Type I and type II errors happen when you erroneously spot winners in your experiments or fail to spot them. With both errors, you end up going with what appears to work or not. And not with the real results.
Source: Type I and Type II Errors: The Inevitable Errors in Optimization Experiments, an article by Disha Sharma.
This time, we are going to describe our journey of getting Haskell garbage collection times under control when dealing with heap sizes of up to 100 GB per instance. Ultimately, we managed to reduce the garbage collection time while parsing 320,000 product rows from 68s to 3s, reducing the total parsing time from 110s to 46s.
Source: Lessons in Managing Haskell Memory, an article by Fabian Thorand and Yorick Sijsling.
As you probably know, setting things up in a developer's environment is wild — you don't know what to expect. It's hard to set things up deterministically, and reproduce one environment into another. Homebrew for instance, tries to do as best as it can, but since it treats the environment as a global space in which dumping things, alike a singleton class that anyone can modify, that often results in hard-to-debug errors.
The first time that I heard about Nix was on this blog post from Pinterest, but it didn't catch my attention until now. I started reading about it and watching some internal videos that Burke is creating to evangelize the idea. The more I read about it, the more amazed I am with the idea.
Source: Diving into Nix, an article by Pedro Piñera.