Ah I didn't actually follow the link to the github repo. I don't think that percludes pre-processing the input into something that needs no parsing, however, or to spawn a background daemon that prefetches everything in the right order.
The first run is indeed runtime, not build time, so technically I think I still count with pre-computation, sending that to the daemon (or just stashing it somewhere in /tmp, or even crazier patching the jarfile), and just printing it back out on the second run onwards.
The first run is indeed runtime, not build time, so technically I think I still count with pre-computation, sending that to the daemon (or just stashing it somewhere in /tmp, or even crazier patching the jarfile), and just printing it back out on the second run onwards.