I read quite a lot about the visitor pattern and its supposed advantages. To me however it seems they are not that much advantages when applied in practice:
"Convenient" and "elegant" seems to mean lots and lots of boilerplate code
Therefore, the code is hard to follow. Also 'accept'/'visit' is not very descriptive
Even uglier boilerplate code if your programming language has no method overloading (i.e. Vala)
You cannot in general add new operations to an existing type hierarchy without modification of all classes, since you need new 'accept'/'visit' methods everywhere as soon as you need an operation with different parameters and/or return value (changes to classes all over the place is one thing this design pattern was supposed to avoid!?)
Adding a new type to the type hierarchy requires changes to all visitors. Also, your visitors cannot simply ignore a type - you need to create an empty visit method (boilerplate again)
It all just seems to be an awful lot of work when all you want to do is actually:
// Pseudocode
int SomeOperation(ISomeAbstractThing obj) {
switch (type of obj) {
case Foo: // do Foo-specific stuff here
case Bar: // do Bar-specific stuff here
case Baz: // do Baz-specific stuff here
default: return 0; // do some sensible default if type unknown or if we don't care
}
}
The only real advantage I see (which btw i haven't seen mentioned anywhere): The visitor pattern is probably the fastest method to implement the above code snippet in terms of cpu time (if you don't have a language with double dispatch or efficient type comparison in the fashion of the pseudocode above).
Questions:
So, what advantages of the visitor pattern have I missed?
What alternative concepts/data structures could be used to make the above fictional code sample run equally fast?
For as far as I have seen so far there are two uses / benefits for the visitor design pattern:
Double dispatch
Separate data structures from the operations on them
Double dispatch
Let's say you have a Vehicle class and a VehicleWasher class. The VehicleWasher has a Wash(Vehicle) method:
VehicleWasher
Wash(Vehicle)
Vehicle
Additionally we also have specific vehicles like a car and in the future we'll also have other specific vehicles. For this we have a Car class but also a specific CarWasher class that has an operation specific to washing cars (pseudo code):
CarWasher : VehicleWasher
Wash(Car)
Car : Vehicle
Then consider the following client code to wash a specific vehicle (notice that x and washer are declared using their base type because the instances might be dynamically created based on user input or external configuration values; in this example they are simply created with a new operator though):
Vehicle x = new Car();
VehicleWasher washer = new CarWasher();
washer.Wash(x);
Many languages use single dispatch to call the appropriate function. Single dispatch means that during runtime only a single value is taken into account when determining which method to call. Therefore only the actual type of washer we'll be considered. The actual type of x isn't taken into account. The last line of code will therefore invoke CarWasher.Wash(Vehicle) and NOT CarWasher.Wash(Car).
If you use a language that does not support multiple dispatch and you do need it (I can honoustly say I have never encountered such a situation though) then you can use the visitor design pattern to enable this. For this two things need to be done. First of all add an Accept method to the Vehicle class (the visitee) that accepts a VehicleWasher as a visitor and then call its operation Wash:
Accept(VehicleWasher washer)
washer.Wash(this);
The second thing is to modify the calling code and replace the washer.Wash(x); line with the following:
x.Accept(washer);
Now for the call to the Accept method the actual type of x is considered (and only that of x since we are assuming to be using a single dispatch language). In the implementation of the Accept method the Wash method is called on the washer object (the visitor). For this the actual type of the washer is considered and this will invoke CarWasher.Wash(Car). By combining two single dispatches a double dispatch is implemented.
Now to eleborate on your remark of the terms like Accept and Visit and Visitor being very unspecific. That is absolutely true. But it is for a reason.
Consider the requirement in this example to implement a new class that is able to repair vehicles: a VehicleRepairer. This class can only be used as a visitor in this example if it would inherit from VehicleWasher and have its repair logic inside a Wash method. But that ofcourse doesn't make any sense and would be confusing. So I totally agree that design patterns tend to have very vague and unspecific naming but it does make them applicable to many situations. The more specific your naming is, the more restrictive it can be.
Your switch statement only considers one type which is actually a manual way of single dispatch. Applying the visitor design pattern in the above way will provide double dispatch.
This way you do not necessarily need additional Visit methods when adding additional types to your hierarchy. Ofcourse it does add some complexity as it makes the code less readable. But ofcourse all patterns come at a price.
Ofcourse this pattern cannot always be used. If you expect lots of complex operations with multiple parameters then this will not be a good option.
