|
14 | 14 |
|
15 | 15 | use std::sync::Arc;
|
16 | 16 |
|
| 17 | +use databend_common_exception::ErrorCode; |
17 | 18 | use databend_common_exception::Result;
|
18 | 19 | use databend_common_expression::converts::arrow::table_schema_to_arrow_schema_ignore_inside_nullable;
|
| 20 | +use databend_common_expression::types::DataType; |
| 21 | +use databend_common_expression::types::DecimalDataType; |
| 22 | +use databend_common_expression::types::NumberDataType; |
19 | 23 | use databend_common_expression::BlockEntry;
|
20 | 24 | use databend_common_expression::DataBlock;
|
21 | 25 | use databend_common_expression::TableSchema;
|
22 | 26 | use databend_storages_common_table_meta::meta::StatisticsOfColumns;
|
23 | 27 | use databend_storages_common_table_meta::table::TableCompression;
|
| 28 | +use ethnum::I256; |
24 | 29 | use parquet_rs::arrow::arrow_to_parquet_schema;
|
25 | 30 | use parquet_rs::arrow::ArrowWriter;
|
26 | 31 | use parquet_rs::basic::Encoding;
|
@@ -107,6 +112,9 @@ fn choose_compression_scheme(
|
107 | 112 | if can_apply_dict_encoding(physical_type, distinct_of_values, num_rows, col)? {
|
108 | 113 | let col_path = ColumnPath::new(vec![table_field.name().clone()]);
|
109 | 114 | props = props.set_column_dictionary_enabled(col_path, true);
|
| 115 | + } else if can_apply_delta_binary_pack(physical_type, col, num_rows)? { |
| 116 | + let col_path = ColumnPath::new(vec![table_field.name().clone()]); |
| 117 | + props = props.set_column_encoding(col_path, Encoding::DELTA_BINARY_PACKED); |
110 | 118 | }
|
111 | 119 | }
|
112 | 120 | Type::GroupType {
|
@@ -137,3 +145,195 @@ fn can_apply_dict_encoding(
|
137 | 145 | let avg_bytes_per_value = total_bytes as f64 / num_rows as f64;
|
138 | 146 | Ok(is_low_cardinality && avg_bytes_per_value < AVG_BYTES_PER_VALUE)
|
139 | 147 | }
|
| 148 | + |
| 149 | +fn can_apply_delta_binary_pack( |
| 150 | + physical_type: &PhysicalType, |
| 151 | + col: &BlockEntry, |
| 152 | + num_rows: usize, |
| 153 | +) -> Result<bool> { |
| 154 | + const MAX_DELTA: i64 = 1 << 3; |
| 155 | + if !matches!( |
| 156 | + physical_type, |
| 157 | + PhysicalType::INT32 | PhysicalType::INT64 | PhysicalType::INT96 |
| 158 | + ) { |
| 159 | + return Ok(false); |
| 160 | + } |
| 161 | + if num_rows == 0 { |
| 162 | + return Ok(false); |
| 163 | + } |
| 164 | + let col = col.value.convert_to_full_column(&col.data_type, num_rows); |
| 165 | + match col.data_type().remove_nullable() { |
| 166 | + DataType::Number(NumberDataType::UInt8) => { |
| 167 | + let mut max_delta = 0; |
| 168 | + let col = col.as_number().unwrap().as_u_int8().unwrap(); |
| 169 | + let mut col_iter = col.iter(); |
| 170 | + let mut prev = *col_iter.next().unwrap(); |
| 171 | + for &v in col_iter { |
| 172 | + let delta = if v > prev { v - prev } else { prev - v }; |
| 173 | + if delta > max_delta { |
| 174 | + max_delta = delta; |
| 175 | + } |
| 176 | + prev = v; |
| 177 | + } |
| 178 | + Ok(max_delta as i64 <= MAX_DELTA) |
| 179 | + } |
| 180 | + DataType::Number(NumberDataType::UInt16) => { |
