|
| 1 | +use std::{ |
| 2 | + fs::File, |
| 3 | + sync::Arc |
| 4 | +}; |
| 5 | +use arrow::{ |
| 6 | + array::{Int16Array, Int64Array, StringArray, StructArray}, |
| 7 | + datatypes::{DataType, Field, Schema}, |
| 8 | + compute as cp, |
| 9 | + error::Result, |
| 10 | + record_batch::RecordBatch |
| 11 | +}; |
| 12 | +use datafusion::{ |
| 13 | + datasource::file_format::FileFormat, |
| 14 | + prelude::* |
| 15 | +}; |
| 16 | +use parquet; |
| 17 | + |
| 18 | +fn create_batch() -> RecordBatch { |
| 19 | + let schema = Schema::new(vec![ |
| 20 | + Field::new("a", DataType::Int64, false), |
| 21 | + Field::new("b", DataType::Int64, false), |
| 22 | + Field::new("c", DataType::Int64, false) |
| 23 | + ]); |
| 24 | + let array_a = Int64Array::from([0, 1, 2, 3, 4, 5, 6, 7, 8, 9]); |
| 25 | + let array_b = Int64Array::from([9, 8, 7, 6, 5, 4, 3, 2, 1, 0]); |
| 26 | + let array_c = Int64Array::from([1, 2, 1, 2, 1, 2, 1, 2, 1, 2]); |
| 27 | + RecordBatch::try_new(Arc::new(schema), vec![array_a, array_b, array_c]).unwrap() |
| 28 | +} |
| 29 | + |
| 30 | +fn create_sample_dataset( |
| 31 | + filesystem: &const std::shared_ptr<fs::FileSystem>, |
| 32 | + root_path: &str) -> std::string { |
| 33 | + let base_path = root_path + "/parquet_dataset"; |
| 34 | + ABORT_ON_FAIL(filesystem.CreateDir(base_path)); |
| 35 | + let batch = create_batch(); |
| 36 | + let output = |
| 37 | + filesystem.OpenOutputStream(base_path + "/data1.parquet").unwrap(); |
| 38 | + ABORT_ON_FAIL(parquet::arrow::WriteTable(*batch.Slice(0, 5), |
| 39 | + arrow::default_memory_pool(), output, |
| 40 | + /*chunk_size*/ 2048)); |
| 41 | + output = |
| 42 | + filesystem.OpenOutputStream(base_path + "/data2.parquet").unwrap(); |
| 43 | + ABORT_ON_FAIL(parquet::arrow::WriteTable(*batch.Slice(5), |
| 44 | + arrow::default_memory_pool(), output, |
| 45 | + /*chunk_size*/ 2048)); |
| 46 | + base_path |
| 47 | +} |
| 48 | + |
| 49 | +fn scan_dataset( |
| 50 | + filesystem: const std::shared_ptr<fs::FileSystem>&, |
| 51 | + format: impl FileFormat, |
| 52 | + base_dir: &str) -> std::shared_ptr<arrow::Table> { |
| 53 | + fs::FileSelector selector; |
| 54 | + selector.base_dir = base_dir; |
| 55 | + let factory = |
| 56 | + ds::FileSystemDatasetFactory::Make(filesystem, selector, format, |
| 57 | + ds::FileSystemFactoryOptions()) |
| 58 | + .unwrap(); |
| 59 | + |
| 60 | + let dataset = factory.finish().unwrap(); |
| 61 | + let fragments = dataset.GetFragments().unwrap(); |
| 62 | + for fragment in fragments { |
| 63 | + println!("Found Fragment: {}", fragment); |
| 64 | + } |
| 65 | + |
| 66 | + let scan_builder = dataset.NewScan().unwrap(); |
| 67 | + let scanner = scan_builder.finish().unwrap(); |
| 68 | + return scanner.ToTable().unwrap(); |
| 69 | +} |
| 70 | + |
| 71 | +fn filter_and_select( |
| 72 | + filesystem: const std::shared_ptr<fs::FileSystem>&, |
| 73 | + format: const std::shared_ptr<ds::FileFormat>&, |
