Q&A

How to parse a CSV file in Java

By AlgoInsight Team2026-06-07java

Overview

CSV ("comma-separated values") looks simple but has subtle rules: fields may be quoted, quoted fields may contain commas or newlines, and quotes inside fields are escaped by doubling. A naive split(",") ignores all of this. For anything beyond toy data, use a library that implements the CSV format (RFC 4180) correctly.

graph TD
    A[Parse CSV] --> B[Trivial, no quotes/commas-in-fields]
    A --> C[Real-world CSV]
    B --> B1[BufferedReader + split(',')]
    C --> C1[Apache Commons CSV]
    C --> C2[OpenCSV]
    style C1 fill:#1e1b4b,color:#fff
    style C2 fill:#1e1b4b,color:#fff

Approach 1: BufferedReader + split (simple files only)

import java.io.*;
import java.nio.file.*;

try (BufferedReader br = Files.newBufferedReader(Paths.get("data.csv"))) {
    String line;
    boolean header = true;
    while ((line = br.readLine()) != null) {
        if (header) { header = false; continue; }   // skip header row
        String[] fields = line.split(",", -1);       // -1 keeps trailing empties
        String name = fields[0];
        String email = fields[1];
        // ...
    }
}

This works only if no field contains a comma, quote, or newline. The -1 limit to split preserves trailing empty fields (otherwise they're dropped). Convenient but fragile.

Why split(",") Fails on Real CSV

"Smith, John","john@x.com","Says ""hi"""

A single record where the first field contains a comma and the third contains escaped quotes. split(",") would wrongly break "Smith and John" into two fields. Correct parsing requires honoring quotes - exactly what libraries do.

Approach 2: Apache Commons CSV (recommended)

import org.apache.commons.csv.*;
import java.io.*;
import java.nio.file.*;

try (Reader reader = Files.newBufferedReader(Paths.get("data.csv"));
     CSVParser parser = CSVParser.parse(reader,
         CSVFormat.DEFAULT.builder().setHeader().setSkipHeaderRecord(true).build())) {

    for (CSVRecord record : parser) {
        String name  = record.get("name");      // access by header name
        String email = record.get("email");
        // or by index: record.get(0)
    }
}

It handles quoted fields, embedded commas/newlines, escaped quotes, and header-based access automatically.

Approach 3: OpenCSV

import com.opencsv.CSVReader;
import java.io.*;

try (CSVReader reader = new CSVReader(new FileReader("data.csv"))) {
    String[] row;
    reader.readNext();                  // skip header
    while ((row = reader.readNext()) != null) {
        String name = row[0];
        // ...
    }
}

OpenCSV also supports bean mapping - binding rows directly to Java objects via annotations:

List<Person> people = new CsvToBeanBuilder<Person>(reader)
    .withType(Person.class).build().parse();

Choosing an Approach

• split(",") - only for throwaway scripts on data you fully control (no quotes/commas in fields).
• Apache Commons CSV - robust, header access, great general choice.
• OpenCSV - robust, plus convenient bean (object) mapping.

Practical Considerations

• Encoding - specify charset (e.g. UTF-8) when reading; watch for a BOM.
• Headers - use header-aware parsing to access fields by name (resilient to column reordering).
• Large files - stream record-by-record (as above) rather than loading everything into memory.
• Delimiters - real "CSV" files sometimes use ; or tabs; configure the format accordingly.

Why This Matters in Interviews

This tests file I/O and awareness of real-world data messiness. The key insight interviewers want: split(",") is wrong for general CSV because it can't handle quoted fields containing commas, newlines, or escaped quotes - so use Apache Commons CSV or OpenCSV. Demonstrating try-with-resources for the reader, header-based access, streaming for large files, and encoding awareness signals production experience.

Common Pitfalls

• Using split(",") and breaking on quoted fields with embedded commas/newlines.
• split(",") dropping trailing empty fields (pass -1 as the limit).
• Not skipping the header row, or hard-coding column indices that shift.
• Ignoring character encoding (mojibake / BOM issues).
• Loading a huge file fully into memory instead of streaming records.

References

• Apache Commons CSV: https://commons.apache.org/proper/commons-csv/
• OpenCSV: https://opencsv.sourceforge.net/

More answers