In Eiffel, Void Safety is a way to statically prevent dereferencing uninitialised ("null") objects. The way it works is that first, the object has to be declared as detachable, and then you need to check in an if-block whether the object is actually attached (i.e. has some value) before you can use it.
This is how I have been using it until now:
some_object: detachable TYPE
...
if attached some_object then
some_object.method
end
Works perfectly fine: Without the attached-check, compiling fails with an "Target of the Object_call might be void" error.
However, after actually reading the documentation on Void Safety, I learned that this is actually how it's supposed to look like:
some_object: detachable TYPE
...
if attached some_object as l_some_object then
l_some_object.method
end
In this form, l_some_object is a variable local to the if-block which points to the same object as some_object but is statically guaranteed to be non-void.
However, I don't see the reason for the existence of this as-clause. As I pointed out above, apparently the original some_object is already statically guaranteed to be non-void within the if-block, so what's the point of introducing another variable?
What are the differences between some_object and l_some_object, apart from the scope?
Short answer
If some_object is a local variable, there is no point to introduce an object test local l_some_object.
Long answer
The general form of an object test is
attached {SOME_TYPE} expr as var
where {SOME_TYPE} and var are optional. When type ({SOME_TYPE} in the example above) is not used, the object test just checks whether expr is attached or not and assigns its value to var when it is attached.
In theory something like the following could be expected to be void-safe:
if attached expr then
expr.do_something
end
However this is not allowed in general case because expr might have side effects, so that the second time it is computed, a different value is returned, and this value might be void making the code void-unsafe:
if attached foo then -- On first call function foo returns non-void value.
foo.do_something -- On second call function foo returns void: BOOM!
end
Another possibility is an intermediate call that changes value of the expression, for example,
if attached attr then -- Attribute attr is attached here.
bar -- bar sets attr to Void.
attr.do_something -- BOOM!
end
If bar sets attribute attr to void (this can be done indirectly), the code is void-unsafe again.
Finally in a multithreading environment another thread may change the value of attr after the check and before its use inside "then" part even without any intermediate feature call:
if attached attr then -- Attribute attr is attached here.
-- Another thread sets attr to Void.
attr.do_something -- BOOM!
end
To prevent these situations, the var part is used. This object test local is read-only and is not affected by an evaluation of the same expression, by any intermediate feature call or by another thread. In other words it is always attached.
Still in some situations an object tests expression is not affected by these factors:
Arguments are read-only, so it is always sufficient to use the short form
attached arg
and it makes no sense to introduce an object test local because it will always be equal to the argument.
Local variables and Result may only become Void if they are assigned a detachable expression. If there is no such an assignment, the same
attached local_var
is just fine. However as soon as the local is assigned a detachable expression, it is not considered attached anymore:
if attached local_var then
... -- OK to use local_var as attached.
local_var := detachable_expression
... -- No guarantees about local_var attachment status.
end
If this scenario is not desired, the long form of the object test can be used
attached local_var as attached_local_var
and it guarantees that attached_local_var is always attached.
Related
When we return a value from a C++ function copy-initialisation happens. Eg:
std::string hello() {
std::string x = "Hello world";
return x; // copy-init
}
Assume that RVO is disabled.
As per copy-init rule if x is a non-POD class type, then the copy constructor should be called. However for C++11 onward, I see move-constrtuctor being called. I could not find or understand the rules regarding this https://en.cppreference.com/w/cpp/language/copy_initialization. So my first question is -
What does the C++ standard say about move happening for copy-init when value is returned from function?
As an extension to the above question, I would also like to know in what cases move does not happen. I came up with the following case where copy-constructor is called instead of move:
std::string hello2(std::string& param) {
return param;
}
Finally, in some library code I saw that std::move was being explicitly used when returning (even if RVO or move should happen). Eg:
std::string hello3() {
std::string x = "Hello world";
return std::move(x);
}
What is the advantage and disadvantage of explicitly using std::move when returning?
You are confused by the fact that initialization via the move constructor is a special case of "copy initialization", and does not come as seperate concept. Check the notes on the cppreference page.