| 181 | + let mut max_delta = 0; |
| 182 | + let col = col.as_number().unwrap().as_u_int16().unwrap(); |
| 183 | + let mut col_iter = col.iter(); |
| 184 | + let mut prev = *col_iter.next().unwrap(); |
| 185 | + for &v in col_iter { |
| 186 | + let delta = if v > prev { v - prev } else { prev - v }; |
| 187 | + if delta > max_delta { |
| 188 | + max_delta = delta; |
| 189 | + } |
| 190 | + prev = v; |
| 191 | + } |
| 192 | + Ok(max_delta as i64 <= MAX_DELTA) |
| 193 | + } |
| 194 | + DataType::Number(NumberDataType::UInt32) => { |
| 195 | + let mut max_delta = 0; |
| 196 | + let col = col.as_number().unwrap().as_u_int32().unwrap(); |
| 197 | + let mut col_iter = col.iter(); |
| 198 | + let mut prev = *col_iter.next().unwrap(); |
| 199 | + for &v in col_iter { |
| 200 | + let delta = if v > prev { v - prev } else { prev - v }; |
| 201 | + if delta > max_delta { |
| 202 | + max_delta = delta; |
| 203 | + } |
| 204 | + prev = v; |
| 205 | + } |
| 206 | + Ok(max_delta as i64 <= MAX_DELTA) |
| 207 | + } |
| 208 | + DataType::Number(NumberDataType::UInt64) => { |
| 209 | + let mut max_delta = 0; |
| 210 | + let col = col.as_number().unwrap().as_u_int64().unwrap(); |
| 211 | + let mut col_iter = col.iter(); |
| 212 | + let mut prev = *col_iter.next().unwrap(); |
| 213 | + for &v in col_iter { |
| 214 | + let delta = if v > prev { v - prev } else { prev - v }; |
| 215 | + if delta > max_delta { |
| 216 | + max_delta = delta; |
| 217 | + } |
| 218 | + prev = v; |
| 219 | + } |
| 220 | + Ok(max_delta as i64 <= MAX_DELTA) |
| 221 | + } |
| 222 | + DataType::Number(NumberDataType::Int8) => { |
| 223 | + let mut max_delta = 0; |
| 224 | + let col = col.as_number().unwrap().as_int8().unwrap(); |
| 225 | + let mut col_iter = col.iter(); |
| 226 | + let mut prev = *col_iter.next().unwrap(); |
| 227 | + for &v in col_iter { |
| 228 | + let delta = if v > prev { v - prev } else { prev - v }; |
| 229 | + if delta > max_delta { |
| 230 | + max_delta = delta; |
| 231 | + } |
| 232 | + prev = v; |
| 233 | + } |
| 234 | + Ok(max_delta as i64 <= MAX_DELTA) |
| 235 | + } |
| 236 | + DataType::Number(NumberDataType::Int16) => { |
| 237 | + let mut max_delta = 0; |
| 238 | + let col = col.as_number().unwrap().as_int16().unwrap(); |
| 239 | + let mut col_iter = col.iter(); |
| 240 | + let mut prev = *col_iter.next().unwrap(); |
| 241 | + for &v in col_iter { |
| 242 | + let delta = if v > prev { v - prev } else { prev - v }; |
| 243 | + if delta > max_delta { |
| 244 | + max_delta = delta; |
| 245 | + } |
| 246 | + prev = v; |
| 247 | + } |
| 248 | + Ok(max_delta as i64 <= MAX_DELTA) |
| 249 | + } |
| 250 | + DataType::Number(NumberDataType::Int32) => { |
| 251 | + let mut max_delta = 0; |
| 252 | + let col = col.as_number().unwrap().as_int32().unwrap(); |
| 253 | + let mut col_iter = col.iter(); |
| 254 | + let mut prev = *col_iter.next().unwrap(); |
| 255 | + for &v in col_iter { |
| 256 | + let delta = if v > prev { v - prev } else { prev - v }; |
