Skip to content

Latest commit

 

History

History
517 lines (340 loc) · 28.3 KB

developer-guide.md

File metadata and controls

517 lines (340 loc) · 28.3 KB
Raw

Developer guide

This guide describes:

  • how to download source code for Nextclade CLI and Nextclade Web
  • how to setup a development environment
  • how to build and run Nextclade CLI and Nextclade Web
  • how the official distributions are maintained, released and deployed

This is only useful if you know programming at least a little or is curations about how Nextclade is built.

⚠️ If you are Nextclade user or is looking to familiarize yourself with Nextclade usage and features, then refer to Nextclade user documentation instead.

⚠️ This guide assumes basic familiarity with Nextclade Web and/or Nextclade CLI as well as certain technical skills.

⚠️ Datasets are managed in a separate repository

Developing locally

Nextclade CLI

Nextclade CLI is written in Rust programming language. The usual rustup & cargo workflow can be used.

If you are not familiar with Rust, please refer to documentation:

  • Rust - the programming language itself
  • Rustup - Rust toolchain installer and version manager
  • Cargo - Rust package manager

as well as to the --help text for each tool.

Steps

  1. Obtain source code (once)

    Make sure you have git installed.

    Clone Nextclade git repository:

    git clone https://github.com/nextstrain/nextclade
cd nextclade

    💡 We accept pull requests on GitHub. If you want to submit a with new feature or a bug fix, then make a GitHub account, make a fork of the origin Nextclade repository and clone your forked repository instead. Refer to GitHub documentation "Contributing to projects" for more details.

    💡 Make sure you keep your local code up to date with the origin repo, especially if it's forked.

    💡 If you are a member of Nextstrain team, then you don't need a fork and you can contribute directly to the origin repository. Still, in most cases, please submit pull requests for review, rather than pushing changes to major branches directly.

  2. Install Rust if not already (https://www.rust-lang.org/tools/install):

    # [once] Install Rustup, the Rust version manager
curl --proto '=https' --tlsv1.2 -sSf https://sh.rustup.rs | sh

# [once] Add Rust tools to the $PATH
export PATH="$PATH:$HOME/.cargo/bin"

# [once] [Linux only] install openssl and pkgconfig. Example for Ubuntu:
sudo apt-get update
sudo apt-get install --yes libssl-dev pkg-config

# Check your installed versions of Rust compiler, Cargo and Rustup
$ rustc -V
$ cargo -V
$ rustup -V

    ⚠️ We don't support Rust installations deviating from the officially recommended steps. If you install Rust from Linux OS package repositories, Homebrew, Conda etc., things may or may not work, or they may work but produce wrong results. Nextclade team don't have bandwidth to try every platform and config, so if you decide to go unofficial route, then you are on your own. But feel free to open pull requests with fixes, where necessary.

    💡 Note, Rustup allows to install multiple versions of Rust and to switch between them. This repository contains a rust-toolchain.toml file, which describes which version of Rust is currently in use by Nextclade official build. Cargo and Rustup should be able to pick it up automatically, install the required toolchain and use it when you type cargo commands. Any other versions of Rust toolchain are not supported.

  3. Prepare environment variables which configure Nextclade build-time settings (once). Optionally adjust the variables in the .env file to your needs.

    # [once] Prepare dotenv file with default values
cp .env.example .env

  4. Build and run Nextclade CLI in debug mode (convenient for development - faster to build, slow to run, has more debug info in the executable, error messages are more elaborate):

    # (Re-)build Nextclade in debug mode.
# By default, the resulting executable will be in `target/debug/nextclade`.
cargo build --bin=nextclade

# (Re-)build Nextclade in debug mode and run `nextclade --help` to print Nextclade CLI main help screen. The arguments after the `--` are passed to nextclade executable, as if you'd run it directly. You can also refer to Nextclade user documentation (https://docs.nextstrain.org/projects/nextclade/en/stable/index.html) for explanation of arguments.
cargo run --bin=nextclade -- --help

# (Re-)build Nextclade in debug mode and use it to download a dataset to `data_dev/` directory.
cargo run --bin=nextclade -- dataset get \
  --name='sars-cov-2' \
  --output-dir='data_dev/sars-cov-2'

# (Re-)build Nextclade in debug mode and run the analysis using the dataset we just downloaded (to `data_dev/`) and output results to the `out/` directory.
cargo run --bin=nextclade -- run \
  'data_dev/sars-cov-2/sequences.fasta' \
  --input-dataset='data_dev/sars-cov-2/' \
  --output-all='out/'

    💡 Note, depending on your computer hardware and internet speed, your first build can take significant amount of time, because the necessary Rust toolchain version and all dependency packages (crates) will be downloaded and compiled. Next time the existing toolchain and cached packages are used, so the repeated builds should be much faster.