If other is an rvalue expression, move constructor will be selected by overload resolution and called during copy-initialization. There is no such term as move-initialization.
For returning a value from the function, check the description of returning on cppreference. It says in a box called "automatic move from local variables and parameters", where expression refers to what you return (warning: that quote is shortened! read the original for full details about other cases):
If expression is a (possibly parenthesized) id-expression that names a variable whose type is [...] a non-volatile object type [...] and that variable is declared [...] in the body or as a parameter of the [...] function, then overload resolution to select the constructor to use for initialization of the returned value is performed twice: first as if expression were an rvalue expression (thus it may select the move constructor), and if the first overload resolution failed [...] then overload resolution is performed as usual, with expression considered as an lvalue (so it may select the copy constructor).
So in the special case of returning a local variable, the variable can be treated as r-value, even if normal syntactic rules would make it a l-value. The spirit of the rule is that after the return, you can't find out whether the value of the local variable has been destroyed during the copy-initialization of the returned value, so moving it does not do any damage.
Regarding your second question: It is considered bad style to use std::move while returning, because moving will happen anyway, and it inhibits NRVO.
Quoting the C++ core guidelines linked above:
Never write return move(local_variable);, because the language already knows the variable is a move candidate. Writing move in this code won’t help, and can actually be detrimental because on some compilers it interferes with RVO (the return value optimization) by creating an additional reference alias to the local variable.
So that library code you quote is suboptimal.
Also, you can not implicitly move from anything that is not local to the function (that is local variables and value parameters), because implicit moving may move from something that is still visible after the function returned. In the quote from cppreference, the important point is "a non-volatile object type". When you return std::string& param, that is a variable with reference type.
I have the following ruby method: a single do iteration without any break, next, or return. Here, cats is an array of cat objects; is_cat_red evaluates to true if cat has a color property of red.
def get_non_red_cats cats
cats.each do |cat|
!is_cat_red?(cat)
end
end
What does the method return (what does the loop evaluate to)?
This is some unusual code and it depends entirely on what the cats method does. You can pass a block to any Ruby method and that method can get executed zero more more times at any point between immediately and the end of the program's execution.
The return value is whatever cats returns, which is not clear from this snippet.
Imagine this in JavaScript terms as that language is a lot less ambiguous:
function get_non_red_cats(cats) {
return cats(function(cat) {
return !is_cat_red?(cat);
}
}
Where this shows that cats is just a function that, potentially, takes a function. It might ignore your function, too.
Now if this is cats.each that changes things as that's probably the Enumerable each method which has well-defined behaviour.
In that case the return value is whatever cats is.
There is no loop in your code. Ruby has two kinds of loops: while and for/in. (Actually, the latter is just syntactic sugar for each.)
In Ruby, an expression evaluates to the value of the last sub-expression evaluated inside the expression. A message send evaluates to the return value of the method that was executed as a result of the message send. The return value of a method is either explicitly the value of the return expression that ended the method execution or implicitly the value of the last expression evaluated inside the method body. (Note that the last expression evaluated inside the body is also what a module or class definition expression evaluates to. A method definition expression however evaluates to a Symbol denoting the name of the method.)
So, what does get_non_red_cats return? Well, there is no return in it, so it returns the value of the last expression evaluated inside the method body. The last expression evaluated inside the method body is a message send of the message each to the object referenced by the parameter binding cats. Ergo, the return value of get_non_red_cats is the return value of the method that gets executed as a result of sending the each message to cats.
And that is all we positively know.
We can make some assumptions, though. In general, each should return self. That's what all implementations of each in the entire core library and standard library do, and it is part of the standard "Iterable" Protocol in Ruby. It would be highly unusual and highly confusing if that were not the case. So, we can assume that whatever implementation of each ends up being executed, it will return self, i.e. the receiver of the message send, i.e. the object referenced by the parameter binding cats.
In other words: the method get_non_red_cats simply returns whatever was passed in as an argument. It is a pretty boring method. In fact, it is the identity method, which is pretty much the most boring method possible.
