// Minimal example for Arrow CSV parsing bug
// Only crate dependency is Arrow - tested against 43.0.0 and 46.0.0

use std::{io::Cursor, sync::Arc}; 

use arrow::{
    datatypes::{Schema, Field, DataType}, 
    csv::reader::ReaderBuilder
};

fn main() {
    
    // Simple three-column schema
    let schema = Schema::new(vec![
        Field::new("field 1",     DataType::Utf8, false),
        Field::new("string list", DataType::Utf8, false),
        Field::new("field 3",     DataType::Utf8, false)
    ]);
    
    // CSV sample - first record OK, second record doesn't quote the string
    // list so ends up with too many columns when parsed. 
    let sample = String::from("something,\"this,is,a,list\",something\nsomething,this,is,another,list,whoops\n").into_bytes(); 
    
    // Build a CSV reader to read the sample using the provided schema
    let reader = ReaderBuilder::new(Arc::new(schema))
        .build(Cursor::new(sample))
        .unwrap();

    // Reader errors on row 2 because it has too many columns and doesn't fit
    // the schema. The reader doesn't move on to the next row after the 
    // error, it reads in more bytes and keeps adding them to the current line. 
    // This eventually causes an out of memory error; the reader never exits 
    // because it never reaches end of stream.
    //
    // Uncomment the println! statement to show the CSVError. (Leaving the 
    // println! commented out demonstrates the OOM quicker.)
    for batch in reader {
        match batch {
            Ok (next)  => println!("Got a batch of {} records", next.num_rows()),
            Err(_error) => {
                //println!("{:?}", _error); 
                continue;
            }
        }
    }
    
}