Writing by Peter Hilton

Why CSV survives

The enduring nature of comma-separated values (CSV) - 21 December 2021 #data

Classic car, old building with columns

Serge Kutuzov

Although comma-separated values (CSV) files often don’t work, because they have the wrong delimiters or the wrong encoding, CSV lives on. CSV has retained its popularity for decades, and continues to do so despite a series of potential replacements.

None of the CSV replacements replaced CSV

In the decades since CSV first gained popularity, many alternative text-based data interchange formats appeared. Most improved on CSV by avoiding its ambiguities, or escaping its limitations.

XML has risen and fallen. JSON is just a flash in the pan. YAML is a poisoned chalice. CSV will outlast them all.
- Leon Bambrick, Awesome CSV

But cinema didn’t replace radio, and television didn’t replace cinema. New options add choices, and obsolescence occurs separately. In particular, new data formats aimed for many goals, but never more convenience.

Convenience beats quality

Despite losing the fashionability it had twenty years ago, XML succeeds where it offers more convenience than legacy binary formats, that required far more specification. However, in the world of text-based formats that we send over HTTP, XML occupies a kind of high-quality high-complexity high ground compared to the 1990s’ text formats.

XML provides a proper engineering solution to data exchange, so we use JSON instead. JSON gives software developers a more convenient experience by not addressing the hard problems that XML took on. In general, convenience beats quality, because inconvenience usually costs more than an imperfect solution.

People prefer tabular data

Most people associate the word data with a two-dimensional table. A one-dimensional list feels like something simpler, and hierarchies belong to people with exposure to recursion and nested data structures, such as programmers. While programming languages naturally support hierarchical data structures, genealogists and human resources departments need specialist software for their family trees and organisation charts.

Everyone else uses spreadsheets. People (non-programmers) generally prefer tabular data to more exotic structures. For many of them, data only exists in spreadsheets, which you can only share as CSV or a proprietary spreadsheet file format.

Excel can’t import JSON

JSON deserves most of its popularity to how it just works in JavaScript. JSON doesn’t work with Excel, though. And while CSV arguably doesn’t work particular well in Excel either, it works well enough.

If Excel can’t import your data format, does it even exist?

Don’t expect CSV to go away any time soon. Instead, make sure that the software you build uses it properly.