Springbatch read from csv, how does it work?

I am new to springbatch, and I wonder how this reader/processor/writer works if I am reading a csv file which contains 10k rows, use a chunk size of 10 and output to a csv file.

My questions is:

Does springbatch loads all 10k rows from csv in one time, process individually(10k times), and then store all of them into the destination file in one go? If so, what's the point of using springbatch? I can have three methods doing the same job right?

Or:

Does springbatch opens up a stream reading 10k rows from csv, each time it reads 10 rows, process 10 rows, and open a output stream write/append those 10 rows into destination file? Basically repeats 10k/10 = 1k times.

@Configuration
public class SampleJob3 {

    @Bean
    public Job job3(JobRepository jobRepository, PlatformTransactionManager transactionManager) {
        return new JobBuilder("Job3", jobRepository)
                .incrementer(new RunIdIncrementer()) // work with program args
                .start(step(jobRepository, transactionManager))
                .build();
    }

    private Step step(JobRepository jobRepository, PlatformTransactionManager transactionManager) {
        return new StepBuilder("Job3 Step started ")
                .<Student, Student>chunk(3)
                .repository(jobRepository)
                .transactionManager(transactionManager)
                .reader(reader(true))
                .processor(student -> {
                    System.out.println("processor");
                    return new Student(student.getId(), student.getFirstName() + "!", student.getLastName() + "!", student.getEmail() + "!");
                })
                .writer(writer())
                .build();
    }


    private FlatFileItemReader<Student> reader(boolean isValid) {
        System.out.println("reader");
        FlatFileItemReader<Student> reader = new FlatFileItemReader<>();
        // using FileSystemResource if file stores in a directory instead of resource folder
        reader.setResource(new PathMatchingResourcePatternResolver().getResource(isValid ? "input/students.csv" : "input/students_invalid.csv"));
        reader.setLineMapper(new DefaultLineMapper<>() {
            {
                setLineTokenizer(new DelimitedLineTokenizer() {{
                    setNames("ID", "First Name", "Last Name", "Email");
                }});

                setFieldSetMapper(new BeanWrapperFieldSetMapper<>() {{
                    setTargetType(Student.class);
                }});
            }
        });
        reader.setLinesToSkip(1);
        return reader;
    }

    //@Bean
    public FlatFileItemWriter<Student> writer() {
        System.out.println("writer");
        FlatFileItemWriter<Student> writer = new FlatFileItemWriter<>();
        writer.setResource(new FileSystemResource("output/students.csv"));
        writer.setHeaderCallback(writer1 -> writer1.write("Id,First Name,Last Name, Email"));
        writer.setLineAggregator(new DelimitedLineAggregator<>() {{
            setFieldExtractor(new BeanWrapperFieldExtractor<>() {{
                setNames(new String[]{"id", "firstName", "lastName", "email"});
            }});
        }});

        writer.setFooterCallback(writer12 -> writer12.write("Created @ " + Instant.now()));
        return writer;
    }


}

My last question basically the same, but datasource is database. e.g. reading a table contains 10k data from dbA and write to dbB. Am I able to read 10 rows at a time, process them and write them to dbB? If so, can you share some sudocode?

A chunk-oriented step in Spring Batch will not read the entire file or table at once. It will rather stream data from the source in chunks (of a configurable size).

You can find more details about the processing model in the reference documentation here: Chunk-oriented Processing.