|
| 1 | +// Licensed to the Apache Software Foundation (ASF) under one |
| 2 | +// or more contributor license agreements. See the NOTICE file |
| 3 | +// distributed with this work for additional information |
| 4 | +// regarding copyright ownership. The ASF licenses this file |
| 5 | +// to you under the Apache License, Version 2.0 (the |
| 6 | +// "License"); you may not use this file except in compliance |
| 7 | +// with the License. You may obtain a copy of the License at |
| 8 | +// |
| 9 | +// http://www.apache.org/licenses/LICENSE-2.0 |
| 10 | +// |
| 11 | +// Unless required by applicable law or agreed to in writing, |
| 12 | +// software distributed under the License is distributed on an |
| 13 | +// "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY |
| 14 | +// KIND, either express or implied. See the License for the |
| 15 | +// specific language governing permissions and limitations |
| 16 | +// under the License. |
| 17 | + |
| 18 | +//! Compare DistinctCount for string with naive HashSet and Short String Optimized HashSet |
| 19 | +
|
| 20 | +use std::sync::Arc; |
| 21 | + |
| 22 | +use arrow::array::ArrayRef; |
| 23 | +use arrow::record_batch::RecordBatch; |
| 24 | +use arrow_array::{Array, GenericStringArray, OffsetSizeTrait, UInt32Array}; |
| 25 | + |
| 26 | +use arrow_array::cast::AsArray; |
| 27 | +use datafusion::datasource::MemTable; |
| 28 | +use rand::rngs::StdRng; |
| 29 | +use rand::{thread_rng, Rng, SeedableRng}; |
| 30 | +use std::collections::HashSet; |
| 31 | +use tokio::task::JoinSet; |
| 32 | + |
| 33 | +use datafusion::prelude::{SessionConfig, SessionContext}; |
| 34 | +use test_utils::stagger_batch; |
| 35 | + |
| 36 | +#[tokio::test(flavor = "multi_thread")] |
| 37 | +async fn distinct_count_string_test() { |
| 38 | + // max length of generated strings |
| 39 | + let mut join_set = JoinSet::new(); |
| 40 | + let mut rng = thread_rng(); |
| 41 | + for null_pct in [0.0, 0.01, 0.1, 0.5] { |
| 42 | + for _ in 0..100 { |
| 43 | + let max_len = rng.gen_range(1..50); |
| 44 | + let num_strings = rng.gen_range(1..100); |
| 45 | + let num_distinct_strings = if num_strings > 1 { |
| 46 | + rng.gen_range(1..num_strings) |
| 47 | + } else { |
| 48 | + num_strings |
| 49 | + }; |
| 50 | + let generator = BatchGenerator { |
| 51 | + max_len, |
| 52 | + num_strings, |
| 53 | + num_distinct_strings, |
| 54 | + null_pct, |
| 55 | + rng: StdRng::from_seed(rng.gen()), |
| 56 | + }; |
| 57 | + join_set.spawn(async move { run_distinct_count_test(generator).await }); |
| 58 | + } |
| 59 | + } |
| 60 | + while let Some(join_handle) = join_set.join_next().await { |
| 61 | + // propagate errors |
| 62 | + join_handle.unwrap(); |
| 63 | + } |
| 64 | +} |
| 65 | + |
| 66 | +/// Run COUNT DISTINCT using SQL and compare the result to computing the |
| 67 | +/// distinct count using HashSet<String> |
| 68 | +async fn run_distinct_count_test(mut generator: BatchGenerator) { |
| 69 | + let input = generator.make_input_batches(); |
| 70 | + |
| 71 | + let schema = input[0].schema(); |
| 72 | + let session_config = SessionConfig::new().with_batch_size(50); |
| 73 | + let ctx = SessionContext::new_with_config(session_config); |
| 74 | + |
| 75 | + // split input into two partitions |
| 76 | + let partition_len = input.len() / 2; |
| 77 | + let partitions = vec![ |
| 78 | + input[0..partition_len].to_vec(), |
| 79 | + input[partition_len..].to_vec(), |
| 80 | + ]; |
| 81 | + |
| 82 | + let provider = MemTable::try_new(schema, partitions).unwrap(); |
| 83 | + ctx.register_table("t", Arc::new(provider)).unwrap(); |
| 84 | + // input has two columns, a and b. The result is the number of distinct |
| 85 | + // values in each column. |
| 86 | + // |
| 87 | + // Note, we need at least two count distinct aggregates to trigger the |
| 88 | + // count distinct aggregate. Otherwise, the optimizer will rewrite the |
| 89 | + // `COUNT(DISTINCT a)` to `COUNT(*) from (SELECT DISTINCT a FROM t)` |
| 90 | + let results = ctx |
| 91 | + .sql("SELECT COUNT(DISTINCT a), COUNT(DISTINCT b) FROM t") |
| 92 | + .await |
| 93 | + .unwrap() |
| 94 | + .collect() |
| 95 | + .await |
| 96 | + .unwrap(); |
| 97 | + |
| 98 | + // get all the strings from the first column of the result (distinct a) |
| 99 | + let expected_a = extract_distinct_strings::<i32>(&input, 0).len(); |
