midgard/mapcomplete
1
0
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity
mapcomplete/Docs/Development_deployment.md

7.2 KiB
Raw Blame History

Development and deployment

There are various scripts to help set up MapComplete for developing and for deployment.

This documents attempts to shed some light on these scripts.

Note: these scripts change every now and then - if the documentation here is incorrect or you run into troubles, do leave a message in the issue tracker

Architecture overview

At its core, MapComplete is a static (!) website. There are no servers to host.

The data is fetched from Overpass/OSM/Wikidata/Wikipedia/Mapillary/... and written there directly. This means that any static file server will do to create a self-hosted version of MapComplete.

Dependencies

make , python3, g++

(Nix users may run nix-env -iA nixos.gnumake nixos.gdc nixos.python3)

Development using *NIX

You need at least 3GB RAM available to run MapComplete, but you'll preferably have 8GB of free RAM available.

To develop and build MapComplete, you

  1. Make a fork and clone the repository. (We recommend a shallow clone with git clone --filter=blob:none <repo>)
  2. Install python3 if you do not have it already - On Linux: sudo apt install python3
  3. Install nvm to easily install node:
    • wget -qO- https://raw.githubusercontent.com/nvm-sh/nvm/v0.37.2/install.sh | bash
    • Restart your terminal
    • Run nvm install and nvm use to install and use the correct version of node. (_Note: nvm might complain that the relevant version is not yet installed. It'll have it installed only for the current user account but not system-wide - which is fine)
  4. Run npm run init (including run, not npm init)which …
    • runs npm ci for you
    • generates some additional dependencies and files
    • does various housekeeping and setup. This can take a few minutes the first time as some PNGs need to be created
  5. Run npm run start to host a local testversion at http://localhost:1234/
  6. By default, a landing page with available themes is served. In order to load a single theme, use layout=themename or userlayout=true#<layout configuration> as Query parameter. Note that the shorter URLs (e.g. bookcases.html, aed.html, ...) don't exist on the development version.

The previous instructions were tested on 2023-03-09 on a Ubuntu 22.04 machine.

Development using Windows

You need at least 3GB RAM available to run MapComplete, but you'll preferably have 8GB of free RAM available.

For Windows you can use the devcontainer, or the WSL subsystem.

To use the devcontainer in Visual Studio Code:

  1. Make sure you have installed the Remote - Containers extension and it's dependencies.
  2. Make a fork and clone the repository.
  3. After cloning, Visual Studio Code will ask you if you want to use the devcontainer.
  4. Then you can either clone it again in a volume (for better performance), or open the current folder in a container.
  5. By now, you should be able to run npm run start to host a local testversion at http://localhost:1234/index.html
  6. By default, a landing page with available themes is served. In order to load a single theme, use layout=themename or userlayout=true#<layout configuration> as Query parameter. Note that the shorter URLs ( e.g. bookcases.html, aed.html, ...) don't exist on the development version.

To use the WSL in Visual Studio Code:

  1. Make sure you have installed the Remote - WSL extension and it's dependencies.
  2. Open a remote WSL window using the button in the bottom left.
  3. Make a fork and clone the repository.
  4. Install npm using sudo apt install npm.
  5. Run npm run init and generate some additional dependencies and generated files. Note that it'll install the dependencies too
  6. Run npm run start to host a local testversion at http://localhost:1234/index.html
  7. By default, a landing page with available themes is served. In order to load a single theme, use layout=themename or userlayout=true#<layout configuration> as Query parameter. Note that the shorter URLs ( e.g. bookcases.html, aed.html, ...) don't exist on the development version.

Automatic deployment

Currently, the master branch is automatically deployed to https://mapcomplete.osm.be/ by a GitHub action.

Every branch is automatically built (upon push) to https://pietervdvn.github.io/mc/<branchname> by a GitHub action.

Deploying a fork

A script creates a webpage for every theme automatically, with some customizations in order to:

  • to have shorter URLs
  • have individual social images
  • have individual web manifests

This script can be invoked with npm run prepare-deploy

If you want to deploy your fork:

  1. npm run prepare-deploy
  2. npm run build
  3. Copy the entire dist folder to where you host your website. Visiting index.html gives you the landing page, visiting yourwebsite/<theme> should bring you to the appropriate theme.

Weird errors

Try removing node_modules, package-lock.json and .cache.

Misc setup

The json-git-merger is used to quickly merge translation files, documentation here. This merge driver is broken and would sometimes drop new questions or duplicate them... Not a good idea!

Overview of package.json-scripts

  • increase-memory: give Node.js some more RAM since this is a big (and memory intensive) project to build and run
  • start: start a development server
  • test: run the unit tests
  • init: generate and download various assets which are needed to compile
  • generate:editor-layer-index: download the editor-layer-index-json from osmlab.github.io
  • generate:images: compile the SVGs into an asset
  • generate:translations: compile the translation file into a JavaScript file
  • generate:layouts: use index.html as template to create all the theme index pages. You'll want to run clean when done
  • generate:docs: generate various documents, such as information about available metatags, information to put on the OSM wiki, ...
  • generate:report: download statistics from OsmCha, compile neat graphs
  • generate:cache:speelplekken: create an offline copy of all the data required for one specific (paid for) theme
  • generate:layeroverview: read all the theme and layer configurations, compiles them into a single JSON.
  • reset:layeroverview: if something is wrong with the layer overview, create an empty one
  • generate:licenses: compile all the license info of images into a single JSON
  • optimize:images: attempt to make smaller PNGs - optional to run before a deployment
  • generate: run all the necessary generate-scripts
  • build: actually bundle all the files into a single dist/ folder
  • prepare-deploy: create the layouts
  • deploy:staging,deploy:pietervdvn, deploy:production: deploy the latest code on various locations
  • lint: get depressed by the amount of warnings
  • clean: remove some generated files which are annoying in the repo