Skip to content

Latest commit

 

History

History
447 lines (311 loc) · 29.1 KB

README.md

File metadata and controls

447 lines (311 loc) · 29.1 KB

Unofficial Meteor FAQ

Just answering some common questions that aren’t answered on the meteor site. Thanks to all the contributors: https://github.com/oortcloud/unofficial-meteor-faq/graphs/contributors !

Table of Contents

  • How do I update this FAQ?
  • How can I find out more about Meteor?
  • Customizing Meteor
  • Reactivity and Rendering
  • Animation
  • Subscriptions and Methods
  • Deployment
  • Best practices
  • Troubleshooting errors
    • "Uncaught SyntaxError: Unexpected token Y"
    • "TypeError: Object # has no method '_defineMutationMethods'"
    • "Uncaught TypeError: Converting circular structure to JSON"
    • "Unexpected mongo exit code 100. Restarting."
    • @importing in less or sass files causes errors related to variables not found.
    • Connecting with default MongoDB database
    • ###How do I update this FAQ?

      Send me (Tom Coleman, @tmeasday) a message on github (or otherwise) or fork the FAQ and send a pull request (we can discuss changes and ideas as part of the pull request).

      ###How can I find out more about Meteor?

      Customizing Meteor

      ###How do I use a preview release of meteor? Find the git tag you'd like to use in the meteor repository's release list, then use the --release flag with the meteor command. For example:

      $ meteor --release template-engine-preview-10.1

      Meteor will handle updating any smart packages and npm dependencies automatically.

      ###How do I use a unreleased branch of meteor? You can git clone it anywhere on your system and then run it with:

      $ path/to/checked/out/meteor/meteor

      Or you could let meteorite do this for you.

      ###How can I use someone else's javascript in my meteor project?

      It depends on how it's packaged. Search for it on Atmosphere, and if there's a existing package, it should be as simple as meteor add X to get it.

      If there's no Atmosphere package yet, consider repackaging the external library for meteor.

      If it's a simple client side JS script, you can include it in client/lib/ or lib/, although it might be nicer to create a smart package to wrap it, and publish that package on atmosphere. There are some good instructions on the atmosphere page about how to do that.

      ###What about npm (node) modules?

      Meteor by default comes bundled with a set of node modules, which you can access via:

      var foo = Npm.require('foo');

      If you need a npm package that is not included in Meteor, you need to create a Meteor package that wraps it. Avital Oliver created a simple example for the xml2js npm library.

      Note that you cannot Npm.require such npm packages from within your app, you can only require them within the Meteor package that depends on them. This is due to different Meteor packages potentially depending on different versions of the same npm package.

      There is another way you can use NPM modules with your app.

      • first you need to install npm smart package with - mrt add npm
      • then add a file named packages.json and add npm modules

      Read this guide to learn more on using NPM modules with a meteor app.

      Reactivity and Rendering

      ###How do I stop meteor from reactively overwriting DOM changes from outside meteor?

      Since the rendering-system blaze has come out, this shouldn't happen anymore. You may want to read the templates section of the docs. You may also be interested in the using-blaze article.

      How do I ensure control state preservation across live page updating?

      Add the preserve-inputs package.

      From the docs: "This preserves all elements of type input, textarea, button, select, and option that have unique id attributes or that have name attributes that are unique within an enclosing element with an id attribute."

      Also, make sure to store related client-side data in the Session object (not in JavaScript variables)

      NOTE: form data will still get cleared across hot-code pushes unfortunately.

      ###How do I get a callback that runs after my template is attached to the DOM?

      This is now straightforward with the rendered callback.

      ###How do I get reactive HTML in the <head> tag?

      Unfortunately, this isn't yet supported. However, you can work around it, like so:

      Meteor.startup(function() {
        Deps.autorun(function() {
          document.title = Session.get('document-title');
        });
      });

      Note also that this code should work fine with the spiderable package, as long as the document-title session var is set when your site initially loads (for instance in a router).

      How can I tell why my code is getting re-run?

      If you are using an autorun block, you could try this:

      Deps.autorun(function(computation) {
        computation.onInvalidate(function() {
          console.trace();
        });
        
        // standard autorun code...
      });

      Animation

      How do I animate when meteor changes things under me?

      This is a similar problem to above. I outlined some techniques that have worked for me in a blog post.

      ###How do I animate things adding/being removed from collections?

