Im reading Metaprogramming in Ruby.
Here are two code snippets from the book:
my_var = "Success"
MyClass = Class.new do
puts "#{my_var} in the class definition!"
define_method :my_method do
puts "#{my_var} in the method!"
end
end
MyClass.new.my_method
⇒ Success in the class definition!
Success in the method!
and:
def define_methods
shared = 0
Kernel.send :define_method, :counter do
shared
end
Kernel.send :define_method, :inc do |x|
shared += x
end
end
define_methods
counter # => 0
inc(4)
counter # => 4
I wonder why one doesnt have to use dynamic dispatch (use of Kernel.send) when defining the method in the first example while one has to use it in the second example.
I given it some thoughts but cant understand it.
Thanks!
Module#define_method is private. Private methods can only be called without a receiver. So, in the second example, you need to use reflection (i.e. send) to circumvent the access restrictions.
Note that the first example still uses dynamic dispatch. In fact, in Ruby, everything (except variable access and some builtin operators) uses dynamic dispatch.
Contrary to what the book says, it works without Kernel.send, you just need a context where the class for self is including Kernel - which is almost always the case.
for example, if you do this in the main scope, it's there:
irb(main):024:0> self.class.ancestors
=> [Object, Kernel, BasicObject]
So, using
define_method :counter do
shared
end
works fine.
You only need that trick if within the current context Kernel is not included, e.g. if you're inside a plain BasicObject, or some user created descendant class.
Related
Given the following program, in which I want to:
Create a Struct with some keys
Provide some default customization
Allow a block to be passed for further customization
module Magic
def self.MagicStruct(keys_array, &block)
Struct.new(*keys_array) do
#keys = keys_array
def self.magic_class_method
puts "constructed with #{#keys}"
end
def magic_instance_method
puts "instance method"
end
# --- DOESN'T WORK, CONTEXT IS OUTSIDE OF MODULE --- #
# yield if block_given?
instance_eval(&block) if block_given?
end
end
end
Foo = Magic.MagicStruct([:a, :b, :c]) do
puts "class customizations executing..."
def self.custom_class_method
puts "custom class method"
end
def custom_instance_method
puts "custom instance method"
end
end
Foo.magic_class_method # works
Foo.custom_class_method # works
x = Foo.new({a: 10, b: 20, c: 30})
x.magic_instance_method # works
x.custom_instance_method # fails
Output:
class customizations executing...
constructed with [:a, :b, :c]
custom class method
instance method
Traceback (most recent call last):
`<main>': undefined method `custom_instance_method' for #<struct Foo a={:a=>10, :b=>20, :c=>30}, b=nil, c=nil> (NoMethodError)
Why is the self.custom_class_method correctly added to the Foo class, but the custom_instance_method is not? This usage is clearly stated in the Struct documentation, so I'm afraid there's some kind of scoping or context issue I'm missing here.
I would prefer to keep the nice def method() ... end syntax rather than resorting to having a strict requirement to use define_method("method") in the customization block, which does happen to work.
In Ruby there is the notion of a current class which is the target of keywords such as def.
When you use instance_eval, the current class is set to self.singleton_class. In other words, def x and def self.x are equivalent. In your code, custom_instance_method is defined on the singleton class of the newly created Struct, making it a class method.
When you use class_eval, the current class is set to self. Since this method is only available on classes, it will set the current class to the one you called the method on. In other words, def x will define a method that's available to all objects of that class. This is what you wanted.
For more details, see my other answer.
This is a gem I wrote that I believe does most, if not all that you want (not so sure about adding class methods though, I'd have to play with that). While the Struct created is immutable, you can have a look at the code and alter it to meet your needs; it's pretty simple. What you'd be interested in, would be here, which basically amounts to what you see below. The extension of the modules uses instance_eval in a way that I believe is what you want as well:
# https://github.com/gangelo/immutable_struct_ex/blob/main/lib/immutable_struct_ex.rb
# Defines the methods used to create/manage the ImmutableStructEx struct.
module ImmutableStructEx
class << self
# Most likely changing this method name to #create in subsequent version, but alas...
def new(**hash, &block)
Struct.new(*hash.keys, keyword_init: true, &block).tap do |struct|
return struct.new(**hash).tap do |struct_object|
struct_object.extend Comparable
struct_object.extend Immutable
end
end
end
end
end
Usage details can be found in the README.md in the github repository.
