The new Grammar::generate logic made sense. But benchmarking on my local machine showed ~55% performance impact (likely the recursive allocation of vectors). With a little refactoring, my local perf got back within noise of the performance on main branch.
The new
Grammar::generatelogic made sense. But benchmarking on my local machine showed ~55% performance impact (likely the recursive allocation of vectors). With a little refactoring, my local perf got back within noise of the performance onmainbranch.