      One way to animate things being added is to transition from a .loading class when it's rendered. First, add a .loading class to to the element in the HTML.

      Then, in the relevant template's rendered function, remove the class just after it attaches to the document:

      Template.item.rendered = function() {
        var $item = $(this.find('.item'));
        Meteor.defer(function() {
          $item.removeClass('loading');
        });
      }

      The defer (which corresponds to a setTimeout of 0) is necessary as otherwise the browser won't have had a chance to render the item in the .loading state (i.e. hidden / offscreen / whatever).

      Removal animations are more difficult, because Meteor will remove the element from the DOM immediately upon the item leaving the collection. Some hacks that you could try (please show me if you get anything working!):

      • Not deleting the element immediately but waiting for the animation to complete (ughhh, not great interactivity between users).
      • Hooking up your own observe calls on the collection so you can handle removed differently - ie. making an 'animating' cursor. (Much better, perhaps over-engineering the problem)

      You may also want to render items added by the client in a different state, until they're confirmed server side.

      ###How do I route my app between different views/pages?

      Ideally, you'd use an existing JS router and combine it with a reactive variable (e.g. in the session) which defines the visible page. Or you could just try Iron Router, which supersedes my reactive router package, which did this for you with the backbone router. Worth a look even if you want to use a different router as it could give you some ideas.

      ###How do I animate/transition such view changes?

      You can view these articles (part 1 and part 2) on animating using _uihooks.

      ###How do I animate "drag-n-drop", "add" and "remove" actions on lists?

      Example implementation here: meteor-todos-sortable-animation.

      Subscriptions and Methods

      How can I alter every document before it is added to the database?

      There is a feature request to add this to core - please vote for it. Right now you can hack around by using a deny to achieve what you want on the server. For example, to timestamp each document as it goes into MongoDB:

      Posts.deny({insert: function(userId, doc) {
        doc.createdAt = new Date().valueOf();
        return false;
      }})

      You can do something similar with update to for example simulate a synthetic updatedAt field.

      Avoid the tempation to use allow for this, as there's no guarantee it will run (only one allow needs to be successful for the insert to go ahead).

      Until the feature gets implemented in core, Mathieu Bouchard (@matb33) has published collection-hooks.

      How do I know when my subscription is "ready" and not still loading?

      When you subscribe to a collection you are returned a handle. This handle has a reactive ready() method that lets you know when the initial snapshot of the collection has been sent to the client.

      globalSubscriptionHandles.push(Meteor.subscribe('foo'));
      
      ...
      
      Template.item.areCollectionsReady = function()
      {	
      	var isReady = globalSubscriptionHandles.every(function(handle)
      	{
      	    return handle.ready();
      	});
      
      	return isReady;
      }

      Why does observe fire a bunch of added events for existing documents?

      The default behaviour is to do this, because it's often what you want (think about how the {{#each}} tag works for instance). You can avoid it with a hidden options though:

      Posts.find().observe({
        _suppress_initial: true,
        added: function() {
          // ...
        }
      })

      Note that this option only works on the client. Also note that considering this option is undocumented (and underscore-cased), it could silently change in the future.

      Why do I sometimes see two copies of a document during a latency compensated method?

      If you insert a document in both the real (server-side) method and the simulation (client-side), but then subsequently do more work on the server, there will be a time period where, for the a single browser, there will be two copies of the document in the database.

      Meteor dev Nick Martin wrote:

      There are a couple ways to work around this:

      a) pick the same id on the client and the server. The easiest way to do this (and what the default 'insert' method does) is simply pick an id on the client and pass it along with the document to be inserted.

      Doing this will result in their only being one document in the database, so you won't see double. It will still be the case that the server changes to the document don't show up locally until the method is done, though.

      b) make the method finish faster. The latency compensation is triggered when the method finishes. If the thing you are blocking on doesn't do any writes (or its writes don't need to be latency compensated), you can have the method return early and complete its work asynchronously.

      c) do the writes at the end of the method on the server. batch up all the database modifications and wait till after all the blocking stuff is done. this is gross, but should work to reduce the time where the user may see two items.