I hope this help, at least in part.
I'm not sure where I have seen this, or if I just think I have seen it, but I would like to be able to call a method that creates an instance of a class with the same name. So, instead of:
# The class is called 'Vector3', for example:
Vector3.new(x,y,z)
I would like to have an eponymous method that instantiates the class, like so:
Vector3(x,y,z) #<- returns instance of Vector3 class
How would you define that in Ruby?
As #Eli mentioned, you can define a method in the kernel:
def Kernel.Vector3(*args)
Vector3.new(*args)
end
Ruby does this for Array, Complex, Float, Hash, Integer, Rational and String where you probably saw it.
The reason it works is that Object includes Kernel, hence all objects in your program (except ones directly inheriting from BasicObject) will have that method.
However, doing so is unidiomatic, unnecessary (you clutter all objects with an additional method), confusing (capitalized identifiers should be constants) and generally looked down upon.
AFAIK you can't. You can do something similar, Vector3[x, y, z].
class Vector3
def initialize(x, y, z)
# ...
end
def self.[](*args)
self.new(*args)
end
end
Note that the Ruby library uses this device as well. There's Hash.new(...) and Hash[...] , but no Hash(...). This parallels how Proc objects are invoked:
greet = Proc.new { |name| puts "Hello, #{name}" }
greet["Amadan"]
EDIT: I stand corrected:
module Kernel
def Vector3(*args)
Vector3.new(*args)
end
end
But, as Eli Sadoff said, it is impractical, violates encapsulation, and Ruby style (functions and methods should be lowercase).
This answers what I understood the question to be from the title. (Somehow I overlooked the example that contradicts that.) I will leave my answer because I think it includes some interesting elements.
class MyClass
def hi
"hi"
end
my_alias = to_s
singleton_class.class_eval { alias_method(my_alias, :new) }
end
MyClass.methods(false)
#=> [:MyClass]
my_instance = MyClass.MyClass
#=> #<MyClass:0x007fadc2092320>
my_instance.hi
#=> "hi"
Note that this works when the alias of new is passed arguments and/or a block.
See Object#singleton_class and Module#class_eval.
Just want to enquire what the right practice is.
My preference is to use procs, simply because I think that defining methods inside of methhods is a bit untidy and should be done only when necessary. To get around it, I simply use procs.
What is the right / better way to do it and why? (apart from the proc's ability to access the main method's variables defined before itself)
def meth( params_prime )
calculations = do_something_with_whatever
def sub_meth( params_sub )
do_something_with_params_sub
end
sub_meth_params(calculations) # is this better?
proc1 = proc{ |params_sub| do_something_with_params_sub }
proc1.call(calculations) # or is this?
end
It is not clear what your specific use-case is, but I would definitely go for procs or lambdas. There is less overhead when defining a proc or lambda dynamically, they are passable, so if needed you could return them and they could be used outside the function.
Using "def" exposes the method as an instance method outside of the current method scope (so in the containing class, which could be Object in your case). This may or may not be with you want. If you want to use an anonymous function only available in the local scope, use a lambda.
Also Proc vs Lambda: I generally prefer to use lambdas since they behave a little more "predictable", meaning: as you would expect (check passed variables, and return just returns from the lambda, proc returns from the called scope). But from your example it is hard to deduce what would apply. I think the key-difference is: lambas are ment to be passed around, and thus behave a little more sanely. If this is not your use-case, use Proc :) (a write-up of the difference).
If you want to use sub_func to encapsulate it from call from other methods you can use a class to group function and sub_func together and make sub_func private. Otherwise if you want to pass this function as a parameter further you can declare it as lamda.
def func params_prime
sub_func = ->(params_sub){do_something_with_params}
sub_func.call(params_prime)
end
Defining methods inside methods is a feature of Ruby that may have its use. But something is telling me that you are asking a very advanced question while you are still a beginner level Rubyist. Do you know what default definee is? If not, check this article by Yugui.
