CSV to Array of Arrays (No Headers)

CSV to Array of Arrays

Not every CSV-to-JSON conversion needs named keys. Sometimes the most appropriate output is a simple array of arrays, where each inner array represents one row of values.

Input CSV

2024-01,150,89.5
2024-02,203,92.1
2024-03,178,87.3

Output JSON

[
  ["2024-01", "150", "89.5"],
  ["2024-02", "203", "92.1"],
  ["2024-03", "178", "87.3"]
]

When to use array-of-arrays format

1. Headerless CSV data. Sensor readings, log entries, and matrix data often lack a header row. Forcing artificial key names adds no value.
2. Compact output. By omitting repeated key names, the JSON output can be 30-50% smaller. For a CSV with 10 columns and 1,000 rows, that eliminates 10,000 repeated key strings.
3. Grid/table rendering. Frontend table components often accept [headers[], ...rows[][]] format directly.
4. Data pipelines. When the consumer already knows the schema, named keys are unnecessary overhead.

Implementation

function csvToArrays(csv) {
  return csv
    .trim()
    .split("\n")
    .map(line => line.split(",").map(cell => cell.trim()));
}

With optional header separation

If the CSV does have headers but you want arrays:

{
  "headers": ["month", "sales", "satisfaction"],
  "rows": [
    ["2024-01", "150", "89.5"],
    ["2024-02", "203", "92.1"]
  ]
}

This hybrid format gives you the compactness of arrays with the self-documenting benefit of header names. It is commonly used by charting libraries and spreadsheet APIs.

Caveats

• Column order matters. Unlike objects, arrays are positional, so reordering columns changes the meaning.
• Missing values must be represented by empty strings or null placeholders to maintain alignment.
• Consumers must reference values by index (row[0]) rather than by name (row.month), which is less readable.