EMBL-EBI MGnify user guides and resources

This repository contains the user documentation for MGnify, written in Markdown. It also contains example notebooks, written in Python and R, for using the MGnify API.

There are various ways to use these Notebooks, including opening them locally in an existing Jupyter environment or via cloud services.

The Docs are in the src/docs dir. The Notebooks are in the src/notebooks dir.

Quick edits to the docs

If you're making small edits to the documentation only (Markdown files), you can just use the GitHub interface to make a code branch, edit Markdown directly, and open a Pull Request.

Please check the authoring guidance below for info about features beyond standard Markdown (e.g. figures etc).

Development prerequisites

For major documentation edits or any edits to the notebooks, you need Docker installed. (Podman will also work for basics like editing the notebooks.)

You need Task installed for handy shortcut commands. If you don't want to install that, check Taskfile.yml for the long commands.

Note : To run the Docker containers on ARM-based systems, such as Apple Silicon, follow these steps:

1. Install Rosetta on your system using the command (for macOS) : softwareupdate --install-rosetta
2. Go to Docker Desktop > Settings > Features in development > Beta Features > Check "Use Rosetta for x86/amd64 emulation on Apple Silicon"

Opening the notebooks (for use or development): use Docker

Between the base image and the extra requirements (dependecies/*), the Docker contains all the libraries we need.

To add a new notebook

Python

TITLE='My New Notebook' AUTHOR='My Name' task add-py-notebook

This copies and fills in a template notebook stub file with a standard header, into src/notebooks/Python Examples.

R

TITLE='My New Notebook' AUTHOR='My Name' task add-r-notebook

This copies and fills in a template notebook stub file with a standard header, into src/notebooks/R Examples.

To open the notebooks server in edit mode

task edit-notebooks

This runs the Docker image quay.io/microbiome-informatics/emg-notebooks.dev:latest which will either by pulled from Quay, or run from your local image if you have built it (see below).

The folder src/notebooks will be mounted into the Docker container, so that changes you make are reflected in the repository.

Open a web browser to one of the URLs printed in your console to see Jupyter Lab. It should be localhost port 8888, with a random token.

When you're finished editing, use normal git add and git commit to contribute your changes.

For info, "jovyan" is always the user for these Jupyter Docker images. Jovyan as in jovian (a being from the planet Jupiter), but from Jupyter!

Guidance for authoring notebooks

• Notebooks should be complete examples, that can be run with zero code changes needed
• Notebooks should showcase good practice and use of popular libraries
• Use the Jupyter menu option to "Clear All Outputs" before checking changes into git
• Datasets should run reasonably quickly (i.e. no step should take more than a few minutes)
  • If large (slow) data fetches are needed, these should be cached in the Docker image.

Caching data in the image

MGnifyR uses a cache of pulled MGnify data. This is populated during the Docker build, into /home/jovyan/.mgnify_cache, by the script in dependencies/populate-mgnify-cache.R. Add commands to this to include other datasets in the cache. The cache is zipped and checked into the repo for faster population during builds (dependencies/mgnify-cache.tgz), since it rarely changes. To check in an updated version of the cache...

task update-mgnifyr-cache
git add dependencies/mgnify-cache.tgz

Changing dependencies and Docker build

You can temporarily try things by opening a Terminal inside Jupyter Lab and mambda installing the package(s). But make sure you reflect everything in the conda environment file.

Then check the environment builds by (re)building the Docker:

task build-notebook-docker

Generating the documentation site

There is a setup to use Quarto to render the notebooks – including inputs and outputs – as well as the static documentation, to a documentation website. This allows us to use both the docs and notebooks interchangebly to provide user guidance.

There is a Dockerfile to add Quarto on top of the regular Docker stack: docker/docs.Dockerfile.

To preview the statically rendered notebooks, use:

task render-static

This builds a docker image tagged as notebooks-static, runs Quarto inside it, executes all cells of the notebooks, and renders the completed notebooks to the _site folder (which is mounted from your host machine into Docker).

You can browse the generated HTML with

task serve-static

Or use

task preview-static

to render the notebook in watch-mode and serve them to serve a preview of them.

Docs authoring guidance

Mostly, you can just use normal Markdown – but there are some handy extra features. We frequently use: YAML Front Matter (metadata for each docs page); Callout Blocks; Figures. E.g.:

---
title: My new docs page
author:
  - name: MGnify
    url: https://www.ebi.ac.uk/metagenomics
    affiliation: EMBL-EBI
    affiliation-url: https://www.ebi.ac.uk
date: last-modified
citation: true
description: This page explains everything
---

## Steps to do everything

::: {.callout-warning}

### By the way

This sentence will be rendered as an attention-grabbing box
:::

![This figure shows everything](images/mypage/everything.png){#fig-my-everything .tall-figure fig-align="left"}

As you can see in @fig-my-everything, ...