Procs are very important in Ruby, but newbies tend to use them instead of defining methods in appropriate objects, which is the exact smell I'm getting from your question. The normal way of doing things in OO languages of Ruby family is to define methods on objects:
class Foo
def bar *params
# do something with params
end
end
Since you do not understand the meaning of defining methods inside methods, refrain from doing it for the next 6 months. Once you understand objects, you can start experimenting with this very advanced feature again.
APPENDIX:
Since you demonstrated intrest, let me show you that using def in def at the top level is a frownable-upon thing to do. Normally, when you define a method on some class without further adornment, it becomes a public instance method of that class:
class X
def foo; "foo" end
end
X.instance_methods.include? :foo
#=> true
When you use def in a def, the definee for the inner def is going to be X:
class X
def bar
def baz
"baz"
end
"bar"
end
end
When you execute the above code, instance method #bar becomes defined on X:
X.instance_methods.include? :bar
#=> true
But #baz not yet:
X.instance_methods.include? :baz
#=> false
Only after you call #bar at least once does the method become defined on X:
X.new.bar
#=> "bar"
X.instance_methods.include? :baz
#=> true
And now I would like to ask you to appreciate how terrible thing just happened: An instance just modified its mother class. That's a violation. A violation of such a basic principle of OO design, that I'm not even sure it has a name. This technique is great for obfuscated coding competitions, but in production, it's taboo. Ruby gives you the freedom to break that taboo, gives you the rope to hang yourself on, but you don't do it under any kind of normal circumstances.
So what can be worse than a def inside a def in a class definition? The answer is, a def inside a def at the top level. Let me show you why. Normally, when you define methods with def at the top level, the default definee is Object, but the top level defnitions become private instance methods of object. This is to prevent the unintended consequence of top level defs, because almost all Ruby objects inherit from Object. For example, if you define:
class Object
def foo; "foo" end
end
Now all your objects will respond to foo:
foo #=> "foo"
1.foo #=> "foo"
[].foo #=> "foo
When we define methods at the top level, we usually just intend to use the method at the top level, and don't want every single object to inherit it. For that reason, top level defs become private:
hello #=> NameError: undefined local variable or method `hello' for main:Object
1.hello #=> NoMethodError: undifined method 'hello' for 1:Fixnum
Now we use def at the top level:
def hello; "hello" end
We can see that method #hello is has not become an instance methods of Object:
Object.instance_methods.include? :hello
#=> false
Mysteriously, it became its private method:
Object.private_instance_methods.include? :hello
#=> true
This way, we avoid the unintended consequence of defining #hello method for every single object. But the inheritance is there. The error message has changed:
1.hello #=> NoMethodError: private method 'hello' called for 1:Fixnum
And we can forcibly call the method via #send:
1.send :hello
#=> "hello"
Mysteriously, at the top level, we are allowed to call this private method without #send:
hello
#=> "hello"
And now, what happens when you do def in def at the top level:
def bar
def baz; "baz" end
"bar"
end
You define a private instance method Object#bar in an expected way. But when you call it, alas, the top level magic no longer works and a public method Object#baz gets defined:
bar #=> "bar"
This way, not just the top level, but every single Ruby object got polluted with your #baz method:
1.baz #=> "baz"
Class.baz #=> "baz"
This is why I told you to refrain from using this idiom until you progress from the level of unconscious incompetence to the level of conscious incompetence. I recommend you to read more about top level methods in Ruby.
I am reading the Module documentation but can't seem to understand their differences and which should be used where.
How is the eval different than exec?
I'm going to answer a bit more than your question by including instance_{eval|exec} in your question.
All variations of {instance|module|class}_{eval|exec} change the current context, i.e. the value for self:
class Array
p self # prints "Array"
43.instance_eval{ p self } # prints "43"
end
Now for the differences. The eval versions accepts a string or a block, while the exec versions only accept a block but allow you to pass parameters to it:
def example(&block)
42.instance_exec("Hello", &block)
end
example{|mess| p mess, self } # Prints "Hello" then "42"
The eval version does not allow to pass parameters. It provides self as the first parameter, although I can't think of a use for this.
