thanks for the outlook, good to know.

Did you try the latest version? At least with LDC and using full
optimizations we finally got some numbers that were faster than Crystal
or RapidJSON. DMD is still behind RapidJSON, but I think it was on par
with Crystal. There is still potential for optimization, so I'm pretty
sure that we'll have a very compelling package in the end.

yes, after some profiling (and trying out the byLineFast implementation that is floating in the dlang forum) i realized that most of my lost time was due to inefficient by line parsing in std.stdio. bummer, i totally did not expect this.

so yep, its actually now on par/faster than crystal. i think its pretty awesome that you can get C speed with very clean and dense code.