code review

[pixelzoom]

This sim will be ready for code review around the middle of next week (Tuesday/Wednesday). To stay on schedule, code review (including addressing code review issues) needs to be completed before the sim enters dev testing <= October 18.

Assigning to @ariel-phet to identify a reviewer. Please assign this back to me, then I'll let the reviewer know when it's ready to review.

[pixelzoom]

@ariel-phet if you'd like to have one of the newer developers do the review, it might be good to have them tag along when I do my "self review", where I go through the checklist.

[ariel-phet]

@pixelzoom I probably need to bring this up at dev meeting, but there should be an initial dev test before code review (a confusion that seems to have cropped up several times). That way if there are bugs, the reviewer is not reviewing buggy code. I imagine a dev test of your code will not reveal much, but we should still take that step.

[pixelzoom]

Ah right, dev test before code review.

[pixelzoom]

Via Slack, @ariel-phet asked me to have @chrisklus participate in my "self code review" (where I go through the checklist myself before passing it on to a reviewer), and "we will count that as the code review of that sim." And ideally we should do dev test before code review. Scheduling for this review is still TBD.

[pixelzoom]

Code review is pushed to (at the earliest) the first week of November.

[pixelzoom]

@chrisklus I don't anticipate any big changes to the code base, so you can start looking around whenever you'd like. Start with model.md and implementation-notes.md. We should be able to pair on the checklist sometime next week.

[chrisklus]

@pixelzoom sounds great, thanks.

[pixelzoom]

Please mark failed items with ❌

PhET code-review checklist

Build and Run Checks

• Does the sim build without warnings or errors?
• Does the html file size seem reasonable, compared to other similar sims?
• Does the sim start up? (requirejs and built versions)
• Does the sim experience any assertion failures? (run with query parameter ea)
• Does the sim pass a scenery fuzz test? (run with query parameters fuzz&ea)

Memory Leaks

• Does a heap comparison using Chrome Developer Tools indicate a memory leak? (Describing this process is beyond the scope of this document.) Test on a version built using grunt --mangle=false. There should be a GitHub issue showing the results of testing done by the primary developer.
• For each common-code component (sun, scenery-phet, vegas, …) that opaquely registers observers or listeners, is there a call to that component’s dispose function, or documentation about why dispose is unnecessary?
• Are there leaks due to registering observers or listeners? These guidelines should be followed, or documentation added about why following them is not necessary:
  • AXON: Property.link is accompanied by Property.unlink.
  • AXON: Creation of DerivedProperty is accompanied by dispose.
  • AXON: Creation of Multilink is accompanied by dispose.
  • AXON: Events.on is accompanied by Events.off.
  • AXON: Emitter.addListener is accompanied by Emitter.removeListener.
  • SCENERY: Node.on is accompanied by Node.off
  • TANDEM: tandem.addInstance is accompanied by tandem.removeInstance, or use PhetioObject constructor+dispose
• Do all types that require a dispose function have one? This should expose a public dispose function that calls this.disposeMyType(), where disposeMyType is a private function declared in the constructor. MyType should exactly match the filename.
• ❌see test for memory leaks (brand=phet-io) #50. PhET-iO instantiates different objects and wires up listeners that are not present in the PhET-branded simulation. It needs to be tested separately for memory leaks. To help isolate the nature of the memory leak, this test should be run separately from the PhET brand memory leak test. Test with the "console" and "studio" wrappers (easily accessed from phetmarks)

Performance, Usability

• Does the sim perform as desired across the range of supported platforms? (eg, not too slow on slow platforms, not too fast on fast platforms)
• If the sim uses WebGL, does it have a fallback? Does the fallback perform reasonably well? (run with query parameter webgl=false)
• Are UI components sufficiently responsive? (especially continuous UI components, such as sliders)
• Are pointer areas optimized, especially for touch? (run with query parameter showPointerAreas)
• Do pointer areas overlap? (run with query parameter showPointerAreas) Some overlap may be OK depending on the z-ordering (if the frontmost object is supposed to occlude touch/mouse areas)