    💡 Add -v to Nextclade arguments to make console output more verbose. Add more occurrences, e.g. -vv, for even more verbose output.

  5. Build and run Nextclade CLI in release mode (slow to build, fast to run, very little debug info):

    # Build Nextclade in release mode.
# By default, the resulting executable will be in `target/release/nextclade`.
cargo build --bin=nextclade --release

# Run Nextclade release binary
./target/release/nextclade run \
  'data_dev/sars-cov-2/sequences.fasta' \
  --input-dataset='data_dev/sars-cov-2' \
  --output-fasta='out/nextclade.aligned.fasta' \
  --output-tsv='out/nextclade.tsv' \
  --output-tree='out/nextclade.tree.json' \
  --in-order \
  --include-reference

    💡 Debug builds are incremental, i.e. only the files that have changed since the last build are compiled. But release builds are not. If you need to quickly iterate on features, then use debug builds. If you are measuring performance, or make a build for daily usage, always use release builds.

Nextclade Web

Nextclade Web is a React & Typescript application, which relies on Nextclade WebAssembly (wasm) module to perform the computation. This WebAssembly module shares the same Rust code for algorithms as Nextclade CLI. So building Nextclade Web involves 2 steps:

  • building WebAssembly module
  • building the web application itself

Install Node.js version 14+ (latest LTS release is recommended), by either downloading it from the official website: https://nodejs.org/en/download/, or by using nvm.

⚠️ We don't have bandwidth to support Node.js installations from Linux OS package repositories, Homebrew, Conda and everything else deviating from the officially recommended setup. If you decide to go that route - things may or may not work - you are on your own. But feel free to open pull requests with fixes if necessary.

Steps

  1. Obtain source code (once)

    Make sure you have git installed.

    Clone Nextclade git repository:

    git clone https://github.com/nextstrain/nextclade
cd nextclade

    💡 We accept pull requests on GitHub. If you want to submit a with new feature or a bug fixe, then make a GitHub account, make a fork of the origin Nextclade repository and clone your forked repository instead. Refer to GitHub documentation "Contributing to projects" for more details.

    💡 Make sure you keep your local code up to date with the origin repo, especially if it's forked.

    💡 If you are a member of Nextstrain team, then you don't need a fork and you can contribute directly to the origin repository. Still, in most cases, please submit pull requests for review, rather than pushing changes to branches directly.

  2. Install Rust if not already (https://www.rust-lang.org/tools/install):

    # [once] Install Rustup, the Rust version manager
curl --proto '=https' --tlsv1.2 -sSf https://sh.rustup.rs | sh

# [once] Add Rust tools to the $PATH
export PATH="$PATH:$HOME/.cargo/bin"

# [once] [Linux only] install openssl and pkgconfig. Example for Ubuntu:
sudo apt-get update
sudo apt-get install --yes libssl-dev pkg-config

# Check your installed versions of Rust compiler, Cargo and Rustup
$ rustc -V
$ cargo -V
$ rustup -V

    ⚠️ We don't support Rust installations deviating from the officially recommended steps. If you install Rust from Linux OS package repositories, Homebrew, Conda etc., things may or may not work, or they may work but produce wrong results. Nextclade team don't have bandwidth to try every platform and config, so if you decide to go unofficial route, then you are on your own. But feel free to open pull requests with fixes, where necessary.

    💡 Note, Rustup allows to install multiple versions of Rust and to switch between them. This repository contains a rust-toolchain.toml file, which describes which version of Rust is currently in use by Nextclade official build. Cargo and Rustup should be able to pick it up automatically, install the required toolchain and use it when you type cargo commands. Any other versions of Rust toolchain are not supported.