| 100 | + let result_a = extract_i64(&results, 0); |
| 101 | + assert_eq!(expected_a, result_a); |
| 102 | + |
| 103 | + // get all the strings from the second column of the result (distinct b( |
| 104 | + let expected_b = extract_distinct_strings::<i64>(&input, 1).len(); |
| 105 | + let result_b = extract_i64(&results, 1); |
| 106 | + assert_eq!(expected_b, result_b); |
| 107 | +} |
| 108 | + |
| 109 | +/// Return all (non null) distinct strings from column col_idx |
| 110 | +fn extract_distinct_strings<O: OffsetSizeTrait>( |
| 111 | + results: &[RecordBatch], |
| 112 | + col_idx: usize, |
| 113 | +) -> Vec<String> { |
| 114 | + results |
| 115 | + .iter() |
| 116 | + .flat_map(|batch| { |
| 117 | + let array = batch.column(col_idx).as_string::<O>(); |
| 118 | + // remove nulls via 'flatten' |
| 119 | + array.iter().flatten().map(|s| s.to_string()) |
| 120 | + }) |
| 121 | + .collect::<HashSet<_>>() |
| 122 | + .into_iter() |
| 123 | + .collect() |
| 124 | +} |
| 125 | + |
| 126 | +// extract the value from the Int64 column in col_idx in batch and return |
| 127 | +// it as a usize |
| 128 | +fn extract_i64(results: &[RecordBatch], col_idx: usize) -> usize { |
| 129 | + assert_eq!(results.len(), 1); |
| 130 | + let array = results[0] |
| 131 | + .column(col_idx) |
| 132 | + .as_any() |
| 133 | + .downcast_ref::<arrow::array::Int64Array>() |
| 134 | + .unwrap(); |
| 135 | + assert_eq!(array.len(), 1); |
| 136 | + assert!(!array.is_null(0)); |
| 137 | + array.value(0).try_into().unwrap() |
| 138 | +} |
| 139 | + |
| 140 | +struct BatchGenerator { |
| 141 | + //// The maximum length of the strings |
| 142 | + max_len: usize, |
| 143 | + /// the total number of strings in the output |
| 144 | + num_strings: usize, |
| 145 | + /// The number of distinct strings in the columns |
| 146 | + num_distinct_strings: usize, |
| 147 | + /// The percentage of nulls in the columns |
| 148 | + null_pct: f64, |
| 149 | + /// Random number generator |
| 150 | + rng: StdRng, |
| 151 | +} |
| 152 | + |
| 153 | +impl BatchGenerator { |
| 154 | + /// Make batches of random strings with a random length columns "a" and "b": |
| 155 | + /// |
| 156 | + /// * "a" is a StringArray |
| 157 | + /// * "b" is a LargeStringArray |
| 158 | + fn make_input_batches(&mut self) -> Vec<RecordBatch> { |
| 159 | + // use a random number generator to pick a random sized output |
| 160 | + |
| 161 | + let batch = RecordBatch::try_from_iter(vec![ |
| 162 | + ("a", self.gen_data::<i32>()), |
| 163 | + ("b", self.gen_data::<i64>()), |
| 164 | + ]) |
| 165 | + .unwrap(); |
| 166 | + |
| 167 | + stagger_batch(batch) |
| 168 | + } |
| 169 | + |
| 170 | + /// Creates a StringArray or LargeStringArray with random strings according |
| 171 | + /// to the parameters of the BatchGenerator |
| 172 | + fn gen_data<O: OffsetSizeTrait>(&mut self) -> ArrayRef { |
| 173 | + // table of strings from which to draw |
| 174 | + let distinct_strings: GenericStringArray<O> = (0..self.num_distinct_strings) |
| 175 | + .map(|_| Some(random_string(&mut self.rng, self.max_len))) |
| 176 | + .collect(); |
| 177 | + |
| 178 | + // pick num_strings randomly from the distinct string table |
| 179 | + let indicies: UInt32Array = (0..self.num_strings) |
| 180 | + .map(|_| { |
| 181 | + if self.rng.gen::<f64>() < self.null_pct { |
| 182 | + None |
| 183 | + } else if self.num_distinct_strings > 1 { |
| 184 | + let range = 1..(self.num_distinct_strings as u32); |
| 185 | + Some(self.rng.gen_range(range)) |
| 186 | + } else { |
| 187 | + Some(0) |
| 188 | + } |
| 189 | + }) |
| 190 | + .collect(); |
| 191 | + |
| 192 | + let options = None; |
| 193 | + arrow::compute::take(&distinct_strings, &indicies, options).unwrap() |
| 194 | + } |
| 195 | +} |
| 196 | + |
| 197 | +/// Return a string of random characters of length 1..=max_len |
| 198 | +fn random_string(rng: &mut StdRng, max_len: usize) -> String { |
| 199 | + // pick characters at random (not just ascii) |
| 200 | + match max_len { |
| 201 | + 0 => "".to_string(), |
| 202 | + 1 => String::from(rng.gen::<char>()), |
| 203 | + _ => { |
| 204 | + let len = rng.gen_range(1..=max_len); |
| 205 | + rng.sample_iter::<char, _>(rand::distributions::Standard) |
| 206 | + .take(len) |
| 207 | + .map(char::from) |
| 208 | + .collect::<String>() |
| 209 | + } |
| 210 | + } |
| 211 | +} |
0 commit comments