quantecon/jupinx repository overview

⁠Overview

This is a docker specification which provides a uniform environment within which to generate the QuantEcon lectures (Julia⁠, Python⁠, and Data Science⁠.)

There will be a new build (retag latest) every time there's a git push to this repo.

Currently we do a new build every day, mainly to capture changes in Jupinx⁠.

⁠Usage

1. Install Docker⁠.

2. Run docker pull quantecon/jupinx.

3. In a terminal, cd to the source repository you want to build, and run docker run --name quantecon-docker -it -d -v "$(pwd)":/home/jovyan/work quantecon/jupinx from inside the directory (Linux/macOS). It should spit out a container ID string then exit. Try ${PWD} on Windows, but your mileage may vary.

   :warning: In order to guarantee reproducibility, you should either be mounting a fresh clone of the source repo (e.g. https://github.com/QuantEcon/lecture-source-jl⁠), or sanitize things by running git clean -xdff (remove uncommitted/untracked files) and git reset --hard (reset to the last git state.) Otherwise, local variance in the mounted files may impact your results.

4. In the same terminal (i.e., not inside the container), run docker exec quantecon-docker bash -c "cd work && make jupyter". (or whatever make statement you want to run.)

5. To open up an interactive REPL, run docker exec -it quantecon-docker /bin/bash.

6. After it's done, in a terminal run docker stop quantecon-docker and docker rm quantecon-docker. This will garbage-collect the container, and free the name quantecon-docker for your next run. If you're having trouble, run docker rm -f quantecon-docker to force removal.

⁠Troubleshooting

Solution: Add ARG OPENBLAS_CORETYPE=haswell to the top of the Dockerfile, if you're building. And run with the option -e OPENBLAS_CORETYPE=haswell. This is a known issue with the Julia docker, and comes from an upstream issue in OpenBLAS.