  3. Prepare environment variables which configure Nextclade build-time settings (once). Optionally adjust the variables in the .env file to your needs.

    cp .env.example .env

  4. Install other required tools (once)

    cargo install wasm-pack
    🍏 Extra requirements for macOS [click to expand]

    For macOS, you will also have to install llvm:

    brew install llvm

    Furthermore, you will need to set the following environment variables before invoking yarn wasm-prod:

    export CC=/opt/homebrew/opt/llvm/bin/clang
export AR=/opt/homebrew/opt/llvm/bin/llvm-ar

  5. Install NPM dependencies (once)

    cd packages/nextclade-web
yarn install

    ⚠️ Nextclade uses yarn to manage NPM dependencies. While you could try npm or other tools instead, we don't support this.

  6. Build the WebAssembly module

    cd packages/nextclade-web
yarn wasm-prod

    This step might take a lot of time. The WebAssembly module and accompanying Typescript code should be been generated into packages/nextclade-web/src/gen/. The web application should be able to find it there.

    Repeat this step every time you are touching Rust code.

  7. Build and serve the web app

    We are going to run a development web server, which runs continuously (it does not yield terminal prompt until you stop it). It's convenient to do it in a separate terminal instance from WebAssembly module build to allow rebuilding the app and the module independently.

    The development version can be built using:

    cd packages/nextclade-web
yarn dev

    Open http://localhost:3000/ in the browser. Typescript code changes should trigger rebuild and fast refresh of the app. If you rebuild the WebAssembly module (ina separate terminal instance), it should also pick up the changes automatically.

    Alternatively, the optimized ("production") version of the web app can be built and served with

    yarn prod:build
yarn prod:serve

    Open http://localhost:8080/ in the browser.

    The resulting HTML, CSS and JS files should be available under packages/nextclade-web/.build/production/web.

    Production build does not have automatic rebuild and reload. You need to do full rebuild on every code change.

    The yarn prod:serve command runs Express underneath and it is just an example of a simple (also slow and insecure) local file web server. But the produced HTML, CSS and JS files can be served using any static file web server or static file hosting service. The official deployment uses AWS S3 + Cloudfront.

Linting (static analysis)

Linting Rust code

Rust code is linted with Clippy:

cargo clippy

Automatic fixes can be applied using:

cargo clippy --fix

Clippy is configured in clippy.toml and in .cargo/config.toml.

For routine development, it is recommended to configure your text editor to see the Rust compiler and linter errors.

💡 In VSCode [click to expand]

(these instructions can go out of date with time, so make sure you check VSCode community for what's latest and greatest)

Make sure you have "Rust Analyzer" extension (and not deprecated "Rust" extension), and configure it to use clippy: hit Ctrl+Shit+P, then find "Preferences: Open user settings (JSON)", then add:

"rust-analyzer.check.command": "clippy",

Now the warnings and errors will be shown as yellow and red squiggles. If you mouse hover a squiggle, there will appear a tooltip with explanation and a link to even more details. Sometimes there will be a link in the bottom of the tooltip to apply a "Quick fix" for this particular mistake. And there is also a light bulb in the editor to do the same.

You can disable the pesky inline type hints (for all languages) by adding this to your preferences JSON:

"editor.parameterHints.enabled": false,
"editor.inlayHints.enabled": "off",

An extension "Error lens" allows to see error and warning text inline in the editor.

💡 In Jetbrains CLion [click to expand]

(these instructions can go out of date with time, so make sure you check Jetbrains docs for what's latest and greatest)

Install Intellij Rust plugin.

In main menu, "File | Settings | Languages & Frameworks | Rust | External Linters", set "External tool" to "Clippy" and check the checkbox "Run external linter to analyze code on the fly".

You should now see red and yellow squiggles if there are problems. Mouse hover to read the message and recommendations.

Install Inspection Lens plugin to see the messages inline in the code.

Linting Typescript and JavaScript

The web app is linted using eslint and tsc as a part of development command, but the same lints also be run separately:

cd packages/nextclade-web
yarn lint

The eslint configuration is in .eslintrc.js. tsc configuration is in tsconfig.json.

Modern text editors should be able to display ESLint warnings out of the box as soon as you install NPM dependencies (the yarn install command in the build steps). Refer to the documentation of you text editor if it does not.

