strider-extension-loader

Load Strider extensions

Strider Extension Loader

Strider is an extensible CI system, written in node. Strider extensions are simply NPM packages with additional metadata contained in a file named strider.json. This metadata tells Strider which JavaScript source files should be loaded and initialized.

Hence, to install a new Strider extension, you can just npm install it in your strider repositiory.

This is a small Node.JS library for loading Strider extensions.

var Loader = require('strider-extension-loader');
var loader = new Loader();
  • lesspaths is an optional list of directories that will be made available while compiling plugins' less style files.
  • isNamespaced is for backwards compatibility with older versions where the default type controllers e.g. JobController were not namespaced. For versions < 1.6.0 this property should NOT be set.

Collect all strider extensions found in the given directories.

Load the "webapp" portion of all extensions.

extensions looks like { plugintype: { pluginid: loadedPlugin, ... }, ... }

The structure of the loadedPlugin object depend on the plugin type.

  • job:

Same as initWebAppExtensions but for the worker portion.

Load all of the templates from all extensions. templates looks like { templatename: stringtemplate, ... }.

Register the /static/ directories of all plugins to map to /ext/:pluginid.

Assets for configuring plugins on the /my/project/config page.

Collect the html, js, and css for all plugins. This is per-project config. js scripts will each be wrapped in an anonymous function to namespace it, and concatenated into one file (in future it will also be minified). Stylesheets will also be concatenated together, and .less files will be compiled to css (with the lesspaths available for imports).

Then the js and css are written to the files specified jspath and csspath.

Html for the templates are available on the configs objects.

Configs look like { plugintype: { pluginid: config, ...}, ... } and config looks like:

{
  id: "myplugin",
  controller: "MyController", // defaults to [plugintype]Controller 
  html: "<the>html</the>",    // loaded from config.template 
  icon: "icon.png",           // relative to the plugin's `static` directory 
  title: "My Plugin"
}

Basically, this is constructed from the strider section of package.json.

"strider": {
  "id": "myplugin",
  "title": "My Plugin",
  "icon": "icon.png", // should be in the ./static dir 
  "config": {
    "controller": // defaults to "Config.JobController" for job plugins, "Config.ProviderController", etc. 
    "script":     // path where the js should be loaded from. Path defaults to "config/config.js" 
    "style":      // defaults to "config/config.less" 
    "template":   // defaults to "config/config.html" 
  }
}

I hope that's clear.

If you don't need to do anything fancy, you can just use the default controller for your plugin type. Take a look in strider's client/config/controllers directory for the source of those controllers. Basically, each controller makes available a config object on the scope, which is populated by the plugin's config for the currently selected branch. Also a save() function is available on the scope.

So, for example, the simplest configuration template for any plugin type could just have

<input ng-model="config.oneAttr" placeholder="One Attribute Here">
<button class="btn" ng-click="save()">Save</button>

No javascript required. Just put that in "config/config.html" and you're done.

This is very similar to initConfig, but for per-user as opposed to per-project config. For provider plugins, the default file name is config/accountConfig.html, js, less, and for all other plugin types it's config/userConfig.html, js, less.

Status blocks allow plugins to

  • runner: runs the jobs, like strider-simple-runner
  • provider: gets the source code for a project, like strider-github, strider-bitbucket or strider-git. Can be hosted or regular.
  • job: setup the environment, run tests, deploy etc. like strider-node or strider-sauce
  • basic: do whatever you want. If you need more power, use this, but you don't get the helpers provided by the more specific plugin types.

There are two environments where plugins are loaded, webapp and worker.

Effects the way the strider webapp works, how it looks, etc. You can define templates, serve static files, listen to global strider events, and other great things.

This code is loaded for each job that is run, by the process that is running the job. This may be the same process as the webapp (as when using strider-simple-runner), or it might be somewhere else entirely. Accordingly, it is recommended that you not depend on network connections unless absolutely necessary. In many cases, you can pass a message up to the strider app and handle it in your webapp code.

To declare your npm package as a strider plugin, include a strider.json in the base directory. Alternatively, you can have a strider section to your package.json.

strider.json schema:

{
  "id": "pluginid", // must be unique. 
  "title": "Human Readable",
  "icon": "icon.png", // relative to the plugin's `static` directory 
  "type": "runner | provider | job | basic", // defaults to basic 
  "webapp": "filename.js", // loaded in the webapp environment 
  "worker": "filename.js", // loaded in the worker environment 
  "templates": {
    "tplname": "<div>Hello {{ name }}</div>",
    "tplname": "path/to/tpl.html"
  },
  "config": { // project-specific configuration 
    "controller": // defaults to "Config.JobController" for job plugins, "Config.ProviderController", etc. 
    "script":     // path where the js should be loaded from. Path defaults to "config/config.js" 
    "style":      // defaults to "config/config.less". Can be less or css 
    "template":   // defaults to "config/config.html" 
  }
  // other configurations 
}

Additionally, if there is a /static/ directory, the files within it will be accessible at the path /ext/:pluginid.