Internationalization

• Are there any strings that are not being internationalized? (run with query parameter stringTest=x, you should see nothing but 'x' strings)
• Does the sim layout gracefully handle internationalized strings that are twice as long as the English strings? (run with query parameter stringTest=double)
• Does the sim layout gracefully handle internationalized strings that are exceptionally long? (run with query parameter stringTest=long)
• Does the sim layout gracefully handle internationalized strings that are shorter than the English strings? (run with query parameter stringTest=X)
• Does the sim stay on the sim page (doesn't redirect to an external page) when running with the query parameter stringTest=xss? This test passes if sim does not redirect, OK if sim crashes or fails to fully start. Only test on one desktop platform.
• Use named placeholders (e.g. "{{value}} {{units}}") instead of numbered placeholders (e.g. "{0} {1}").
• Make sure the string keys are all perfect - they are difficult to change after 1.0.0 is published. Strings keys should generally match their values, such as "binaryProbability": { "value": "Binary Probability" }.
• String keys for screen names should have the general form "screen.{{screenName}}", e.g. "screen.lab".
• String patterns that contain placeholders (e.g. "My name is {{first}} {{last}}") should use keys that are unlikely to conflict with strings that might be needed in the future. For example, for "{{price}}" consider using key "pricePattern" instead of "price", if you think there might be a future need for a "price" string.

Repository structure

• Are all required files and directories present?
  For a sim repository named “my-repo”, the general structure should look like this (where assets/, audio/ or images/ may be omitted if the sim doesn’t have those types of assets).

   my-repo/
      assets/
      audio/
         license.json
      doc/
         images/
               *see annotation
         model.md
         implementation-notes.md
      images/
         license.json
      js/
         (see section below)
      dependencies.json
      .gitignore
      my-repo_en.html
      my-repo-strings_en.json
      Gruntfile.js
      LICENSE
      package.json
      README.md

*Any images used in model.md or implementation-notes.md should be added here. Images specific to aiding with documentation do not need their own license.

• Is the js/ directory properly structured?
  All JavaScript source should be in the js/ directory. There should be a subdirectory for each screen (this also applies for single-screen sims, where the subdirectory matches the repo name). For a multi-screen sim, code shared by 2 or more screens should be in a js/common/ subdirectory. Model and view code should be in model/ and view/ subdirectories for each screen and common/. For example, for a sim with screens “Introduction” and “Lab”, the general directory structure should look like this:

   my-repo/
      js/
         common/
            model/
            view/
         introduction/
            model/
            view/
         lab/
            model/
            view/
         my-repo-config.js
         my-repo-main.js
         myRepo.js

• Do filenames use an appropriate prefix? Some filenames may be prefixed with the repository name, e.g. MolarityConstants.js in molarity. If the repository name is long, the developer may choose to abbreviate the repository name, e.g. EEConstants.js in expression-exchange. If the abbreviation is already used by another respository, then the full name must be used. For example, if the "EE" abbreviation is already used by expression-exchange, then it should not be used in equality-explorer. Whichever convention is used, it should be used consistently within a repository - don't mix abbreviations and full names.
• Is there a file in assets/ for every resource file in audio/ and images/? Note that there is not necessarily a 1:1 correspondence between asset and resource files; for example, several related images may be in the same .ai file. Check license.json for possible documentation of why some reesources might not have a corresponding asset file.
• Was the README.md generated by grunt published-README or grunt unpublished-README?
• Does package.json refer to any dependencies that are not used by the sim?
• Is the sim's -config.js up-to-date (generated by grunt generate-config)
• Is the LICENSE file correct? (Generally GPL v3 for sims and MIT for common code, see this thread for additional information).
• Does .gitignore match the one in simula-rasa?
• Does a GitHub issue exist for tracking credits, to ensure that they are correct before publication?
• In GitHub, verify that all non-release branches have an associated issue that describes their purpose.
• Are there any GitHub branches that are no longer needed and should be deleted?
• Does model.md adequately describe the model, in terms appropriate for teachers?
• Does implementation-notes.md adequately describe the implementation, with an overview that will be useful to future maintainers?
• Are sim-specific query parameters (if any) identified and documented in one .js file in js/common/ or js/ (if there is no common/)? The .js file should be named {{REPO}}QueryParameters.js, for example ArithmeticQueryParameters.js for the aritmetic repository.

Coding Conventions

• Is the code formatted according to PhET conventions? See phet-idea-code-style.xml for IntelliJ IDEA code style.
• Are copyright headers present and up to date? Run grunt update-copyright-dates.
• Names (types, variables, properties, functions,...) should be sufficiently descriptive and specific, and should avoid non-standard abbreviations. For example:

var numPart;            // incorrect
var numberOfParticles;  // correct

var width;              // incorrect
var beakerWidth;        // correct

• Require statements should be organized into blocks, with the code modules first, followed by strings, images and audio (any order ok for strings/images/audio). For modules, the var name should match the file name. Example below.

