|
| 1 | +//! Row Expansion System |
| 2 | +//! |
| 3 | +//! This module provides a generic framework for functions that multiply rows. |
| 4 | +//! Unlike regular functions (value → value) or generators (args → table), |
| 5 | +//! row expanders take a single input row and produce multiple output rows. |
| 6 | +//! |
| 7 | +//! ## Use Cases |
| 8 | +//! - UNNEST: Split delimited strings into separate rows |
| 9 | +//! - EXPLODE: Expand JSON arrays |
| 10 | +//! - GENERATE_SERIES: Create multiple rows from a range |
| 11 | +//! - Custom expansions for domain-specific data |
| 12 | +
|
| 13 | +use crate::data::datatable::DataValue; |
| 14 | +use anyhow::Result; |
| 15 | + |
| 16 | +/// Result of row expansion: a vector of values for this column across N rows |
| 17 | +#[derive(Debug, Clone)] |
| 18 | +pub struct ExpansionResult { |
| 19 | + /// The expanded values (one per output row) |
| 20 | + pub values: Vec<DataValue>, |
| 21 | +} |
| 22 | + |
| 23 | +impl ExpansionResult { |
| 24 | + /// Create a new expansion result |
| 25 | + pub fn new(values: Vec<DataValue>) -> Self { |
| 26 | + Self { values } |
| 27 | + } |
| 28 | + |
| 29 | + /// Number of rows this expansion will create |
| 30 | + pub fn row_count(&self) -> usize { |
| 31 | + self.values.len() |
| 32 | + } |
| 33 | +} |
| 34 | + |
| 35 | +/// Trait for row expander implementations |
| 36 | +/// |
| 37 | +/// Row expanders are special expressions that cause a single input row |
| 38 | +/// to be multiplied into multiple output rows. |
| 39 | +pub trait RowExpander: Send + Sync { |
| 40 | + /// Name of the expander (e.g., "UNNEST") |
| 41 | + fn name(&self) -> &str; |
| 42 | + |
| 43 | + /// Description of what this expander does |
| 44 | + fn description(&self) -> &str; |
| 45 | + |
| 46 | + /// Expand a single value into multiple values |
| 47 | + /// |
| 48 | + /// # Arguments |
| 49 | + /// * `value` - The input value to expand (e.g., delimited string) |
| 50 | + /// * `args` - Additional arguments (e.g., delimiter) |
| 51 | + /// |
| 52 | + /// # Returns |
| 53 | + /// An ExpansionResult containing the array of values for output rows |
| 54 | + fn expand(&self, value: &DataValue, args: &[DataValue]) -> Result<ExpansionResult>; |
| 55 | +} |
| 56 | + |
| 57 | +/// Registry of available row expanders |
| 58 | +pub struct RowExpanderRegistry { |
| 59 | + expanders: std::collections::HashMap<String, Box<dyn RowExpander>>, |
| 60 | +} |
| 61 | + |
| 62 | +impl RowExpanderRegistry { |
| 63 | + /// Create a new registry with default expanders |
| 64 | + pub fn new() -> Self { |
| 65 | + let mut registry = Self { |
| 66 | + expanders: std::collections::HashMap::new(), |
| 67 | + }; |
| 68 | + |
| 69 | + // Register built-in expanders |
| 70 | + registry.register(Box::new(unnest::UnnestExpander)); |
| 71 | + |
| 72 | + registry |
| 73 | + } |
| 74 | + |
| 75 | + /// Register a new expander |
| 76 | + pub fn register(&mut self, expander: Box<dyn RowExpander>) { |
| 77 | + self.expanders |
| 78 | + .insert(expander.name().to_uppercase(), expander); |
| 79 | + } |
| 80 | + |
| 81 | + /// Get an expander by name |
| 82 | + pub fn get(&self, name: &str) -> Option<&dyn RowExpander> { |
| 83 | + self.expanders.get(&name.to_uppercase()).map(|e| e.as_ref()) |
| 84 | + } |
| 85 | + |
| 86 | + /// Check if an expander exists |
| 87 | + pub fn contains(&self, name: &str) -> bool { |
| 88 | + self.expanders.contains_key(&name.to_uppercase()) |
| 89 | + } |
| 90 | + |
| 91 | + /// List all registered expanders |
| 92 | + pub fn list(&self) -> Vec<&str> { |
| 93 | + self.expanders.keys().map(|s| s.as_str()).collect() |
| 94 | + } |
| 95 | +} |
| 96 | + |
| 97 | +impl Default for RowExpanderRegistry { |
