Opal is a source-to-source compiler, so there is no VM as such and the compiled code aims to be as fast and efficient as possible, mapping directly to underlying javascript features and objects where possible.
self is mostly compiled to this. Methods and blocks are implemented as javascript functions, so their this value will be the right self value. Class bodies and the top level scope use a self variable to improve readability.
true and false are compiled directly into their native boolean equivalents. This makes interaction a lot easier as there is no need to convert values to opal specific values. It does mean that there is only a Boolean ruby class available, not seperate TrueClass and FalseClass classes.
nil is compiled to null, so it can easily be passed between ruby and javascript code. undefined is also a valid value for nil, and both can be passed seamlessly. You can still call methods on nil as it is treated as an instance of NilClass.
nil # => null true # => true false # => false self # => self
Ruby strings are compiled directly into javascript strings for performance as well as readability. This has the side effect that Opal does not support mutable strings - i.e. all strings are immutable.
For performance reasons, symbols compile directly into strings. Opal supports all the symbol syntaxes, but does not have a real Symbol class. Symbols and Strings can therefore be used interchangeably.
"hello world!" # => "hello world!" :foo # => "foo" <<-EOS # => "\nHello there.\n" Hello there. EOS
In Opal there is a single class for numbers; Numeric. To keep opal as performant as possible, ruby numbers are mapped to native numbers. This has the side effect that all numbers must be of the same class. Most relevant methods from Integer, Float and Numeric are implemented on this class.
42 # => 42 3.142 # => 3.142
Ruby arrays are compiled directly into javascript arrays. Special ruby syntaxes for word arrays etc are also supported.
[1, 2, 3, 4] # => [1, 2, 3, 4] %w[foo bar baz] # => ["foo", "bar", "baz"]
Inside a generated ruby script, a function __hash is available which creates a new hash. This is also available in javascript as Opal.hash and simply returns a new instance of the Hash class.
{ :foo => 100, :baz => 700 } # => __hash("foo", 100, "baz", 700) { foo: 42, bar: [1, 2, 3] } # => __hash("foo", 42, "bar", [1, 2, 3])
Similar to hash, there is a function __range available to create range instances.
1..4 # => __range(1, 4, true) 3...7 # => __range(3, 7, false)
Opal fully supports method_missing and it is turned on by default. Opal can also send methods to nil, and the receiver is checked inline to be able to dispatch calls to native null or undefined.
In ruby, all math operators are method calls, but compiling this into javascript would end up being too slow. For this reason, math operators are optimized to test first if the receiver is a number, and if so then to just carry out the math call.
3 + 4
This ruby code will then be compiled into the following javascript:
(a = 3, b = 4, typeof(a) === "number" ? a + b : /* method call */)
This ternary statement falls back on sending a method to the receiver so all non-numeric receivers will still have the normal method call being sent. This optimization makes math operators a lot faster. Currently, the optimized method calls are +, -, * and /.
As per ruby, Opal treats only false and nil as falsy, everything else is a truthy value including "", 0 and []. This differs from javascript as these values are also treated as false.
For this reason, most truthy tests must check if values are false or nil. Note: as nil compiles to null, the truthyness tests will check null and undefined as both values are valid nil.
Taking the following test:
val = 42 if val return 3.142; end
This would be compiled into:
var val = 42; if (val !== false && val != null) { return 3.142; }
This makes the generated truthy tests (if statements, and checks and or statements) a litle more verbose in the generated code.
Instance variables in Opal work just as expected. When ivars are set or retrieved on an object, they are set natively without the @ prefix. This allows real javascript identifiers to be used which is more efficient then accessing variables by string name.
@foo = 200 @foo # => 200 @bar # => nil
This gets compiled into:
this.foo = 200; this.foo; // => 200 this.bar; // => nil
Opal tries to interact as cleanly with javascript and its api as much as possible. Ruby arrays, strings, numbers, regexps, blocks and booleans are just javascript native equivalents. The only boxed core features are hashes.
As most of the corelib deals with these low level details, opal provides a special syntax for inlining javascript code. This is done with x-strings or "backticks", as their ruby use has no useful translation in the browser.
`window.title` # => "Opal: ruby to javascript compiler" %x{ console.log("ruby version is:"); console.log(#{ OPAL_VERSION }); } # => ruby version is: # => 0.3.19
Even interpolations are supported, as seen here.
This feature of inlining code is used extensively, for example in Array#length:
class Array def length `this.length` end end
X-Strings also have the ability to automatically return their value, as used by this example.
Accessing classes and methods defined in Opal from the javascript runtime is possible via the Opal js object. The following class:
class Foo def bar puts "called bar on class Foo defined in ruby code" end end
Can be accessed from javascript like this:
Opal.Foo.$new().$bar(); // => "called bar on class Foo defined in ruby code"
Remember that all ruby methods are prefixed with a '$'.
As described above, a compiled ruby source gets generated into a string of javascript code that is wrapped inside an anonymous function. This looks similar to the following:
(function() { var _klass = Opal.klass, self = Opal.top; // generated code })();
As a complete example, assuming the following code:
puts "foo"
This would compile directly into:
(function() { var _klass = Opal.klass, self = Opal.top; self.$puts("foo"); })();
Most of the helpers are no longer present as they are not used in this example.
If you write the generated code as above into a file app.js and add that to your HTML page, then it is obvious that "foo" would be written to the browser's console.
The opal corelib includes JSON support instead of treating it as an external lib. The JSON module provides the usual parsing methods.
JSON.parse '{"a": 10, "b": [1, 2, 3], "c": null}' # => { "a" => 10, "b" => [1, 2, 3], "c" => nil }
Opal expects JSON to be present in the browser, so older browsers may require a shim (json2.js) to work with opal. Most mobile browsers and modern desktop browsers include json support natively.
Because Opal does not aim to be fully compatible with ruby, there are some instances where things can break and it may not be entirely obvious what went wrong.
As opal just generates javascript, it is useful to use a native debugger to work through javascript code. To use a debugger, simply add an x-string similar to the following at the place you wish to debug:
# .. code `debugger` # .. more code
The x-strings just pass the debugger statement straight through to the javascript output.
Inside methods and blocks, the current self value is always the native this value. You will not see self inside debuggers as it is never used to refer to the actual ruby self value.
All local variables and method/block arguments also keep their ruby names except in the rare cases when the name is reserved in javascript. In these cases, a $ suffix is added to the name (e.g. try => try$).
