Recent stirrings in the JavaScript Twitter community have brought a lot of attention to the shortcomings of JSON as a human-readable format, and proposed allowing JSON5 in place of it. JSON5 is a superset of JSON that allows for comments, trailing commas, and more. This is a great idea for more human-readable JSON, but I was curious about the performance implications of defaulting to JSON5 over JSON, since parsing JSON, being CPU-bound, can already be a bottleneck in many web services.

The benchmark includes 3 tests for each JSON and JSON5. All are valid JSON, which makes up the majority of JSON usage in the wild. All the test payloads have the same structure, but vary in the quantity of data they contain. The tests are as follows:

• Test 1: Small input: 4.36KiB
• Test 2: Medium input: 27.59KiB
• Test 3: Large input: 1407.77KiB

This benchmark is not meant to be a comprehensive comparison of JSON and JSON5. It is a simple benchmark that compares the performance of JSON and JSON5 in an environment where parsing JSON payloads is a common task. This does not represent any meaningful performance difference for 1-time payloads, like reading a configuration file, which JSON5 may be more targeted to.

The benchmark was run on Ubuntu 20.04, Node.js v20.12.2, on an older Intel [email protected] with 32GB of GDDR5 RAM.

npm install
npm start