| 98 | + fn default() -> Self { |
| 99 | + Self::new() |
| 100 | + } |
| 101 | +} |
| 102 | + |
| 103 | +/// UNNEST expander implementation |
| 104 | +pub mod unnest { |
| 105 | + use super::*; |
| 106 | + |
| 107 | + pub struct UnnestExpander; |
| 108 | + |
| 109 | + impl RowExpander for UnnestExpander { |
| 110 | + fn name(&self) -> &str { |
| 111 | + "UNNEST" |
| 112 | + } |
| 113 | + |
| 114 | + fn description(&self) -> &str { |
| 115 | + "Split a delimited string into multiple rows" |
| 116 | + } |
| 117 | + |
| 118 | + fn expand(&self, value: &DataValue, args: &[DataValue]) -> Result<ExpansionResult> { |
| 119 | + // Get the delimiter (first argument) |
| 120 | + let delimiter = match args.first() { |
| 121 | + Some(DataValue::String(s)) => s.as_str(), |
| 122 | + Some(_) => { |
| 123 | + return Err(anyhow::anyhow!( |
| 124 | + "UNNEST delimiter must be a string, got {:?}", |
| 125 | + args.first() |
| 126 | + )) |
| 127 | + } |
| 128 | + None => return Err(anyhow::anyhow!("UNNEST requires a delimiter argument")), |
| 129 | + }; |
| 130 | + |
| 131 | + // Convert value to string |
| 132 | + let text = match value { |
| 133 | + DataValue::String(s) => s.clone(), |
| 134 | + DataValue::Null => { |
| 135 | + // NULL expands to a single NULL |
| 136 | + return Ok(ExpansionResult::new(vec![DataValue::Null])); |
| 137 | + } |
| 138 | + other => other.to_string(), |
| 139 | + }; |
| 140 | + |
| 141 | + // Split the string |
| 142 | + let parts: Vec<DataValue> = if delimiter.is_empty() { |
| 143 | + // Empty delimiter: split into characters |
| 144 | + text.chars() |
| 145 | + .map(|ch| DataValue::String(ch.to_string())) |
| 146 | + .collect() |
| 147 | + } else { |
| 148 | + // Split by delimiter, filter out empty parts |
| 149 | + text.split(delimiter) |
| 150 | + .filter(|s| !s.is_empty()) |
| 151 | + .map(|s| DataValue::String(s.to_string())) |
| 152 | + .collect() |
| 153 | + }; |
| 154 | + |
| 155 | + // If no parts, return single NULL to maintain row presence |
| 156 | + if parts.is_empty() { |
| 157 | + Ok(ExpansionResult::new(vec![DataValue::Null])) |
| 158 | + } else { |
| 159 | + Ok(ExpansionResult::new(parts)) |
| 160 | + } |
| 161 | + } |
| 162 | + } |
| 163 | +} |
| 164 | + |
| 165 | +#[cfg(test)] |
| 166 | +mod tests { |
| 167 | + use super::*; |
| 168 | + |
| 169 | + #[test] |
| 170 | + fn test_unnest_basic() { |
| 171 | + let expander = unnest::UnnestExpander; |
| 172 | + let value = DataValue::String("A|B|C".to_string()); |
| 173 | + let delimiter = DataValue::String("|".to_string()); |
| 174 | + |
| 175 | + let result = expander.expand(&value, &[delimiter]).unwrap(); |
| 176 | + assert_eq!(result.row_count(), 3); |
| 177 | + assert_eq!(result.values[0], DataValue::String("A".to_string())); |
| 178 | + assert_eq!(result.values[1], DataValue::String("B".to_string())); |
| 179 | + assert_eq!(result.values[2], DataValue::String("C".to_string())); |
| 180 | + } |
| 181 | + |
| 182 | + #[test] |
| 183 | + fn test_unnest_null() { |
| 184 | + let expander = unnest::UnnestExpander; |
| 185 | + let value = DataValue::Null; |
| 186 | + let delimiter = DataValue::String("|".to_string()); |
| 187 | + |
| 188 | + let result = expander.expand(&value, &[delimiter]).unwrap(); |
| 189 | + assert_eq!(result.row_count(), 1); |
| 190 | + assert_eq!(result.values[0], DataValue::Null); |
| 191 | + } |
| 192 | + |
| 193 | + #[test] |
| 194 | + fn test_registry() { |
| 195 | + let registry = RowExpanderRegistry::new(); |
| 196 | + assert!(registry.contains("UNNEST")); |
| 197 | + assert!(registry.contains("unnest")); |
| 198 | + |
| 199 | + let expander = registry.get("UNNEST").unwrap(); |
| 200 | + assert_eq!(expander.name(), "UNNEST"); |
| 201 | + } |
| 202 | +} |
0 commit comments