| 257 | + if delta > max_delta { |
| 258 | + max_delta = delta; |
| 259 | + } |
| 260 | + prev = v; |
| 261 | + } |
| 262 | + Ok(max_delta as i64 <= MAX_DELTA) |
| 263 | + } |
| 264 | + DataType::Number(NumberDataType::Int64) => { |
| 265 | + let mut max_delta = 0; |
| 266 | + let col = col.as_number().unwrap().as_int64().unwrap(); |
| 267 | + let mut col_iter = col.iter(); |
| 268 | + let mut prev = *col_iter.next().unwrap(); |
| 269 | + for &v in col_iter { |
| 270 | + let delta = if v > prev { v - prev } else { prev - v }; |
| 271 | + if delta > max_delta { |
| 272 | + max_delta = delta; |
| 273 | + } |
| 274 | + prev = v; |
| 275 | + } |
| 276 | + Ok(max_delta <= MAX_DELTA) |
| 277 | + } |
| 278 | + DataType::Decimal(DecimalDataType::Decimal128(_)) => { |
| 279 | + let mut max_delta = 0; |
| 280 | + let (col, _) = col.as_decimal().unwrap().as_decimal128().unwrap(); |
| 281 | + let mut col_iter = col.iter(); |
| 282 | + let mut prev = *col_iter.next().unwrap(); |
| 283 | + for &v in col_iter { |
| 284 | + let delta = if v > prev { v - prev } else { prev - v }; |
| 285 | + if delta > max_delta { |
| 286 | + max_delta = delta; |
| 287 | + } |
| 288 | + prev = v; |
| 289 | + } |
| 290 | + Ok(max_delta <= MAX_DELTA as i128) |
| 291 | + } |
| 292 | + DataType::Decimal(DecimalDataType::Decimal256(_)) => { |
| 293 | + let mut max_delta: I256 = I256::ZERO; |
| 294 | + let (col, _) = col.as_decimal().unwrap().as_decimal256().unwrap(); |
| 295 | + let mut col_iter = col.iter(); |
| 296 | + let mut prev = *col_iter.next().unwrap(); |
| 297 | + for &v in col_iter { |
| 298 | + let delta = if v > prev { v - prev } else { prev - v }; |
| 299 | + if delta > max_delta { |
| 300 | + max_delta = delta; |
| 301 | + } |
| 302 | + prev = v; |
| 303 | + } |
| 304 | + Ok(max_delta <= I256::from(MAX_DELTA)) |
| 305 | + } |
| 306 | + DataType::Timestamp => { |
| 307 | + let mut max_delta = 0; |
| 308 | + let col = col.as_timestamp().unwrap(); |
| 309 | + let mut col_iter = col.iter(); |
| 310 | + let mut prev = *col_iter.next().unwrap(); |
| 311 | + for &v in col_iter { |
| 312 | + let delta = if v > prev { v - prev } else { prev - v }; |
| 313 | + if delta > max_delta { |
| 314 | + max_delta = delta; |
| 315 | + } |
| 316 | + prev = v; |
| 317 | + } |
| 318 | + Ok(max_delta <= MAX_DELTA) |
| 319 | + } |
| 320 | + DataType::Date => { |
| 321 | + let mut max_delta = 0; |
| 322 | + let col = col.as_date().unwrap(); |
| 323 | + let mut col_iter = col.iter(); |
| 324 | + let mut prev = *col_iter.next().unwrap(); |
| 325 | + for &v in col_iter { |
| 326 | + let delta = if v > prev { v - prev } else { prev - v }; |
| 327 | + if delta > max_delta { |
| 328 | + max_delta = delta; |
| 329 | + } |
| 330 | + prev = v; |
| 331 | + } |
| 332 | + Ok(max_delta as i64 <= MAX_DELTA) |
| 333 | + } |
| 334 | + _ => Err(ErrorCode::Internal(format!( |
| 335 | + "Unsupported data type for delta binary pack: {:?}", |
| 336 | + col.data_type() |
| 337 | + ))), |
| 338 | + } |
| 339 | +} |
0 commit comments