However, it could have a side-effect. You didn't ask about side-effects, only the return value, but let's look at it anyway.
Since each is supposed to simply return its receiver, it is in some sense also an identity method and thus extremely boring. However, each is generally supposed to evaluate the block it is passed, passing each element of the collection in turn as an argument. But, it ignores the value that the block evaluates to; the block is evaluated purely for its side-effect. Note that each with a block that has no side-effect makes no sense whatsoever. If the block has no side-effect, then the only thing interesting about the block is its value, but each ignores the block's value, and simply returns self.
foo.each do
# something that has no side-effect
end
is fully equivalent to
foo
Another Ruby convention is that message sends that end in a question mark ? should be used for asking questions (duh!) I.e. a message send that ends in a question mark should return something that is suitable to used as a conditional. It also generally shouldn't have a side-effect. (This is called the Command-Query Separation Principle and is a fundamental design principle of Object-Oriented Software Construction.)
And lastly, the ! unary prefix operator, when applied to something that is intended to be used in a conditional (i.e. a boolean value or something equivalent) is generally not supposed to have side-effect. Ergo, since the message send in the block ends with a question mark, it is not supposed to have a side-effect, and the ! operator is also not supposed to have a side-effect, we can assume that the entire block has no side-effect.
This, in turn, means that each shouldn't have a side-effect, and thus get_non_red_cats doesn't have a side-effect. As a result, the only other thing get_non_red_cats can do, is return a value, and it very likely simply returns the value that was passed in.
Ergo, the entire method is equivalent to
def get_non_red_cats(cats)
cats
end
All of this is assuming that the author followed standard Ruby conventions. If she didn't, then this method could do and return anything whatsoever, it could format your harddrive, launch a nuclear attack, return 42, or do absolutely nothing at all.
I have the following code:
# Assuming each element in foo is an array.
foo.each do |bar|
zaz = bar.first
# Other code using zaz, but not modifying it.
end
Will zaz local variable be modified on each iteration inside this loop, making it mutable? I am not sure about the behavior of Ruby here.
It depends on the code before the loop, really.
If that is all the code, then zaz is a block-local variable, and a new zaz variable will be created every time the loop body is evaluated.
If, however, there is a zaz local variable in the surrounding scope, then zaz is a free variable in the block, and since block scopes nest in their surrounding scope, the existing zaz variable outside the block will be re-assigned over and over again, every time the block is evaluated.
You can ensure that zaz is always treated as a block-local variable and never looked up in the surrounding scope, by explicitly declaring it as a block-local variable in the block's parameter list:
foo.each do |bar; zaz|
zaz = bar.first
end
Note, however, that your code only makes sense IFF your code is impure and mutable:
You assign to zaz but never actually use it inside the block. So, the only way that this makes sense at all is if zaz is a local variable in the outer scope and you are assigning it. Although in that case, your entire loop is just equivalent to zaz = foo.last.first.
each evaluates the block only for its side-effects. Without side-effects, each makes no sense at all, so the fact that you are using each implies that you have side-effects.
Note that the term "immutable" without additional qualification usually refers to values. When talking about "immutable variables", we usually say "immutable variable" explicitly, to make clear that we are only talking about whether or not a variable can be re-bound, not about mutating object state. Or, one could just say "constant", which is the technical term for "immutable variable" … although that term already has a specific meaning in Ruby.
each loops often mutate the object. Each has to do something.
Because each doesn't return anything useful - it returns the array itself, It won't mutate the object if it sends every element somewhere, like to be printed on screen.
foo.each do |bar|
# do something with element like
# print it, change it, save it
end
Functional alterantive is map
foo.map { |bar| bar.something }
It returns new array which is original array processed in immutable way. Obviously you have to be careful to use immutable methods. This would not be immutable:
foo.map { |bar| bar.something! }
Here something! does something destructive to the element of array.
However I have never seen map used like that. Use each for something destructive.
The following Ruby code
def a(b,c) b+c end
is the same as follows with Python
def a(b,c): return b+c
It looks like that ruby has the special storage(stack or something) that stores the final evaluation result and returns the value when a function is called.
