul_p);
+        vm.register_fn("decimal-div-p", decimal_div_p);
+    }
+
+    /// Register precision control functions
+    pub fn register_precision_control(vm: &mut Engine) {
+        vm.register_fn("set-precision", set_precision);
+        vm.register_fn("get-precision", get_precision);
+        vm.register_fn("clear-precision", clear_precision);
+        vm.register_fn("decimal-format", decimal_format);
+    }
+
    /// Register advanced mathematical functions
    pub fn register_advanced_math(vm: &mut Engine) {
        vm.register_fn("decimal-pow", decimal_pow);
--- a/steel_decimal/tests/function_tests.rs
+++ b/steel_decimal/tests/function_tests.rs
@@ -243,3 +243,55 @@ fn test_scientific_notation(#[case] a: &str, #[case] b: &str, #[case] expected:
    let result = decimal_add(a.to_string(), b.to_string()).unwrap();
    assert_eq!(result, expected);
 }
+
+// Test precision behavior
+#[rstest]
+#[case("5", "0", "5")]           // Integer + integer = integer
+#[case("5.0", "0", "5.0")]       // Decimal + integer = decimal
+#[case("5.00", "0.00", "5.00")]  // Preserves highest precision
+#[case("5.1", "0.23", "5.33")]   // Normal decimal arithmetic
+fn test_precision_preservation(#[case] a: &str, #[case] b: &str, #[case] expected: &str) {
+    let result = decimal_add(a.to_string(), b.to_string()).unwrap();
+    assert_eq!(result, expected);
+}
+
+// Test explicit precision functions
+#[rstest]
+#[case("5.123", "2.456", 0, "8")]     // 0 decimal places
+#[case("5.123", "2.456", 2, "7.58")]  // 2 decimal places
+#[case("5.123", "2.456", 4, "7.5790")] // 4 decimal places
+fn test_explicit_precision(#[case] a: &str, #[case] b: &str, #[case] precision: u32, #[case] expected: &str) {
+    let result = decimal_add_p(a.to_string(), b.to_string(), precision).unwrap();
+    assert_eq!(result, expected);
+}
+
+// Test scientific notation edge cases
+#[rstest]
+#[case("1e0", "1")]       // Simple case
+#[case("1.0e0", "1.0")]   // Preserves decimal
+#[case("1e-2", "0.01")]   // Negative exponent
+#[case("1.5e-3", "0.0015")] // Decimal + negative exponent
+#[case("2.5e2", "250.0")] // Decimal + positive exponent
+fn test_scientific_edge_cases(#[case] input: &str, #[case] expected: &str) {
+    let result = to_decimal(input.to_string()).unwrap();
+    assert_eq!(result, expected);
+}
+
+// Test precision functions
+#[test]
+fn test_precision_functions() {
+    // Test setting precision
+    assert_eq!(set_precision(2), "Precision set to 2 decimal places");
+    assert_eq!(get_precision(), "2");
+
+    // Test with precision set
+    let result = decimal_add("1.567".to_string(), "2.891".to_string()).unwrap();
+    assert_eq!(result, "4.46");
+
+    // Test clearing precision
+    assert_eq!(get_precision(), "full");
+
+    // Test with full precision
+    let result = decimal_add("1.567".to_string(), "2.891".to_string()).unwrap();
+    assert_eq!(result, "4.458");
+}