Finally, module_{eval|exec} is the same as the corresponding class_{eval|exec}, but they are slightly different from instance_{eval|exec} as they change what is the current opened class (i.e. what will be affected by def) in different ways:
String.instance_eval{ def foo; end }
Integer.class_eval { def bar; end }
String.method_defined?(:foo) # => false
String.singleton_methods.include?(:foo) # => true
Integer.method_defined?(:bar) # => true
So obj.instance_{eval|exec} opens the singleton class of obj, while mod.{class|module}_{eval|exec} opens mod itself.
Of course, instance_{eval|exec} are available on any Ruby object (including modules), while {class|module}_* are only available on Module (and thus Classes)
To answer your last question first, eval (in all its variations) is completely different from exec. exec $command will start a new process to run the command you specify and then exit when that finishes.
class_eval and module_eval have the power to redefine classes and modules -- even those that you yourself did not write. For example, you might use class eval to add a new method that did not exist.
Fixnum.class_eval { def number; self; end }
7.number # returns '7'
class_eval can be used to add instance methods, and instance_eval can be used to add class methods (yes, that part is very confusing). A class method would be something like Thing.foo -- you're literally calling the foo method on the Thing class. An instance method is like the example above, using class_eval I've added a number method to every instance of Fixnum.
Okay, so that's the *_eval class of methods. The exec methods are similar, but they allow you to look inside a class and execute a block of code as though it was defined as a method on that class. Perhaps you have a class that looks like this:
class Foo
##secret = 'secret key'
##protected = 'some secret value'
def protected(key)
if key == ##secret
return ##protected
end
end
end
The class Foo is just a wrapper around some secret value, if you know the correct key. However, you could trick the class into giving you its secrets by executing a block inside the context of the class like so:
Foo.class_exec { ##secret = 'i'm a hacker' }
Foo.protected('i'm a hacker') #returns the value of ##protected because we overwrote ##secret
In general, with a lot of the tools in ruby, you could use any of these to solve a lot of problems. A lot of the time you probably won't even need to unless you want to monkey patch a class some library you use has defined (although that opens up a whole can of worms). Try playing around with them in irb and see which you find easier. I personally don't use the *_exec methods as much as the *_eval methods, but that's a personal preference of mine.
To avoid ambiguity I'm going to call a method that belongs to (owned by) a singleton class a singleton method. The rest are instance methods. Although one might say that a singleton method of an object is an instance method of its singleton class.
tl;dr Use class_eval/module_eval on a class/module to define instance methods, and instance_eval on a class/module to define class methods (or to be more precise, use instance_eval to define singleton methods). Additionally you can use instance_eval to access instance variables.
A terminology is a bit lacking in this case. ruby maintains a stack of class references (cref for short). When you open/reopen a class, the corresponding class reference is pushed to the stack. And the current class refernece affects where def defines methods (to which class/module they're added).
Now, class_eval/module_eval and class_exec/module_exec are aliases.
The *_exec() variants don't accept strings, and allow to pass arguments to the block. Since the *_eval() variants are mainly used I'll focus on them.
class_eval/module_eval changes cref and self to the receiver (Thing in Thing.module_eval(...)):
rb_mod_module_eval() -> specific_eval()
yield_under() (for blocks)
vm_cref_push()
eval_under() (for strings)
vm_cref_push()
instance_eval changes cref to the singleton class of the receiver, and self to the receiver.
Let's see them in action:
class A
p self #=> A
#a = 1
def initialize
#b = 2
end
end
p A.instance_variables #=> [:#a]
p A.new.instance_variables #=> [:#b]
#a on a class level adds an instance variable to the class A as an object. I'm adding it here for completeness. But that's not how you add a class variable.