Runner plugins do not get loaded in the worker environment.

module.exports = {
  config: {}, // mongoose schema. This will be per-project config 
  appConfig: {}, // mongoose schema. Global config 
  createfunction (emitteroptionscallback(errrunner)) {  }
}

These are used for the strider admin dashboard.

  • capacity
  • running number of jobs currently running
  • queued length of the queue
  • job.new (job) see strider-runner-core for a description of the job data
  • job.cancel (jobid) if the runner has the specified job, either queued or in process, stop it and fire the job.canceled event

Runners are only expected to handle a job if job.project.runner.id identifies it as belonging to this runner.

  • browser.update (eventname, data) this is for proxying internal job.status events up to the browser
  • job.queued (jobid, time)
  • job.done (data)
  • panel see the job plugin section
  • appPanel similar to panel, but for global config

Provider plugins that need an ssh keypair are encouraged to use the privkey and pubkey that are defined for each project. They are attributes on the project object.

module.exports = {
  // mongoose schema for project-specific config 
  config: {},
 
  // mongoose schema for user-level config (like a github OAuth token) and/or cache 
  userConfig: {},
  // optional; used by services such as github, bitbucket, etc. 
 
  // getFile: used to get the .strider.json file 
  getFilefunction (filepathuserConfigconfigprojectdone(errfilecontents))
 
  // getBranches: get the branches for a repository 
  getBranchesfunction (userConfigconfigprojectdone(err, [branchname, ...]))
 
  // listRepos: only used by providers that connect to a hosted service. 
  // repos: { groupname: [repo, ...], groupname: ... } 
  listReposfunction (userConfigdone(errrepos)) {},
 
  // if this provider plugin needs setup (in github's case, oauth) this string 
  // represents the href link to the page to handle that. 
  setupLink: "/ext/github/oauth",
 
  // determine whether or not this provider is setup for this user. 
  // e.g. for github, that we have an oauth key 
  // returns boolean 
  isSetupfunction (account) {},
 
  initializefunction (userConfigrepodone(errnamedisplay_urlconfig)) {},
  // namespaced to /ext/:pluginid 
  routesfunction (appcontext) {
  }
}

The repos that are returned by listRepos contain objects which, when activated, will be the provider config for the project. As such, it is required to have a url that is unique, a name that looks like "org/name" and it should also define a display_url where appropriate. All other config is up to you.

{
   name: 'some/name', // should have exactly one '/' 
   // this has to be unique; it's how we identify whether a project 
   // has already been configured. 
   url: 'http://example.com/unique/url.git',
   // optional - linked to from the project page 
   display_url: 'http://example.com/path/to/repo',
   // everything else is up to you. 
}

If just a function is exposed, it is assumed to be "fetch".

module.exports = {
  // get the source code for a project. This is where the real work gets done. 
  //   dest: the path to put things 
  //   context: contains runCmd, io (for event passing) 
  fetchfunction (destuserConfigconfigjobcontextdone) {
  },
}

Use panel for project-level config, and userPanel for user-level config.

  • inline_icon you can also define a 24x24 icon for the display_url links. If this is not a path, it is assumed to be the name of an icon from FontAwesome (without the icon- prefix) and will be loaded as such. Defaults to external-link.
{
   config: {}, // mongoose schema, if you need project-specific config 
   // Define project-specific routes 
   //   all routes created here are namespaced within /:org/:repo/api/:pluginid 
   //   req.project is the current project 
   //   req.accessLevel is the current user's access level for the project 
   //      0 - anonymous, 1 - authed, 2 - admin / collaborator 
   //   req.user is the current user 
   //   req.pluginConfig() -> get the config for this plugin 
   //   req.pluginConfig(config, cb(err)) -> set the config for this plugin 
   routesfunction (appcontext) {
   },
   // Define global routes 
   //   all routes namespaced within /ext/:pluginid 
   //   req.user is the current user 
   //   req.user.account_level can be used for authorization 
   //      0 - anonymous, 1 - authed, 2 - admin / collaborator 
   globalRoutesfunction (appcontext) {
   },
   // Listen for global events 
   //   all job-local events that begin with `plugin.` are proxied to 
   //   the main strider eventemitter, so you can listen for them here. 
   //   Other events include `job.new`, `job.done` and `browser.update`. 
   listenfunction (emittercontext) {
   }
}

If only a function is exposed, it is assumed to be the init(config, job, cb) function.

