parquet_derive/
lib.rs

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//! This crate provides a procedural macro to derive
19//! implementations of a RecordWriter and RecordReader
20
21#![doc(
22    html_logo_url = "https://raw.githubusercontent.com/apache/parquet-format/25f05e73d8cd7f5c83532ce51cb4f4de8ba5f2a2/logo/parquet-logos_1.svg",
23    html_favicon_url = "https://raw.githubusercontent.com/apache/parquet-format/25f05e73d8cd7f5c83532ce51cb4f4de8ba5f2a2/logo/parquet-logos_1.svg"
24)]
25#![cfg_attr(docsrs, feature(doc_cfg))]
26#![warn(missing_docs)]
27#![recursion_limit = "128"]
28
29extern crate proc_macro;
30extern crate proc_macro2;
31extern crate syn;
32#[macro_use]
33extern crate quote;
34
35use ::syn::{Data, DataStruct, DeriveInput, ext::IdentExt, parse_macro_input};
36
37mod parquet_field;
38
39/// Derive flat, simple RecordWriter implementations.
40///
41/// Works by parsing a struct tagged with `#[derive(ParquetRecordWriter)]` and emitting
42/// the correct writing code for each field of the struct. Column writers
43/// are generated in the order they are defined.
44///
45/// It is up to the programmer to keep the order of the struct
46/// fields lined up with the schema.
47///
48/// Example:
49///
50/// ```rust
51/// use parquet::file::properties::WriterProperties;
52/// use parquet::file::writer::SerializedFileWriter;
53/// use parquet::record::RecordWriter;
54/// use parquet_derive::ParquetRecordWriter;
55/// use std::fs::File;
56/// use std::sync::Arc;
57///
58/// // For reader
59/// use parquet::file::reader::{FileReader, SerializedFileReader};
60/// use parquet::record::RecordReader;
61/// use parquet_derive::ParquetRecordReader;
62///
63/// #[derive(Debug, ParquetRecordWriter, ParquetRecordReader)]
64/// struct ACompleteRecord {
65///     pub a_bool: bool,
66///     pub a_string: String,
67/// }
68///
69/// fn write_some_records() {
70///     let samples = vec![
71///         ACompleteRecord {
72///             a_bool: true,
73///             a_string: "I'm true".into(),
74///         },
75///         ACompleteRecord {
76///             a_bool: false,
77///             a_string: "I'm false".into(),
78///         },
79///     ];
80///
81///     let schema = samples.as_slice().schema().unwrap();
82///
83///     let props = Arc::new(WriterProperties::builder().build());
84///
85///     let file = File::create("example.parquet").unwrap();
86///
87///     let mut writer = SerializedFileWriter::new(file, schema, props).unwrap();
88///
89///     let mut row_group = writer.next_row_group().unwrap();
90///
91///     samples
92///         .as_slice()
93///         .write_to_row_group(&mut row_group)
94///         .unwrap();
95///
96///     row_group.close().unwrap();
97///
98///     writer.close().unwrap();
99/// }
100///
101/// fn read_some_records() -> Vec<ACompleteRecord> {
102///     let mut samples: Vec<ACompleteRecord> = Vec::new();
103///     let file = File::open("example.parquet").unwrap();
104///
105///     let reader = SerializedFileReader::new(file).unwrap();
106///     let mut row_group = reader.get_row_group(0).unwrap();
107///     samples.read_from_row_group(&mut *row_group, 2).unwrap();
108///
109///     samples
110/// }
111///
112/// pub fn main() {
113///     write_some_records();
114///
115///     let records = read_some_records();
116///
117///     std::fs::remove_file("example.parquet").unwrap();
118///
119///     assert_eq!(
120///         format!("{:?}", records),
121///         "[ACompleteRecord { a_bool: true, a_string: \"I'm true\" }, ACompleteRecord { a_bool: false, a_string: \"I'm false\" }]"
122///     );
123/// }
124/// ```
125///
126#[proc_macro_derive(ParquetRecordWriter)]
127pub fn parquet_record_writer(input: proc_macro::TokenStream) -> proc_macro::TokenStream {
128    let input: DeriveInput = parse_macro_input!(input as DeriveInput);
129    let fields = match input.data {
130        Data::Struct(DataStruct { fields, .. }) => fields,
131        Data::Enum(_) => unimplemented!("Enum currently is not supported"),
132        Data::Union(_) => unimplemented!("Union currently is not supported"),
133    };
134
135    let field_infos: Vec<_> = fields.iter().map(parquet_field::Field::from).collect();
136
137    let writer_snippets: Vec<proc_macro2::TokenStream> =
138        field_infos.iter().map(|x| x.writer_snippet()).collect();
139
140    let derived_for = input.ident;
141    let generics = input.generics;
142
143    let field_types: Vec<proc_macro2::TokenStream> =
144        field_infos.iter().map(|x| x.parquet_type()).collect();
145
146    (quote! {
147    impl #generics ::parquet::record::RecordWriter<#derived_for #generics> for &[#derived_for #generics] {
148      fn write_to_row_group<W: ::std::io::Write + Send>(
149        &self,
150        row_group_writer: &mut ::parquet::file::writer::SerializedRowGroupWriter<'_, W>
151      ) -> ::std::result::Result<(), ::parquet::errors::ParquetError> {
152        use ::parquet::column::writer::ColumnWriter;
153
154        let mut row_group_writer = row_group_writer;
155        let records = &self; // Used by all the writer snippets to be more clear
156
157        #(
158          {
159              let mut some_column_writer = row_group_writer.next_column().unwrap();
160              if let Some(mut column_writer) = some_column_writer {
161                  #writer_snippets
162                  column_writer.close()?;
163              } else {
164                  return Err(::parquet::errors::ParquetError::General("Failed to get next column".into()))
165              }
166          }
167        );*
168
169        Ok(())
170      }
171
172      fn schema(&self) -> ::std::result::Result<::parquet::schema::types::TypePtr, ::parquet::errors::ParquetError> {
173        use ::parquet::schema::types::Type as ParquetType;
174        use ::parquet::schema::types::TypePtr;
175        use ::parquet::basic::LogicalType;
176
177        let mut fields: ::std::vec::Vec<TypePtr> = ::std::vec::Vec::new();
178        #(
179          #field_types
180        );*;
181        let group = ParquetType::group_type_builder("rust_schema")
182          .with_fields(fields)
183          .build()?;
184        Ok(group.into())
185      }
186    }
187  }).into()
188}
189
190/// Derive flat, simple RecordReader implementations.