      We also have an item on the roadmap to make this easy: https://trello.com/card/pattern-for-creating-multiple-database-records-from-a-method/508721606e02bb9d570016ae/57

      How can I add additional RESTful API for my collections

      For adding RESTful methods on top of your data, look into the Collection API written for Meteor:

      https://github.com/crazytoad/meteor-collectionapi

      Deployment

      ###Where can I host my applications for production? Meteor gives you a proper infrastructure to deploy your application:

      meteor deploy myapp.meteor.com

      Meteorpedia lists more deployment options.

      ###How do I hook into an existing, running MongoDB instance? You can start meteor with the MONGO_URL environment var set:

      $ MONGO_URL=mongodb://localhost:27017 meteor

      Note that pre-existing data in mongo may be hard to deal with.

      ###How do I set environment variables on meteor.com? This is not currently possible, unfortunately. You can set process.env.X = Y early on in the startup of your app though.

      Best practices

      ###Should I check in this .meteor/ folder into my git repo? Yes. (Unless you've put unwanted files in there yourself.)

      The .meteor/ folder includes the files packages (a list of the project's current packages) and release (describing what Meteor version your project is using); these are important for other developers to clone and use your project.

      Meteor also generates a .meteor/.gitignore file, which tells Git to ignore the folder .meteor/local (the built development app and database). Users of other source control systems should manually prevent this folder from entering source control.

      ###What are the best practices for form handling? While originally intended to be part of core, forms is not currently on the roadmap.

      The most popular forms package currently in use is autoform. Other maintained packages include simple-form and Mesosphere.

      ###What is the best way to do file uploads? There's brief discussion on telescope. Also, there's a filepicker smart package.

      ###What are best practices for security? Core dev Emily Stark has given some fantastic talks about Meteor security.

      This is not a definitive listing, but some common points that should be followed:

      • A default Meteor project comes with the insecure and autopublish packages enabled to assist with rapid development. These should be disabled before development is completed.
        • With the insecure package removed, Collection insertions, updates and removals can be executed from the client if they pass one Collection.allow method and fail to trigger a Collection.deny method.
      • Code that resides outside the server, private and tests subdirectories is delivered to all clients. Ensure no sensitive code is stored outside these folders.
        • Any API keys or other sensitive information should be stored outside your codebase and delivered via environment variables.
        • Even users who are not logged in have access to the templates of the whole site, possibly including admin templates, etc.
        • Your client code can be unminified and the Meteor.methods discovered. Additionally, your Meteor application can be fuzzed. Don't rely on security through obscurity. Use the Meteor Accounts system to validate and authenticate every point of client access.
      • Server-side code has full read/write access to the database. Additionally, Meteor.methods can be called from the terminal by any client at any time. This means you should:
        • Ensure authentication checks exist for each Meteor.method.
        • Ensure any client-provided data is of the proper type e.g. run through the check package.
        • A companion tool to the check package is the audit-arguments-check package which ensures every argument passed has been run through check.
      • Meteor lets you exactly specify which data is published and to which client.
        • Your application should take care which information is published to which users.
        • Any client-provided information passed to a publish function should be run through the check package.
        • Inside a Meteor.publish function, use the server-provided this.userId, not a client-provided property to determine access.
        • Take extra care with any publish function that deals with the Meteor.users Collection.
      • Using the browser-policy package helps prevent malicious JavaScript from being injected client-side, your site being iframed by a malicious site, etc.
      • You should be using the most up-to-date version of Meteor possible, but minimally Meteor version 0.6.6.2 (0.6.6.2 was a Node.js security patch).

      ###How do I debug my meteor app? Client-side you have the browser console.

      If you're using an IDE like WebStorm, see this SO answer.

      For server-side debugging, use node-inspector and make sure you have meteor v0.5.3 or later, which makes things easier thanks to support for NODE_OPTIONS:

      1. Install node-inspector: npm install -g node-inspector
      2. Start meteor: NODE_OPTIONS='--debug-brk' mrt run
      3. Start node-inspector
      4. In Chrome, go to the URL given by node-inspector
      5. If debugging remotely, make sure the port given at the URL above (8080) is not firewalled
      6. Debug at will

      How do I run a local version of Meteor with the production database?

      Override the MONGO_URL environment variable with the output of meteor mongo, which looks like.

      MONGO_URL=$(meteor mongo --url http://<your-app>.meteor.com/) meteor

      For testing or security or backup's sake, you may also just want to download the production database into your local app. There's a script that makes downloading your production database as easy as

      ./download.sh <yourapp>.meteor.com

      ###What are the best practices for Test-Driven Development? Velocity is Meteor's official testing solution as of 0.8.1.