Note the use of .tall-figure fig-align="left": that is a styling hack for figures that are a tall aspect ratio (e.g. a vertical flow diagram, or a scrolled-page screenshot). It makes those images appear less overwhelming. Don't use it for figures that are square or wide.

There are also examples within the existing docs or on the Quarto website for how to do subfigures (panels), annotated code blocks, citations etc.

Shiny-Proxy application

The notebooks can also be built into a Docker container suitable for running as an Application on ShinyProxy. The configuration for this is in the shiny-proxy dir.

Testing with a locally built Docker

task build-notebook-docker

Running ShinyProxy

• Download the latest version of ShinyProxy (>=2.6 is required). It is a JAR, so you need Java installed. i.e., download ShinyProxy into this repo directory.
• The application.yml file must be in the same directory as the location you launch Shiny Proxy from.
• If you want the currently deployed image instead of your local one... docker pull quay.io/microbiome-informatics/emg-notebooks.dev:latest
• cd shiny-proxy, java -jar shinyproxy-2.6.1.jar
• Browse to the ShinyProxy URL, likely localhost:8080

Jupyter Lab Extension, for deep-linking

This extension was created using the JupyterLab Extension Template copier project, which is CC0 licensed.

This extenion is needed because Shiny Proxy does not pass the URL path beyond an app's identifier down to the iframe running the app (JupyterLab).

The extension does two things:

1. Allows deeplinking to notebooks. It watches for a URL querystring parameter, ?jlpath=, and uses this to forward Jupyter Lab to an internal URI which activates that path. This works because Shiny Proxy does include the query params in the iframe. E.g. browsing to localhost:8080/app/mgnify-notebook-lab?jlpath=notebooks/home.ipynb will trigger Jupyter Lab to open the home.ipynb notebooks once the application initialises. This is a frontend extension in shiny_proxy_jlab_query_params/src/index.ts.
2. Sets ENV VARs based on query params. On Jupyter Lab launch, it sends the querystring parameters to a Jupyter Lab "server-side" extension handler. The handler takes any querystring params beginning ?jlvar_ and sets corresponding ENV VARs. E.g., ?jlvar_MGYS=MGYS007 results in an ENV VAR of MGYS=MGYS007 being available to any kernels launched after this. These ENV VARs can then, of course, be read in Notebooks (e.g. os.getenv('MGYS') in Python or Sys.getenv('MGYS') in R). There are helper utilities for both R and Python, that try to read such an ENV VAR otherwise ask for user input. These are in src/notebooks/{Python Examples | R Examples}/lib/variable_utils.{R|Python}.

Together, this means a URL like: localhost:8080/app/mgnify-notebook-lab?jlpath=notebooks/home.ipynb&jlvar_MGYS=MGYS00005116 will trigger Shiny Proxy to start the container, then open the home notebook, and have an MGYS environment variable ready to use in code.

Jupyter Lab Extension, for MGnify-specific help

There is also an extension to render a MGnify-specific help pane and menu inside Jupyter Lab. This is in the mgnify_jupyter_lab_ui folder.

Testing

A small integration test suite is written using Jest-Puppetteer. You need to have built or pulled the docker/Dockerfile (tagged as quay.io/microbiome-informatics/emg-notebooks.dev:latest), and have Shiny Proxy downloaded first. The test suite runs Shiny Proxy, and makes sure Jupyter Lab opens, the deep-linking works, and variable insertion works in R and Python.

cd tests
npm install
npm test

Deployment

There is an on-push build trigger for this repository that builds images to quay.io/quay.io/microbiome-informatics/emg-notebooks.dev

Just push to the repository (all branches are built and tagged). If you push to the main branch, the :latest tag will point to that version, once it is built.

This built image can be (and is) deployed to multiple servers, e.g. a ShinyProxy instance or the Galaxy project. EMBL's ShinyProxy updates to the latest semver tag at 0300 daily. For Galaxy, open a PR to edit the tool schema.

Releases

Tag releases with semver tags, like v1.0.0.

• Increment by 0.0.1 for bugfixes, new notebooks, new packages etc.
• Increment by 0.1 for changes in the base docker image, or refactors of the docker/environments etc.
• Increment by 1 for changes in how the entire stack is run.

Contributors ✨

Thanks goes to these wonderful people (emoji key):

Sandy Rogers 💻 💡 🤔 🚇 👀	Ales-ibt 💻 💡 🤔	Virginie Grosboillot 🤔 💻 🖋	Björn Grüning 🚇	Bérénice Batut 🚇	Martín Beracochea 🤔 💻 🖋 🧑‍🏫	tgurbich 🤔 💻 🖋
Amartya Nambiar 💻 💡 🤔	Ekaterina Sakharova 🤔 💻 🖋

This project follows the all-contributors specification. Contributions of any kind welcome!