I find it ironic that the concurrent logic added by my colleague, design to (theoretically) save the a nanosecond or two from each API request, has cost me an hour and a half of my time trying to debug it. I doubt the API will see the 27,000,000,000,000 requests needed to offset these “savings”. 😕
Actually, I lie: I don’t find it ironic, I find it maddening. The parallel code is for marshalling from one in-memory data structure to another. The amount of overhead involved with dispatching these goroutines with the scheduler dwarfs any theoretical savings from doing this in parallel. It’s stupid.