On Wednesday, 19 August 2015 at 13:45:32 UTC, Per Nordlöw wrote:
The parsing overhead on my machine is somewhere between 0.5 and 1 second.
I recall that Andrei has mentioned plans on further DMD optimizations that will delay lexing and/or parsing of templates until they are referenced. Only the matching ending brace of the template definition will be searched for during the primary scanning of each source file. The rest is cached as a strings associated with the corresponding template symbol until it is referenced/instantiated.
