I have a lambda function with a graphql server that processes in ~ 1-2ms. If I randomly throw an error in it the response time goes up to to ~150ms. I suspect the amount of delay is correlated with sourcemap size. I believe inspecting or traversing the error as graphql does to generate messages is causing this but need to confirm So now I'm back to recommending disabling --enable-source-maps in production unless you have benchmarked it with exceptions being thrown

Should depend on the error frequency, no? If I'm getting a few an hour, that level of latency is worth the readable stack trace. If I'm getting 100 errors a second...maybe fix the errors? I've been recommending sourcemaps on.

The issue is I use them for control flow not just pure exceptions

Ah okay, well then yes, you're going to have to take a lot of latency. I never do that due to horrible experiences with spring framework 😛

You're right though that it still is the minority of situations where this even happens

I did some benchmarks for sourcemaps and I believe they add latency, but only on exceptions. When did you start supporting them? Must've been recent (and I missed it).

It's sort of an invisible feature, we always dump sourcemap into the bundle

I just found it in the docs. Cool.

Could be worse, I've had a fairly-large (i.e. not one function) project using Bluebird breaching the 5s

jest

limit 😄