Optimizing Ruby’s JSON, Part 8

It has now been about 18 months since I concluded my post series on optimising Ruby’s json. Back then, I covered interesting performance patches that happened between version 2.7.2, the last version published before I took over maintenance, and version 2.9.0, the latest release at that time. During that span, both the parser and the generator became twice as fast on the infamous twitter.json benchmark. == Parsing twitter.json (466906 bytes) ruby 4.0.6 (2026-07-14 revision 03b6d3f889) +YJIT +PRISM [arm64-darwin25] 2.7.2 575.498 (± 2.1%) i/s (1.74 ms/i) - 2.880k in 5.004363s 2.9.0 1.109k (± 0.7%) i/s (901.86 μs/i) - 5.564k in 5.017923s Comparison: 2.7.2: 575.5 i/s 2.9.0: 1108.8 i/s - 1.93x faster == Encoding twitter.json (466906 bytes) ruby 4.0.6 (2026-07-14 revision 03b6d3f889) +YJIT +PRISM [arm64-darwin25] 2.7.2 1.108k (± 0.9%) i/s (902.19 μs/i) - 5.635k in 5.083851s 2.9.0 2.219k (± 0.2%) i/s (450.69 μs/i) - 11.220k in 5.056759s Comparison: 2.7.2: 1108.4 i/s 2.9.0: 2218.8 i/s - 2.00x…