Autodetection rules are only used when a project has no plugins configured.

module.exports = {
  // Initialize the plugin for a job 
  //   config: the config for this job, made by extending the DB config 
  //           with any flat-file config 
  //   job:    see strider-runner-core for a description of that object 
  //   context: currently only defines "dataDir" 
  //   cb(err, initializedPlugin) 
  initfunction (configjobcontextcb) {
    return cb(null, {
      // string or list - to be added to the PATH 
      path: path.join(__dirname, 'bin'),
      // any extra env variables. Will be available during all phases 
      env: {},
      // Listen for events on the internal job emitter. 
      //   Look at strider-runner-core for an 
      //   enumeration of the events. Emit plugin.[pluginid].myevent to 
      //   communicate things up to the browser or to the webapp. 
      listenfunction (emittercontext) {
      },
      // For each phase that you want to deal with, provide either a 
      // shell command [string] for a fn(context, done(err, didrun)) 
      environment: 'nvm install ' + (config.version || '0.10'),
      prepare: 'npm install',
      testfunction (contextdone) {
        checkSomething(context, function (shouldDoThings) {
          if (!shouldDoThings) {
            // Send `false` to indicate that we didn't actually run 
            // anything. This is so we can warn users when no plugins 
            // actually do anything during a test run, and avoid false 
            // positives. 
            return done(null, false);
          }
          doThings(function (err) {
            done(err, true);
          });
        });
      },
      cleanup: 'rm -rf node_modules'
    });
  }
  // this is only used if there is _no_ plugin configuration for a 
  // project. See gumshoe for documentation on detection rules. 
  autodetect: {
    filename: 'package.json',
    exists: true
  }
}
  • job
  • project
  • dataDir
  • phase
{
  cmd: "shell string" || {command: "", args: [], screen: ""},
  env: {}, // any extra env variables
  cwd: "" // defaults to the root directory of the project
}

If the command contains sensitive information (such as a password or OAuth token), you can specify a screen command, which is what will be output.

See strider-runner-core for the job.status. events. This emits a job.status.[type] event with [jobid] + arguments.

Type defaults to stdout. It can be one of stderr, message, error, warn. error and warn are prefixed by a colored [STRIDER] WARN | ERROR and sent to stderr. message is prefixed by a colored [STRIDER] and sent to stdout.

You shouldn't need to use these, but they're there.

  • logger
  • io

Job plugins can also define an icon in the strider.json object, which is the path to a 48x48 image that will be shown on the project configuration page when a user is enabling plugins.

If the plugin requires special configuration, it can also define a panel object in strider.json, which looks like:

"panel": {
  "src": "path/to/file.html",
  "controller": "NameOfCtrl" // defaults to [pluginid]Controller 
}

Define the angular controller in /static/project_config.js, which will be loaded.

See strider-webhooks for an example of a custom config panel.

This is where you do whatever you want. It will not be listed in the UI anywhere automatically, so user configuration will require your own ingenuity. If the need arises, we might expose some kind of config on the system level to strider administrators, but not at the moment.

You can listen for events, but you shouldn't run any tests or interact with the source code in any way. For that, write a job plugin.

module.exports = function (contextjobdone) {
}
module.exports = function (contextdone) {
}

This is what gets passed into the basic init function, as well as the listen and routes functions of various plugin types.

  • config -- main strider config
  • emitter -- for passing events
  • models
  • logger
  • middleware
  • auth
  • app -- the express app
  • registerBlock
ctx.registerBlock('HeaderBrand', function(contextcb) {
  // context has a lot of useful stuff on it: 
 
  var email = context.currentUser.user.email;
 
  // You can do some async processing here, but bear in mind 
  // you'll be blocking the page load. 
 
  cb(null, '<h1>FooStrider</h1>');
});

Because writing a bunch of registerBlock calls for simple pieces of template overrides is a little tedious, you can also use the following shortcut in your strider.json:

{
  "templates": {
    "HeaderBrand": "<h1>An HTML String</h1>",
    "FooterTOS": "./path/to/TOS.html"
  }
}

These are either inline strings or paths to static HTML. There is no templating available for these at present.

Note If more than one override is specified for a block, then the first one will be used. At the moment this means that extensions can squash each other. If you want to simply 'append' to a block, use the registerBlock method and make sure that you prefix the html you return with: ctx.content which will contain either the default html, or the content from previous extensions.