gRPC defined 18 status codes for returning different types of errors.

I think they’re a pretty good reference if you want to design clear error handling for API.

Context: Machine Learning Projects are mainly just a complex interdependent pipeline. We desperately need a better abstraction for them.

I spent most of my life building glue systems.

Sometimes, I got to work on a deep, sparkling project. But most of the time, my work is about gluing things together.

Cardsjl is a simple Julia package which demonstrate many interesting bits of the Julia Programming Language.

Reading it is an enjoyable experience. The note I’ve writtend down is here:

The title sounds obvious. We always program for a machine! A program won’t run itself!

Probably the cleanest way to do it: (pushnew (truename "/projects/app/") ql:*local-project-directories* ) (ql:register-local-projects) (ql:quickload :app) From Use Quicklisp to load personal projects from arbitrary locations. Alternatively, you can create a symlink in ~/quicklisp/local-projects/. » Read More

OCaml’s module system can be a powerful tool for building generic code and structuring systems. Functors are functions from modules to modules and they serve an important role for the power of module system. However, I want to know if functors (and the module system) can be optimized away by the OCaml compiler. » Read More

Modern compilers usually composed by multiple stages: parsers, optimizers, linkers, and assemblers. Let’s go through it one by one to have a better understanding of the OCaml compiler. » Read More

eqaf, a constant-time compare function implementation in OCaml, is a great case to demonstrate the predictable performance of OCaml’s compiler. Why? » Read More