If so, what's the name of the stack, and how can I get that stack?
If not, how does the Ruby code work without returning something?
It's not that magic, Ruby just returns the value returned by the operation that does at the end.
It's synctactic sugar that it's implemented just at parsing level: a statement that calculates something implicitly returns itself without any keyword..
to clarify it a little bit you can imagine both abstract syntax trees of the two snippets: they won't be different.
I don't think it's a stack. The final evaluation of the function is simply the return value, plain and simple. Just your everyday Ruby syntactic sugar.
I don't see any reason why a stack should be required to return a result. A simple pointer to a memory location would be sufficient. I'd guess that would usually be returned in a register, such as EAX.
You get the return value of a function by assigning the function's value to a variable (or doing something else with it). That's the way it was intended to be used, and the only way that works.
Not returning anything is really easy: The called function doesn't put anything into the return location (whatever it may be) and the caller ignores it.
Actually, return is special here, not the standard behavior. Consider:
def foo(ary)
ary.each do |e|
return true if e == 2
end
end
This code actually has more then one stack frame (at least the on for #foo, the one for Array#each and the one for the anonymous function passed to #each). What return does: it does a jump to the stack frame of the outermost lexical scope it is called in (the end of foo) and returns the given value. If you play a lot with anonymous functions, you will find that return is no allowed in all context, while just returning the last computed value is.
So I would recommend never to use return if you don't need it for precisely that reason: breaking and returning from a running iteration.
In Ruby some methods have a question mark (?) that ask a question like include? that ask if the object in question is included, this then returns a true/false.
But why do some methods have exclamation marks (!) where others don't?
What does it mean?
In general, methods that end in ! indicate that the method will modify the object it's called on. Ruby calls these as "dangerous methods" because they change state that someone else might have a reference to. Here's a simple example for strings:
foo = "A STRING" # a string called foo
foo.downcase! # modifies foo itself
puts foo # prints modified foo
This will output:
a string
In the standard libraries, there are a lot of places you'll see pairs of similarly named methods, one with the ! and one without. The ones without are called "safe methods", and they return a copy of the original with changes applied to the copy, with the callee unchanged. Here's the same example without the !:
foo = "A STRING" # a string called foo
bar = foo.downcase # doesn't modify foo; returns a modified string
puts foo # prints unchanged foo
puts bar # prints newly created bar
This outputs:
A STRING
a string
Keep in mind this is just a convention, but a lot of Ruby classes follow it. It also helps you keep track of what's getting modified in your code.
The exclamation point means many things, and sometimes you can't tell a lot from it other than "this is dangerous, be careful".
As others have said, in standard methods it's often used to indicate a method that causes an object to mutate itself, but not always. Note that many standard methods change their receiver and don't have an exclamation point (pop, shift, clear), and some methods with exclamation points don't change their receiver (exit!). See this article for example.
Other libraries may use it differently. In Rails an exclamation point often means that the method will throw an exception on failure rather than failing silently.
It's a naming convention but many people use it in subtly different ways. In your own code a good rule of thumbs is to use it whenever a method is doing something "dangerous", especially when two methods with the same name exist and one of them is more "dangerous" than the other. "Dangerous" can mean nearly anything though.
This naming convention is lifted from Scheme.
1.3.5 Naming conventions
By convention, the names of procedures
that always return a boolean value
usually end in ``?''. Such procedures
are called predicates.
By convention, the names of procedures
that store values into previously
allocated locations (see section 3.4)
usually end in ``!''. Such procedures
are called mutation procedures. By
convention, the value returned by a
mutation procedure is unspecified.
! typically means that the method acts upon the object instead of returning a result. From the book Programming Ruby:
Methods that are "dangerous," or modify the receiver, might be named with a trailing "!".
It is most accurate to say that methods with a Bang! are the more dangerous or surprising version. There are many methods that mutate without a Bang such as .destroy and in general methods only have bangs where a safer alternative exists in the core lib.
For instance, on Array we have .compact and .compact!, both methods mutate the array, but .compact! returns nil instead of self if there are no nil's in the array, which is more surprising than just returning self.