// modules
var inherit = require( 'PHET_CORE/inherit' );
var Line = require( 'SCENERY/nodes/Line' );
var Rectangle = require( 'SCENERY/nodes/Rectangle' );

// strings
var kineticString = require( 'string!ENERGY/energy.kinetic' );
var potentialString = require( 'string!ENERGY/energy.potential' );
var thermalString = require( 'string!ENERGY/energy.thermal' );

// images
var energyImage = require( 'image!ENERGY/energy.png' );

// audio
var kineticAudio = require( 'audio!ENERGY/energy' );

• Do the @author annotations seem correct?

• Are all constructors marked with @constructor? That will make them easier to search and review. This is not necessary
  for ES6 constructors.

• For constructors, use parameters for things that don’t have a default. Use options for things that have a default value. This improves readability at the call site, especially when the number of parameters is large. It also eliminates order dependency that is required by using parameters.

For example, this constructor uses parameters for everything. At the call site, the semantics of the arguments are difficult to determine without consulting the constructor.

/**
 * @param {Ball} ball - model element
 * @param {Property.<boolean>} visibleProperty - is the ball visible?
 * @param {Color|string} fill - fill color
 * @param {Color|string} stroke - stroke color
 * @param {number} lineWidth - width of the stroke
 * @constructor
 */
function BallNode( ball, visibleProperty, fill, stroke, lineWidth ){
   // ...
}

// Call site
var ballNode = new BallNode( ball, visibleProperty, 'blue', 'black', 2 );

Here’s the same constructor with an appropriate use of options. The call site is easier to read, and the order of options is flexible.

/**
 * @param {Ball} ball - model element
 * @param {Property.<boolean>} visibleProperty - is the ball visible?
 * @param {Object} [options]
 * @constructor
 */
function BallNode( ball, visibleProperty, options ) {

  options = _.extend( {
    fill: 'white',  // {Color|string} fill color
    stroke: 'black', // {Color|string} stroke color
    lineWidth: 1 // {number} width of the stroke
  }, options ); 

  // ...
}

// Call site
var ballNode = new BallNode( ball, visibleProperty, {
  fill: 'blue', 
  stroke: 'black', 
  lineWidth: 2 
} );

• When options are passed through one constructor to another, a "nested options" pattern should be used. This helps to avoid duplicating option names and/or accidentally overwriting options for different components that use the same option names.

Example:

/**
 * @param {ParticleBox} particleBox - model element
 * @param {Property.<boolean>} visibleProperty - are the box and its contents visible?
 * @param {Object} [options]
 * @constructor
 */
function ParticleBoxNode( particleBox, visibleProperty, options ) {

  options = _.extend( {
    fill: 'white',  // {Color|string} fill color
    stroke: 'black', // {Color|string} stroke color
    lineWidth: 1, // {number} width of the stroke
    particleNodeOptions: null, // {*} to be filled in with defaults below
  }, options );
  
  options.particleNodeOptions = _.extend( {
    fill: 'red',
    stroke: 'gray',
    lineWidth: 0.5
  }, options.particleNodeOptions );
  
  // add particle
  this.addChild( new ParticleNode( particleBox.particle, options.particleNodeOptions ) );
  
  ...
}

A possible exception to this guideline is when the constructor API is improved by hiding the implementation details, i.e. not revealing that a sub-component exists. In that case, it may make sense to use new top-level options. This is left to developer and reviewer discretion.

For more information on the history and thought process around the "nested options" pattern, please see phetsims/tasks#730.

• Constructor and function documentation. Parameter types and names should be clearly specified for each function and constructor (if there are any parameters) using @param annotations. The description for each parameter should follow a hyphen. Primitive types should use lower case. Constructors should additionally include the @constructor annotation. For example:

/** 
 * The PhetDeveloper is responsible for creating code for simulations
 * and documenting their code thoroughly.
 * 
 * @param {string} name - full name
 * @param {number} age - age, in years
 * @param {boolean} isEmployee - whether this developer is an employee of CU
 * @param {function} callback - called immediate after coffee is consumed
 * @param {Property.<number>} hoursProperty - cumulative hours worked
 * @param {string[]} friendNames - names of friends
 * @param {Object} [options] - optional configuration, see constructor
 * @constructor
 */
function PhetDeveloper( name, age, isEmployee, callback, hoursProperty, friendNames, options ) {}

• For most functions, the same form as above should be used, with a @returns annotation which identifies the return type and the meaning of the returned value. Functions should also document any side effects. For extremely simple functions that are just a few lines of simple code, an abbreviated line-comment can be used, for example: // Computes {Number} distance based on {Foo} foo.