Formatting (code style)

Rust:

cargo fmt --all

Typescript:

cd packages/nextclade-web
yarn format:fix

Maintenance

Continuous integration (CI)

Nextclade build and deployment process is automated using GitHub Actions:

The workflows run on every pull request on GitHub and every push to a major branch.

Deployment environments

Nextclade GitHub repository contains 3 major branches with special meaning: master, staging and release, each has a corresponding domain name for Nextclade Web. Nextclade built from one of these branches fetches datasets from the corresponding dataset deployment environment (See Dataset server maintenance guide)

Other branches are built in the context of GitHub pull requests. If you submit a pull request, then Vercel bot will automatically post a comment message with a URL to the preview deployment of Nextclade Web. After CLI GitHub Actions workflow finishes, you can find the resulting Nextclade CLI executables in the "Artifacts" section of the workflow.

Here is a list of environments:

Nextclade repo branch Nextclade Web domain name Dataset server Meaning
release clades.nextstrain.org data.clades.nextstrain.org Final release, targeting all end users
staging staging.nextstrain.org data.staging.nextstrain.org Staging release, for last-minute testing and fixes before a final release is made, to not block progress on master branch
master master.nextstrain.org data.master.nextstrain.org Main development branch - accumulates features and bug fixes from pull requests
other branches temporary domain on Vercel branch with the same name in dataset GitHub repo if exists, otherwise data.master.nextstrain.org Pull requests - development of new features and bug fixes

Preview versions of Nextclade Web built from pull requests will first try to fetch data from GitHub, from the branch with the same name in the dataset GitHub repository, if such branch exists. If not, the it will fetch from master environment. This is useful during development, when you need to modify both software and data: if you have branches with the same name in both repos, Nextclade Web will fetch the datasets from that branch.

Nextclade CLI built from pull requests in Nextclade repository is always using master deployment.

If you build Nextclade Web or Nextclade CLI locally, you can configure the data environment by setting DATA_FULL_DOMAIN variable in your local .env file. Note that despite the name, variable should contain fUll URL to the dataset server root. This is a build-time setting. You need to rebuild Nextclade every time you set it.

For example, for Nextclade v3 the default setting (master environment) is:

DATA_FULL_DOMAIN=https://data.master.clades.nextstrain.org/v3

You can serve datasets locally and tell Nextclade to use your local server:

DATA_FULL_DOMAIN=http://localhost:3001

You can turn on fetching from the same branch from the dataset repo by setting:

DATA_TRY_GITHUB_BRANCH=1

If you are deploying your own Nextclade instance, although it might be tempting to fetch datasets from GitHub directly, without deploying them to a file server, this is not recommended, because your users will probably hit GitHub's usage limits. i.e. we don't recommend to enable this setting for your major branches and end-user releases.

Trying custom dataset server

There are multiple ways to make Nextclade to use a custom dataset server instead of the default one. This is useful for local testing, when developing datasets or Nextclade software itself.

In all cases you need to have a dataset server directory ready (contained datasets and all the required index files).

Prepare and serve a dataset server directory locally

  • Build a fresh dataset server directory as described in the nextstrain/nextclade_data repo. At the time of writing it simply means to run ./scripts/rebuild and to observe the data_output/ directory created, containing the dataset files and associated index files

  • Serve the output directory locally using any static file server. CORS should be enabled on the server. For example, using serve package from NPM:

    npx serve@latest --cors --listen=tcp://0.0.0.0:3001 data_output/

    Now you should be able to fetch dataset index file with curl:

    curl http://localhost:3001/index.json

    and to see some JSON when navigating to http://localhost:3001/index.json in a web browser.

Temporarily use custom dataset server with Nextclade CLI

Run the usual dataset list and dataset get, with an additional flag:

--server=http://localhost:3001

This will tell Nextclade to use the local dataset server instead of the default one.

See Nextclade CLI user documentation for more details about available command ine arguments. You can type type nextclade --help for help screen. Each subcommand has it's own help screen, e.g nextclade dataset get --help.

