Open trumank opened 9 months ago
LosslessJSON is only used to preserve Int64 numbers. JSON encoding itself does not use too much time comparing to the parsing of the Gvas I think;
I'm unsure if there are any usages of Int64 outside of timestamp which is (or could pratically be) larger than 2^53, and if there's none I'm considering switching back as the error is less than 0.05ms.
Using native JSON cut the time down from about 15 seconds to 7 seconds for me on Chrome.
LosslessJSON is only used to preserve Int64 numbers
Ah that makes sense. They could be serialized as strings to preserve precision.
Doing crude benchmarks by just observing the "deserialize" time in the console, LosslessJSON seems to take about twice as much time as native JSON.
Another interesting observation: Firefox is significantly faster than Chome and taking about 30% of the time.