fast_csv

fast_csv

Classic CSV parsers suitable for most use cases. Pretty fast parsing. With experimental event-based streaming parsing.

Version: 0.2.11

About testing generated parsers

Both parsers have been tested in both operating modes, synchronous and asynchronous modes.
For testing, files from the following source are used:
CSV is a data directory which contains examples of CSV files, a flat file format describing values in a table.
https://people.sc.fsu.edu/~jburkardt/data/csv/csv.html

All files are parsed except for the malformed file mlb_players.csv.
It has unclosed quote on row 1036.

Example of the parser usage

An example of a simple way to parse 1000K rows (53M UTF-16 code units) in chunks (by ~0.53M UTF-16 code units) using events without consuming a lot of memory (during parsing) and simultaneously saving this parsed data into a virtual database (by 10K rows per transaction).

import 'dart:async';

import 'package:fast_csv/csv_converter.dart';

Future<void> main(List<String> args) async {
  _exampleParseString();
  await _exampleParseStreamWithEvents();
}

const _data = '''
Year,Make,Model,Description,Price
1997,Ford,E350,"ac, abs, moon",3000.00
1999,Chevy,"Venture ""Extended Edition""","",4900.00
1999,Chevy,"Venture ""Extended Edition, Very Large""","",5000.00
1996,Jeep,Grand Cherokee,"MUST SELL!
air, moon roof, loaded",4799.00
''';

Stream<String> _createStream() {
  // Create the stream with 1000000 rows
  const count = 1000 * 1000;
  final controller = StreamController<String>();
  final sink = controller.sink;
  const row = '1999,Chevy,"Venture В«Extended EditionВ»","",4900.00';
  const rowsInChunk = count ~/ 100;
  final chunk = List.generate(rowsInChunk, (i) => row).join('\n');
  print('Total data amount ${row.length * count} UTF-16 code units.');
  print('The data will arrive in ${chunk.length} UTF-16 code unit chunks.');
  var i = 0;
  Timer.periodic(Duration.zero, (timer) {
    sink.add(chunk);
    i += rowsInChunk;
    if (i < count) {
      sink.add('\n');
    }

    if (i >= count) {
      controller.close();
      timer.cancel();
    }
  });

  return controller.stream;
}

Future<void> _exampleParseStreamWithEvents() async {
  print('=========================');
  print('Start streaming parsing with events');
  // Get external data
  final stream = _createStream();
  final sw = Stopwatch();
  sw.start();
  final parser = _MyParser(() async {
    print('Saving to virtual database complete in ${sw.elapsed}');
    sw.stop();
  });

  await stream.transform(CsvConverter(parser: parser)).first;
}

void _exampleParseString() {
  print('=========================');
  print('Parsing string');
  final result = CsvConverter().convert(_data);
  print(result.join('\n'));
  for (var i = 1; i < result.length; i++) {
    final row = result[i];
    final car = row[1];
    final price = num.parse(row[4]);
    print('$car $price');
  }
}

class _MyParser extends CsvParser {
  final Future<void> Function()? onComplete;

  int _count = 0;

  int _totalCount = 0;

  int _transactionCount = 0;

  final List<List<String>> _rows = [];

  _MyParser([this.onComplete]);

  @override
  void beginEvent(CsvParserEvent event) {
    if (event == CsvParserEvent.startEvent) {
      _count = 0;
      _totalCount = 0;
      _transactionCount = 0;
      _rows.clear();
    }
  }

  @override
  R? endEvent<R>(CsvParserEvent event, R? result, bool ok) {
    void saveRows(bool isLast) {
      final rows = _rows.toList();
      _rows.clear();
      Timer.run(() async {
        // Asynchronous saving to the database.
        await _saveToDatabase(rows, isLast);
      });
    }

    if (ok) {
      switch (event) {
        case CsvParserEvent.rowEvent:
          final row = result as List<String>;
          _rows.add(row);
          if (_rows.length > 10000) {
            saveRows(false);
          }