The only non-mutating method I've found with a bang is Kernel's .exit! which is more surprising than .exit because you cannot catch SystemExit while the process is closing.
Rails and ActiveRecord continues this trend in that it uses bang for more 'surprising' effects like .create! which raises errors on failure.
From themomorohoax.com:
A bang can used in the below ways, in order of my personal preference.
An active record method raises an error if the method does not do
what it says it will.
An active record method saves the record or a method saves an
object (e.g. strip!)
A method does something “extra”, like posts to someplace, or does
some action.
The point is: only use a bang when you’ve really thought about whether
it’s necessary, to save other developers the annoyance of having to
check why you are using a bang.
The bang provides two cues to other developers.
that it’s not necessary to save the object after calling the
method.
when you call the method, the db is going to be changed.
Simple explanation:
foo = "BEST DAY EVER" #assign a string to variable foo.
=> foo.downcase #call method downcase, this is without any exclamation.
"best day ever" #returns the result in downcase, but no change in value of foo.
=> foo #call the variable foo now.
"BEST DAY EVER" #variable is unchanged.
=> foo.downcase! #call destructive version.
=> foo #call the variable foo now.
"best day ever" #variable has been mutated in place.
But if you ever called a method downcase! in the explanation above, foo would change to downcase permanently. downcase! would not return a new string object but replace the string in place, totally changing the foo to downcase.
I suggest you don't use downcase! unless it is totally necessary.
!
I like to think of this as an explosive change that destroys all that has gone before it. Bang or exclamation mark means that you are making a permanent saved change in your code.
If you use for example Ruby's method for global substitutiongsub!the substitution you make is permanent.
Another way you can imagine it, is opening a text file and doing find and replace, followed by saving. ! does the same in your code.
Another useful reminder if you come from the bash world is sed -i has this similar effect of making permanent saved change.
Bottom line: ! methods just change the value of the object they are called upon, whereas a method without ! returns a manipulated value without writing over the object the method was called upon.
Only use ! if you do not plan on needing the original value stored at the variable you called the method on.
I prefer to do something like:
foo = "word"
bar = foo.capitalize
puts bar
OR
foo = "word"
puts foo.capitalize
Instead of
foo = "word"
foo.capitalize!
puts foo
Just in case I would like to access the original value again.
Called "Destructive Methods" They tend to change the original copy of the object you are referring to.
numbers=[1,0,10,5,8]
numbers.collect{|n| puts n*2} # would multiply each number by two
numbers #returns the same original copy
numbers.collect!{|n| puts n*2} # would multiply each number by two and destructs the original copy from the array
numbers # returns [nil,nil,nil,nil,nil]
My answer explains the significance of Ruby methods with exclamation marks/shebangs in the context of Ruby on Rails (RoR) model validations.
Essentially, whenever developers define Model validations (explained here), their ultimate goal is to decline a database record change & raise/throw the relevant exception(s) in case invalid data has been submitted to update the record in question.
RoR ActiveRecord gem defines various model manipulation methods (Ruby on Rails guides.). Among the methods, the valid? method is the only one that triggers validation without database action/modification. The rest of the methods attempt to change the database.
These methods trigger callbacks whenever they run. Some of the methods in the list feature a sister method with a shebang. What is the difference between the two? It has to do with the form of callback returned whenever a record validation fails.
Methods without the exclamation/shebang merely return a boolean false in the event of record validation failure while the methods with a shebang raise/throw an exception which can then be handled appropriately in code.
Just as a heads-up, since I experienced this myself.
In Ruby, ! mutates the object and returns it. Otherwise it will return nil.
So, if you are doing some kind of operations on an array for example, and call the method .compact! and there is nothig to compact, it will return nil.
Example:
arr = [1, 2, 3, nil]
arr.compact!
=> [1, 2, 3]
Run again arr.compact!
=> nil
It is better to explicitly return again the array arr if you need to use it down the line, otherwise you will get the nil value.
Example:
arr = [1, 2, 3]
arr.compact! => nil
arr # to get the value of the array