This directory contains a simple command line program that compares the speed of JSON5 with the builtin JSON decoder.
On a 2018 Mac Mini with a 3 GHz 6 Core Intel Core i5 and 64 GB of memory running MacOS 14.2.1, JSON5 is from 800-1200x slower than JSON.
The three datasets come from MIT-licensed data grabbed off the web on Mar 3, 2024 around 21:30 GMT. Their accompanying licenses are contained in the LICENSE file.
64KB-min.json was retrieved from https://raw.githubusercontent.com/MicrosoftEdge/Demos/e3b81daee151a225c1d8f24bf82d31c464b0f737/json-dummy-data/64KB-min.json. It looks like that is part of a set of sample data used to benchmark Microsoft Edge‘s JSON viewer, and I’m guessing that the data was synthetically generated.
bitly-usa-gov.json was retrieved from https://raw.githubusercontent.com/wesm/pydata-book/b992071876bb4324b0323170061c886760289d4d/datasets/bitly_usagov/example.txt. and is a data set that was part of the sample data for Python Data Analysis, 3rd Edition. Apparently this data came from bit.ly data for USA.gov, although I have been unable to find either the original source data or a description of the schema for it. The book only references the ‘tz’ field, which appears to be a timezone, the ‘a’ field, which loooks like a web browser User-Agent string. The data was originally a file of newline-separated JSON records, but I merged them into a single array.
twitter.json was retrieved from https://raw.githubusercontent.com/miloyip/nativejson-benchmark/5dbf5a933a850652bf059cde64cc1f0c8d2c5d6f/data/twitter.json. It comes from what appears to be a a repo used for benchmarking for various Native C/C++ JSON parsers. I do not know where this file was retrieved from or what the schema is, but this file has a more complicated schema and contains data in multiple languages.