          // Free memory
          result = const <String>[] as R;
          break;
        case CsvParserEvent.startEvent:
          saveRows(true);
          if (onComplete != null) {
            Timer.run(onComplete!);
          }

          // Completely freeing memory from the entire list
          result = const <List<String>>[] as R;
        default:
      }
    }

    return result;
  }

  Future<void> _saveToDatabase(List<List<String>> rows, bool isLast) async {
    _transactionCount++;
    _count += rows.length;
    _totalCount += rows.length;
    if (_count > 100000 || isLast) {
      print(
          'Saved to virtual database $_totalCount row(s) in $_transactionCount transaction(s)');
      _count = 0;
    }

    if (isLast) {
      print(
          'Totally saved to virtual database $_totalCount row(s) in $_transactionCount transaction(s)');
    }
  }
}

Output:

=========================
Parsing string
[Year, Make, Model, Description, Price]
[1997, Ford, E350, ac, abs, moon, 3000.00]
[1999, Chevy, Venture "Extended Edition", , 4900.00]
[1999, Chevy, Venture "Extended Edition, Very Large", , 5000.00]
[1996, Jeep, Grand Cherokee, MUST SELL!
air, moon roof, loaded, 4799.00]
Ford 3000.0
Chevy 4900.0
Chevy 5000.0
Jeep 4799.0
=========================
Start streaming parsing with events
Total data amount 52000000 UTF-16 code units.
The data will arrive in 529999 UTF-16 code unit chunks.
Saved to virtual database 100010 row(s) in 10 transaction(s)
Saved to virtual database 200020 row(s) in 20 transaction(s)
Saved to virtual database 300030 row(s) in 30 transaction(s)
Saved to virtual database 400040 row(s) in 40 transaction(s)
Saved to virtual database 500050 row(s) in 50 transaction(s)
Saved to virtual database 600060 row(s) in 60 transaction(s)
Saved to virtual database 700070 row(s) in 70 transaction(s)
Saved to virtual database 800080 row(s) in 80 transaction(s)
Saved to virtual database 900090 row(s) in 90 transaction(s)
Saved to virtual database 1000000 row(s) in 100 transaction(s)
Totally saved to virtual database 1000000 row(s) in 100 transaction(s)
Saving to virtual database complete in 0:00:04.087358

An example of using a configurable parser

This parser is slightly slower than the non-configurable parser.

The difference between using a normal parser and using a configurable parser is that you can specify a field separator.
Any value (such as a space or semicolon).

import 'package:fast_csv/csv_ex_converter.dart';

void main(List<String> args) {
  final parser = CsvExParser(separator: ';');
  final result = CsvExConverter(parser: parser).convert(_data);
  print(result.join('\n'));
  for (var i = 1; i < result.length; i++) {
    final row = result[i];
    final car = row[1];
    final price = num.parse(row[4]);
    print('$car $price');
  }
}

const _data = '''
Year;Make;Model;Description;Price
1997;Ford;E350;"ac, abs, moon";3000.00
1999;Chevy;"Venture ""Extended Edition""";"";4900.00
1999;Chevy;"Venture ""Extended Edition, Very Large""";"";5000.00
1996;Jeep;Grand Cherokee;"MUST SELL!
air, moon roof, loaded";4799.00
''';

About the implementation of parsers

Parsers are generated from PEG grammars.
Software used to generate parsers
Below is the source code for one of the grammars.

%%

const CsvParser();

%%

@event
Start = v:Rows Eol? @eof() ;

@inline
Chars = ($[^"]+ / '""' <String>{ $$ = '"'; })* ;

@inline
CloseQuote = '"' Spaces ;

Eol = '\n' / '\r\n' / '\r' ;

@event
@inline
Field = String / Text ;

@inline
OpenQuote = Spaces '"' ;

@event
Row = @list1(Field, ',' ↑ v:Field) ;

@inline
RowEnding = Eol !@eof() ;

Rows = v:@list1(Row, RowEnding ↑ v:Row) ;

Spaces = [ \t]* ;

String
String = OpenQuote ↑ v:Chars CloseQuote { $$ = v.join(); } ;

Text = $[^,"\n\r]* ;