| 74 | + base_dir: &str) -> std::shared_ptr<arrow::Table> { |
| 75 | + fs::FileSelector selector; |
| 76 | + selector.base_dir = base_dir; |
| 77 | + let factory = |
| 78 | + ds::FileSystemDatasetFactory::Make(filesystem, selector, format, |
| 79 | + ds::FileSystemFactoryOptions()) |
| 80 | + .unwrap(); |
| 81 | + |
| 82 | + let dataset = factory.finish().unwrap(); |
| 83 | + let scan_builder = dataset.NewScan().unwrap(); |
| 84 | + ABORT_ON_FAIL(scan_builder.Project({"b"})); |
| 85 | + ABORT_ON_FAIL( |
| 86 | + scan_builder.Filter(cp::less(cp::field_ref("b"), cp::literal(4)))); |
| 87 | + let scanner = scan_builder.finish().unwrap(); |
| 88 | + return scanner.ToTable().unwrap(); |
| 89 | +} |
| 90 | + |
| 91 | +fn derive_and_rename( |
| 92 | + const std::shared_ptr<fs::FileSystem>& filesystem, |
| 93 | + const std::shared_ptr<ds::FileFormat>& format, |
| 94 | + base_dir: &str) -> std::shared_ptr<arrow::Table> { |
| 95 | + fs::FileSelector selector; |
| 96 | + selector.base_dir = base_dir; |
| 97 | + let factory = |
| 98 | + ds::FileSystemDatasetFactory::Make(filesystem, selector, format, |
| 99 | + ds::FileSystemFactoryOptions()) |
| 100 | + .unwrap(); |
| 101 | + |
| 102 | + let dataset = factory.finish().unwrap(); |
| 103 | + let scan_builder = dataset.NewScan().unwrap(); |
| 104 | + let mut names = Vec::new(); |
| 105 | + let mut exprs = Vec::new(); |
| 106 | + for field in dataset.schema().fields() { |
| 107 | + names.push(field.name()); |
| 108 | + exprs.push(cp::field_ref(field.name())); |
| 109 | + } |
| 110 | + names.emplace_back("b_as_float32"); |
| 111 | + exprs.push(cp::call("cast", {cp::field_ref("b")}, |
| 112 | + cp::CastOptions::Safe(arrow::float32()))); |
| 113 | + |
| 114 | + names.emplace_back("b_large"); |
| 115 | + // b > 1 |
| 116 | + exprs.push(cp::greater(cp::field_ref("b"), cp::literal(1))); |
| 117 | + ABORT_ON_FAIL(scan_builder.Project(exprs, names)); |
| 118 | + let scanner = scan_builder.finish().unwrap(); |
| 119 | + return scanner.ToTable().unwrap(); |
| 120 | +} |
| 121 | + |
| 122 | +fn main() { |
| 123 | + std::shared_ptr<fs::FileSystem> filesystem = |
| 124 | + std::make_shared<fs::LocalFileSystem>(); |
| 125 | + let path = create_sample_dataset(filesystem, "/home/zero/sample"); |
| 126 | + println!("{}", path); |
| 127 | + |
| 128 | + std::shared_ptr<ds::FileFormat> format = |
| 129 | + std::make_shared<ds::ParquetFileFormat>(); |
| 130 | + |
| 131 | + let table = |
| 132 | + scan_dataset(filesystem, format, "/home/zero/sample/parquet_dataset"); |
| 133 | + println!("{}", table); |
| 134 | + table = filter_and_select(filesystem, format, |
| 135 | + "/home/zero/sample/parquet_dataset"); |
| 136 | + println!("{}", table); |
| 137 | + table = derive_and_rename(filesystem, format, |
| 138 | + "/home/zero/sample/parquet_dataset"); |
| 139 | + println!("{}", table); |
| 140 | +} |
0 commit comments