A.instance_eval do
p self #=> A
p #a #=> 1
def m() puts 'm' end
end
sclass = A.singleton_class
p sclass.instance_methods(false).include? :m #=> true
A.m #=> m
a = A.new
a.instance_eval do
p self #=> #<A:0x00007fc497661be8 #b=2>
p #b #=> 2
def m2() puts 'm2' end
end
sclass = a.singleton_class
p sclass.instance_methods(false).include? :m2 #=> true
a.m2 #=> m2
So, inside instance_eval def adds a singleton method to the receiver (an instance method to the singleton class of the receiver). For a class/module that means a class/module method. For other objects, a method that is available for that particular object.
A.class_eval do
p self #=> A
p #a #=> 1
def m() puts 'm' end
end
p A.instance_methods(false).include? :m #=> true
A.new.m #=> m
And, inside class_eval def adds an instance method to the receiver itself (the class/module). class_eval is only available for classes/modules.
Also, when class_eval is passed a block, constant/class variable lookup is not affected:
module A
C = 1
##c = 1
class B
C = 2
##c = 2
end
A::B.class_eval { p [C, ##c] } #=> [1, 1]
A::B.class_eval 'p [C, ##c]' #=> [2, 2]
end
The naming is confusing. I might guess that instance in instance_eval suggests that receiver is treated as an instance (allows to change things for a particular instance), and class in class_eval as a class (allows to change things for a class of objects).
When I invoke a method that doesn't exist, method_missing will tell me the name of the method. When I attempt to access a variable that hasn't been set, the value is simply nil.
I'm attempting to dynamically intercept access to nil instance variables and return a value based on the name of the variable being accessed. The closest equivalent would be PHP's __get. Is there any equivalent functionality in Ruby?
I do not believe this is possible in Ruby. The recommended way would be to use a ''user'' method rather than a ''#user'' instance var in your templates.
This is consistent with the way you deal with Ruby objects externally (''obj.user'' is a method which refers to ''#user'', but is actually not ''#user'' itself). If you need any kind of special logic with an attribute, your best bet is to use a method (or method_missing), regardless if you're accessing it from inside or outside the object.
See my answer to another similar question. But just because you can do it doesn't mean that it's a good idea. Sensible design can generally overcome the need for this kind of thing and allow you to produce more readable and hence maintainable code.
instance_variable_get seems to be the closest equivalent of PHP's __get from what I can see (although I'm not a PHP user).
Looking at the relevant Ruby source code, the only 'missing' method for variables is const_missing for constants, nothing for instance variables.
there isn't an instance_variable_missing (at least that I know of)
But why are you accessing randomly named instance variables anyway?
If your thread all the access to the object state through method calls (as you should anyway) then you wouldn't need this.
If you are looking for a way to define magic stuff without messing up with the method lookup, you may want to use const_missing.
A bit late but, instance_variable_missing is the same as method_missing to a point... Take the following class:
class Test
def method_missing(*args)
puts args.inspect
end
end
t = Test.new
Now let's get some instance variables:
t.pineapples #=> [:pineapples]
t.pineapples = 5 #=> [:pineapples=,5]
Not sure why the method is nil for you...
EDIT:
By the sounds of it you want to accomplish:
t = SomeClass.new
t.property.child = 1
So let's try returning a Test object from our previous example:
class Test
def method_missing(*args)
puts args.inspect
return Test.new
end
end
So what happens when we call:
t = Test.new
t.property.child = 1
#=>[:property]
#=>[:child=,1]
So this goes to show that this is indeed possible to do. OpenStruct uses this same technique to set instance variables dynamically. In the below example, I create EternalStruct which does exactly what you wanted:
require 'ostruct'
class EternalStruct < OpenStruct
def method_missing(*args)
ret = super(*args)
if !ret
newES = EternalStruct.new
self.__send__((args[0].to_s + "=").to_sym, newES)
return newES
end
end
end
Usage of EternalStruct:
t = EternalStruct.new
t.foo.bar.baz = "Store me!"
t.foo.bar.baz #=> "Store me!"
t.foo #=> #<EternalStruct bar=#<EternalStruct baz="Store me!">>
t.a = 1
t.a #=> 1
t.b #=> #<EternalStruct:...>
t.b = {}
t.b #=> {}
def t.c(arg)
puts arg
end
t.c("hi there") #=> "hi there"