191///
192/// Works by parsing a struct tagged with `#[derive(ParquetRecordReader)]` and emitting
193/// the correct writing code for each field of the struct. Column readers
194/// are generated by matching names in the schema to the names in the struct.
195///
196/// It is up to the programmer to ensure the names in the struct
197/// fields line up with the schema.
198///
199/// Example:
200///
201/// ```rust
202/// use parquet::record::RecordReader;
203/// use parquet::file::{serialized_reader::SerializedFileReader, reader::FileReader};
204/// use parquet_derive::{ParquetRecordReader};
205/// use std::fs::File;
206///
207/// #[derive(ParquetRecordReader)]
208/// struct ACompleteRecord {
209///     pub a_bool: bool,
210///     pub a_string: String,
211/// }
212///
213/// pub fn read_some_records() -> Vec<ACompleteRecord> {
214///   let mut samples: Vec<ACompleteRecord> = Vec::new();
215///   let file = File::open("some_file.parquet").unwrap();
216///
217///   let reader = SerializedFileReader::new(file).unwrap();
218///   let mut row_group = reader.get_row_group(0).unwrap();
219///   samples.read_from_row_group(&mut *row_group, 1).unwrap();
220///   samples
221/// }
222/// ```
223///
224#[proc_macro_derive(ParquetRecordReader)]
225pub fn parquet_record_reader(input: proc_macro::TokenStream) -> proc_macro::TokenStream {
226    let input: DeriveInput = parse_macro_input!(input as DeriveInput);
227    let fields = match input.data {
228        Data::Struct(DataStruct { fields, .. }) => fields,
229        Data::Enum(_) => unimplemented!("Enum currently is not supported"),
230        Data::Union(_) => unimplemented!("Union currently is not supported"),
231    };
232
233    let field_infos: Vec<_> = fields.iter().map(parquet_field::Field::from).collect();
234    let field_names: Vec<_> = fields.iter().map(|f| f.ident.clone()).collect();
235    // unraw the identifiers, so raw identifiers like `r#type` are looked
236    // up by their column name `type` in the parquet file
237    let field_names_str: Vec<_> = fields
238        .iter()
239        .map(|f| {
240            f.ident
241                .as_ref()
242                .expect("Only structs with named fields are currently supported")
243                .unraw()
244                .to_string()
245        })
246        .collect();
247    let reader_snippets: Vec<proc_macro2::TokenStream> =
248        field_infos.iter().map(|x| x.reader_snippet()).collect();
249
250    let derived_for = input.ident;
251    let generics = input.generics;
252
253    (quote! {
254
255    impl #generics ::parquet::record::RecordReader<#derived_for #generics> for Vec<#derived_for #generics> {
256      fn read_from_row_group(
257        &mut self,
258        row_group_reader: &mut dyn ::parquet::file::reader::RowGroupReader,
259        num_records: usize,
260      ) -> ::std::result::Result<(), ::parquet::errors::ParquetError> {
261        use ::parquet::column::reader::ColumnReader;
262
263        let mut row_group_reader = row_group_reader;
264
265        // key: parquet file column name, value: column index
266        let mut name_to_index = std::collections::HashMap::new();
267        for (idx, col) in row_group_reader.metadata().schema_descr().columns().iter().enumerate() {
268            name_to_index.insert(col.name().to_string(), idx);
269        }
270
271        for _ in 0..num_records {
272          self.push(#derived_for {
273            #(
274              #field_names: Default::default()
275            ),*
276          })
277        }
278
279        let records = self; // Used by all the reader snippets to be more clear
280
281        #(
282          {
283              let idx: usize = match name_to_index.get(#field_names_str) {
284                Some(&col_idx) => col_idx,
285                None => {
286                  let error_msg = format!("column name '{}' is not found in parquet file!", #field_names_str);
287                  return Err(::parquet::errors::ParquetError::General(error_msg));
288                }
289              };
290              if let Ok(column_reader) = row_group_reader.get_column_reader(idx) {
291                  #reader_snippets
292              } else {
293                  return Err(::parquet::errors::ParquetError::General("Failed to get next column".into()))
294              }
295          }
296        );*
297
298        Ok(())
299      }
300    }
301  }).into()
302}
parquet_derive/lib.rs

parquet_derive/
lib.rs
