Wow @Ward, you are a genius. Pre-allocating the string is much much faster! 
Your `newStringOfCap(64)` led me on this path, where turns out it is faster to 
scan the string figure out the exact size then populate the string using a 
second pass. The memory is warm any ways! When I populate the string in the 
second pass I can use unchecked arrays (as the size is well known) for less 
checks and even more speed. Thank to you I got this benchmark to 17.7ms. Way 
faster then TypeScript and still just using plain -d:danger and --gc:arc. 
Thanks!

Reply via email to