Temporarily use custom dataset server with Nextclade Web

To provide Nextclade with the alternative location of the dataset server, add the dataset-server URL parameter with value set to URL of the custom dataset server:

https://clades.nextstrain.org?dataset-server=http://example.com

Local URLs should also work:

https://clades.nextstrain.org?dataset-server=http://localhost:3001

Combining locally built Nextclade Web and local dataset server too:

https://localhost:3000?dataset-server=http://localhost:3001

This instructs Nextclade to disregard the default dataset server URL and fetch data and index files from this custom location instead.

⚠️ Web browser should be able to reach the dataset server address provided. Additionally, make sure Cross-Origin Resource Sharing (CORS) is enabled on your server as well as that all required authentication (if any) is included into the file URL itself.

⚠️The URLs might get quite complex, so don't forget to encode special characters, to keep the URLs valid.

See Nextclade Web user documentation for more details about available URL parameters.

Permanently configure Nextclade CLI and Nextclade Web to use custom dataset server

Open .env file in the root of the project (if you don't have it, create it based on .env.example) and set the DATA_FULL_DOMAIN variable to the address of your local dataset server. In the example above it would be:

DATA_FULL_DOMAIN=http://localhost:3001

Rebuild Nextclade CLI and it will use this address by default for all dataset requests (without need for the additional --server flag).

Rebuild Nextclade Web and it will use this address by default for all dataset requests (without need for the additional dataset-server URL parameter).

Note that this address will be baked into the CLI binaries or into the Web app permanently. Switch to the default value and rebuild to use the default dataset server deployment again.

Any network location can be used, not only localhost.

The same mechanism is used during CI builds for master/staging/production environments, to ensure they use their corresponding dedicated dataset server.

Maintenance

There are 2 release targets, which are released and versioned separately:

  • Nextclade CLI
  • Nextclade Web

Versioning

Nextclade project tries hard to adhere to Semantic Versioning 2.0.0

Releasing

⚠️ We prefer to make releases on weekdays from Tuesday to Thursday, ideally around Wednesday in UTC zone, to ensure that everyone affected (dev team and users) is full of energy and that there's enough time to react to changes and to fix potential breakage without causing overtime hours. We try to avoid releases before and on major holidays and on Fridays to avoid possible weekend/holiday surprises.

Note that due to 3-tier branch system, development is never blocked by the unreleased changes.

Releasing new CLI version
  • Checkout the branch and commit you want to release. Theoretically, you can release any commit, but be nice and stick to releases from master.
  • If you are making a stable release, make sure to fill the CHANGELOG.md and commit changes to your branch. Pay particular attention to headings: CI will extract the text between the two first ## headings, in a very silly way, and will use this text as release notes on GitHub Releases.
  • Make sure there are no uncommitted changes.
  • Follow comments in the script ./scripts/releases on how to install dependencies for this script.
  • Run ./scripts/releases cli <bump_type>, where bump_type signifies by how much you want to increment the version. It should be one of: major, minor, patch, rc, beta, alpha. Note that rc, beta and alpha will make a prerelease, that is - marked as "prerelease" on GitHub Releases and not overwriting "latest" tags on DockerHub.
  • Verify the changes the script applied:
    • versions are bumped as you expect in all Cargo.toml and Cargo.lock files.
    • a local commit created on branch release-cli with a message containing the version number that you expect
  • The script will ask if you want to push the changes. This is the last step. If you agree, then the changes will be pushed to GitHub and CI will start a build. You can track it here. If you refuse this step, you can still push later.
Releasing new Web version
  • There are 3 websites exist, for master, staging and release environments. They map to master, staging and release git branches. Pick an environment you want to deploy the new version to and checkout the corresponding branch.
  • If you are deploying to release, make sure to fill the CHANGELOG.md and commit changes to your branch. Pay particular attention to headings: CI will extract the text between the two first ## headings, in a very silly way, and will use this text as release notes on GitHub Releases.
  • Make sure there are no uncommitted changes.
  • Follow comments in the script ./scripts/releases on how to install dependencies for this script.
  • Run ./scripts/releases web <bump_type>, where bump_type signifies by how much you want to increment the version. It should be one of: major, minor, patch, rc, beta, alpha. It is advised against releasing rc, beta, alpha to release environment.

If you want to deploy the same version to multiple environments, then release to one environment (on one branch) and then promote it to other environments: manually fast-forward other branch(es) to this commit and push.