      TDD support was't official in meteor prior to 0.8.1, but (test) files placed in the tests subdirectory won't be loaded on the client or server. There are various Node.JS modules that help with testing in versions prior to 0.8.1 - see Meteor test driven development on SO.

      The most popular TDD frameworks for meteor 0.8.0 and lower were Laika, RTD, and Safety Harness, which was designed to address some of the shortcomings of the first two - see the feature comparison and the meteor-talk thread where the three authors duke it out. But if you're using 0.8.1 or later, Velocity is where it's at.

      ###Where should I put my files?

      The example apps in meteor are very simple, and don’t provide much insight. Here’s my current thinking on the best way to do it: (any suggestions/improvements are very welcome!)

      lib/                       # <- any common code for client/server.
      lib/environment.js         # <- general configuration
      lib/methods.js             # <- Meteor.method definitions
      lib/external               # <- common code from someone else
      ## Note that js files in lib folders are loaded before other js files.
      
      collections/               # <- definitions of collections and methods on them (could be models/)
      
      client/lib                 # <- client specific libraries (also loaded first)
      client/lib/environment.js  # <- configuration of any client side packages
      client/lib/helpers         # <- any helpers (handlebars or otherwise) that are used often in view files
      
      client/application.js      # <- subscriptions, basic Meteor.startup code.
      client/index.html          # <- toplevel html
      client/index.js            # <- and its JS
      client/views/<page>.html   # <- the templates specific to a single page
      client/views/<page>.js     # <- and the JS to hook it up
      client/views/<type>/       # <- if you find you have a lot of views of the same object type
      client/stylesheets/        # <- css / styl / less files
      
      server/publications.js     # <- Meteor.publish definitions
      server/lib/environment.js  # <- configuration of server side packages
      
      public/                    # <- static files, such as images, that are served directly.
      
      tests/                     # <- unit test files (won't be loaded on client or server)

      For larger applications, discrete functionality can be broken up into sub-directories which are themselves organized using the same pattern. The idea here is that eventually module of functionality could be factored out into a separate smart package, and ideally, shared around.

      feature-foo/               # <- all functionality related to feature 'foo'
      feature-foo/lib/           # <- common code
      feature-foo/models/        # <- model definitions
      feature-foo/client/        # <- files only sent to the client
      feature-foo/server/        # <- files only available on the server

      What IDEs are best suited for meteor?

      While many IDEs support SASS, CoffeeScript etc., at the moment, there are no known IDEs that support meteor. Two efforts are underway:

      • Cloud9, the cloud-based IDE with Node.js support
      • JetBrains WebStorm (Win/Mac/Linux), with existing support for Node.js. If you use Webstorm, make sure to disable auto-save. Webstorm is proprietary software that offers a gratis license to people working on open source projects

      What versions of Internet Explorer does meteor support?

      There isn't an official list of supported versions, but meteor is known to work on IE8+.

      Troubleshooting errors

      "Uncaught SyntaxError: Unexpected token Y"

      Client-side error caused by the server crashing and sending a message starting with "Your app is crashing. Here's the latest log."

      "TypeError: Object #<Object> has no method '_defineMutationMethods'"

      Server-side error. Most likely, you forgot to place "new" before a constructor call in the client. Read more.

      "Uncaught TypeError: Converting circular structure to JSON"

      Check if you're trying to save into Session an object with circular references, such as a Collection. Read more.

      "Unexpected mongo exit code 100. Restarting."

      Mongo was killed without cleaning itself up. Try removing .meteor/local/db/mongod.lock. If that fails do an meteor reset.

      @importing in less or sass files causes errors related to variables not found.

      If you're using a collection of less files that need to be imported in a specific order because of variable dependencies (like a custom Twitter Bootstrap installation):

      • for LESS: you can change the extension of the less files to be imported from .less to .import.less and then change your @import file.less to @import file.import.less.
      • for Sass: rename the files you're importing by adding a _ before them. file.scss becomes _file.scss. And then change @import 'file'; to @import '_file';. This will prevent the less/sass compiler from automatically trying to compile all your import files independently, yet still let you use them in the order you want.

      Connecting with default MongoDB database

      In order to look up content of default MongoDB database, you can retreive MONGO_URI by running following command: meteor mongo -U