• If references are needed to the enclosing object, such as for a closure, self should be defined, but it should only be used in closures. The self variable should not be defined unless it is needed in a closure. Example:

var self = this;
someProperty.link( function(){
  self.doSomething();
} );
this.doSomethingElse();

• Generally, lines should not exceed 120 columns. Break up long statements, expressions, or comments into multiple lines to optimize readability. It is OK for require statements or other structured patterns to exceed 120 columns. Use your judgment!

• Where inheritance is needed, use PHET_CORE/inherit (ES5) or extends (ES6). Add prototype and static functions via the appropriate arguments to inherit. Spaces should exist between the function names unless the functions are all short and closely related. Example:

  return inherit( Object, Line, {

   /**
    * Gets the slope of the line
    * @returns {number}
    */
    getSlope: function() {
      if ( this.undefinedSlope() ) {
        return Number.NaN;
      }
      else {
        return this.rise / this.run;
      }
    },

    /**
     * Given x, solve y = m(x - x1) + y1.  Returns NaN if the solution is not unique, or there is no solution (x can't 
     * possibly be on the line.)  This occurs when we have a vertical line, with no run.
     * @param {number} x - the x coordinate
     * @returns {number} the solution
     */
    solveY: function( x ) {
      if ( this.undefinedSlope() ) {
        return Number.NaN;
      }
      else {
        return ( this.getSlope() * ( x - this.x1 ) ) + this.y1;
      }
    }
  } );

• Functions should be invoked using the dot operator rather than the bracket operator. For more details, please see Style question: bracket notation vs dot notation gravity-and-orbits#9. For example:

// avoid
self[ isFaceSmile ? 'smile' : 'frown' ]();

// OK
isFaceSmile ? self.smile() : self.frown();

// OK
if ( isFaceSmile ) {
  self.smile();
}
else {
  self.frown();
}

• It is not uncommon to use conditional shorthand and short circuiting for invocation.

( expression ) && statement;
( expression ) ? statement1 : statement2;
( foo && bar ) ? fooBar() : fooCat();
( foo && bar ) && fooBar();
( foo && !(bar && fooBar)) && nowIAmConfused();
this.fill = ( foo && bar ) ? 'red' : 'blue'; 

If the expression is only one item, the parentheses can be omitted. This is the most common use case.

assert && assert( happy, ‘Why aren\’t you happy?’ );
happy && smile();
var thoughts = happy ? ‘I am happy’ : ‘I am not happy :(’;

• Naming for Property values: All AXON/Property instances should be declared with the suffix Property. For example, if a visible property is added, it should have the name visibleProperty instead of simply visible. This will help to avoid confusion with non-Property definitions.

• Properties should use type-specific subclasses where appropriate (.e.g BooleanProperty, NumberProperty, StringProperty) or provide documentation as to why they are not.

• Are Property value validation options (valueType, validValues, etc...) utilized? Is their presence or lack thereof properly documented?

• Line comments should generally be preceded by a blank line. For example:

// Randomly choose an existing crystal to possibly bond to
var crystal = this.crystals.get( _.random( this.crystals.length - 1 ) );

// Find a good configuration to have the particles move toward
var targetConfiguration = this.getTargetConfiguration( crystal );

• Line comments should have whitespace between the // and the first letter of the line comment. See the preceding example.

• Differentiate between Property and "property" in comments. They are different things. Property is a type in AXON; property is any value associated with a JavaScript object.

• Files should be named like CapitalizedCamelCasing.js when returning a constructor, or lower-case-style.js when returning a non-constructor function. When returning a constructor, the constructor name should match the filename.

• Every type, method and property should be documented.

• The HTML5/CSS3/JavaScript source code must be reasonably well documented. This is difficult to specify precisely, but the idea is that someone who is moderately experienced with HTML5/CSS3/JavaScript can quickly understand the general function of the source code as well as the overall flow of the code by reading through the comments. For an example of the type of documentation that is required, please see the example-sim repository.

• Assertions should be used appropriately and consistently. Type checking should not just be done in code comments. Use Array.isArray to type check an array.

