Currently compile times for larger projects are acceptable, but not too good.
In fact, I'm working on 2 features that improve the situation:
* Deterministic code generation: Currently Nim uses IDs to generate unique C
names. These IDs are generated by incrementing a global counter, so if you
insert a Nim symbol somewhere all following symbols get a fresh ID causing much
more C code to be recompiled than necessary. Instead the compiler will generate
a unique hash based on the proc and type signatures. This means the C code
cache ("nimcache") will become much more effectively.
* Symbol files: A symbol file is a binary representation of a Nim module
which can be loaded on demand, no parsing or semantic checking is required,
speeding up compile times tremendously.