Visibility Annotations

Because JavaScript lacks visibility modifiers (public, protected, private), PhET uses JSdoc visibility annotations to document the intent of the programmer, and define the public API. Visibility annotations are required for anything that JavaScript makes public. Information about these annotations can be found here. (Note that other documentation systems like the Google Closure Compiler use slightly different syntax in some cases. Where there are differences, JSDoc is authoritative. For example, use Array.<Object> or Object[] instead of Array<Object>). PhET guidelines for visibility annotations are as follows:

• Use @public for anything that is intended to be part of the public API.
• Use @protected for anything that is intended for use by subtypes.
• Use @private for anything that is NOT intended to be part of the public or protected API.
• Put qualifiers in parenthesis after the annotation, for example:
• To qualify that something is read-only, use @public (read-only). This indicates that the given property (AND its value) should not be changed by outside code (e.g. a Property should not have its value changed)
• To qualify that something is public to a specific repository, use (for example) @public (scenery-internal)
• For something made public solely for a11y, use @public (a11y)
• For something made public solely for phet-io, use @public (phet-io)
• Separate multiple qualifiers with commas. For example: @public (scenery-internal, read-only)
• Specify the most general type clients should know about. For example;

// @public (read-only) {Node}
this.myNode = new VeryComplicatedNodeSubclass()

• For JSDoc-style comments, the annotation should appear in context like this:

/**
 * Creates the icon for the "Energy" screen, a cartoonish bar graph.
 * @returns {Node}
 * @public
 */

• For Line comments, the annotation can appear like this:

// @public Adds a {function} listener
addListener: function( listener ) { /*...*/ }

• Verify that every JavaScript property and function has a visibility annotation. Here are some helpful regular expressions to search for these declarations as PhET uses them.

• Regex for property assignment like x.y = something: [\w]+\.[\w]+\s=
• Regex for function declarations: [\w]+: function\(

Math Libraries

• Check that dot.Util.roundSymmetric is used instead of Math.round. Math.round does not treat positive and negative numbers symmetrically, see fix nearest-neighbor rounding in Util.toFixed dot#35 (comment).
• DOT/Util.toFixed or DOT/Util.toFixedNumber should be used instead of toFixed. JavaScript's toFixed is notoriously buggy. Behavior differs depending on browser, because the spec doesn't specify whether to round or floor.
• Check that random numbers are generated using phet.joist.random, and are doing so after modules are declared (non-statically). Including, but not limited to:
• No usage of Math.random
• No usage of _.shuffle
• No usage of _.sample
• No usage of _.random
• No usage of new Random()

IE11

• No usage of Number.parseInt()
• No usage of Array.prototype.find

Organization, Readability, Maintainability

• Does the organization and structure of the code make sense? Do the model and view contain types that you would expect (or guess!) by looking at the sim? Do the names of things correspond to the names that you see in the user interface?
• Are appropriate design patterns used? See https://github.com/phetsims/phet-info/blob/master/doc/phet-software-design-patterns.md.
• Is inheritance used where appropriate? Does the type hierarchy make sense?
• Is there any unnecessary coupling? (e.g., by passing large objects to constructors, or exposing unnecessary properties/functions)
• Is there too much unnecessary decoupling? (e.g. by passing all of the properties of an object independently instead of passing the object itself)?
• Are the source files reasonable in size? Scrutinize large files with too many responsibilities - can responsibilities be broken into smaller delegates?
• Are any significant chunks of code duplicated? This will be checked manually as well as with https://github.com/danielstjules/jsinspect or grunt find-duplicates
• Is there anything that should be generalized and migrated to common code?
• Are there any TODO or FIXME comments in the code? They should be addressed or promoted to GitHub issues.
• Does the implementation rely on any specific constant values that are likely to change in the future? Identify constants that might be changed in the future. (Use your judgement about which constants are likely candidates.) Does changing the values of these constants break the sim? For example, see allow minimum rows to go to "1" and address dependency on current minimum of "5" plinko-probability#84.
• Are all dependent properties modeled as DerivedProperty instead of Property?
• All dynamics should be called from Sim.step(dt), do not use window.setTimeout or window.setInterval. This will help support Legends of Learning and PhET-iO.

PhET-iO

• If the simulation is supposed to be instrumented for PhET-iO, please see How to Instrument a PhET Simulation for PhET-iO
  for the PhET-iO development process.

[chrisklus]

Review complete. I've added a handful of review comments, but things are looking really good. Back to @pixelzoom to address review comments.

[pixelzoom]

I've addressed all REVIEW comments. The 2 that required a bit more work are covered in #55 and #103, which are assigned to @chrisklus to review. The others were trivial (see above commits), so I'm going to close this issue.