An alternative is to use a language that does support multiple dispatch. For instance .NET did not support it until version 4.0 which introduced the dynamic keyword. Then in C# you can do the following:
washer.Wash((dynamic)x);
Because x is then converted to a dynamic type its actual type will be considered for the dispatch and so both x and washer will be used to select the correct method so that CarWasher.Wash(Car) will be called (making the code work correctly and staying intuitive).
Separate data structures and operations
The other benefit and requirement is that it can separate the data structures from the operations. This can be an advantage because it allows new visitors to be added that have there own operations while it also allows data structures to be added that 'inherit' these operations. It can however be only applied if this seperation can be done / makes sense. The classes that perform the operations (the visitors) do not know the structure of the data structures nor do they have to know that which makes code more maintainable and reusable. When applied for this reason the visitors have operations for the different elements in the data structures.
Say you have different data structures and they all consist of elements of class Item. The structures can be lists, stacks, trees, queues etc.
You can then implement visitors that in this case will have the following method:
Visit(Item)
The data structures need to accept visitors and then call the Visit method for each Item.
This way you can implement all kinds of visitors and you can still add new data structures as long as they consist of elements of type Item.
For more specific data structures with additional elements (e.g. a Node) you might consider a specific visitor (NodeVisitor) that inherits from your conventional Visitor and have your new data structures accept that visitor (Accept(NodeVisitor)). The new visitors can be used for the new data structures but also for the old data structures due to inheritence and so you do not need to modify your existing 'interface' (the super class in this case).
In my personal opinion, the visitor pattern is only useful if the interface you want implemented is rather static and doesn't change a lot, while you want to give anyone a chance to implement their own functionality.
Note that you can avoid changing everything every time you add a new method by creating a new interface instead of modifying the old one - then you just have to have some logic handling the case when the visitor doesn't implement all the interfaces.
Basically, the benefit is that it allows you to choose the correct method to call at runtime, rather than at compile time - and the available methods are actually extensible.
For more info, have a look at this article - http://rgomes-info.blogspot.co.uk/2013/01/a-better-implementation-of-visitor.html
By experience, I would say that "Adding a new type to the type hierarchy requires changes to all visitors" is an advantage. Because it definitely forces you to consider the new type added in ALL places where you did some type-specific stuff. It prevents you from forgetting one....
This is an old question but i would like to answer.
The visitor pattern is useful mostly when you have a composite pattern in place in which you build a tree of objects and such tree arrangement is unpredictable.
Type checking may be one thing that a visitor can do, but say you want to build an expression based on a tree that can vary its form according to a user input or something like that, a visitor would be an effective way for you to validate the tree, or build a complex object according to the items found on the tree.
The visitor may also carry an object that does something on each node it may find on that tree. this visitor may be a composite itself chaining lots of operations on each node, or it can carry a mediator object to mediate operations or dispatch events on each node.
You imagination is the limit of all this. you can filter a collection, build an abstract syntax tree out of an complete tree, parse a string, validate a collection of things, etc.
I'll take a real example I have to implement in a program I'm coding:
I have a database that has the score of every game bowled in the past three years in a bowling center. With a GUI, you can choose to either search for the best score on each lane, search for the best score between two dates, for the best score for each week, etc.
I'm wondering what the best way to implement this is. Should I code something like this:
public Vector<Scores> grabMaxScores(sortType, param1, param2)
{
if(sortType.equals("By lane"))
...
else if(sortType.equals("Between given dates")
...
}
Or is it more appropriate to code different methods for each type and call the correct one in the listener?
public Vector<Scores> grabMaxScoresBetweenDates(startDate, endDate)
{
...
}
public Vector<Scores> grabMaxScoresByLane(minLane, maxLane)
{
...
}
I'm not necessarily asking for this particular problem, it's just a question I find asking myself often when I'm coding multiple methods that are alike where the principle is the same, but the parameters are different.
I can see there are good reasons to use each of them, but I want to know if there is a "more correct" or standard way of coding this.
In my personal opinion, I would prefer your second option over the first. This is because you have the opportunity to be precise about things like the types of the parameters. For example, minLane and maxLane may just be integers, but startDate and endDate could very well be Date objects. It's often nicer if you can actually specify what you expect, as it reduces the need for such things as casting and range checks, etc. Also, I would find it more readable, as the function names just say what you are trying to do.
However, I may have an alternative idea, which is kind of a variation on your first example (I actually got this inspiration from Java's Comparator, in case you're familiar with that). Rather than pass a string as the first argument, pass some sort of Selector object. Selector would be the name of a class or a interface, which would look something like so (in Java):
interface Selector {
public void select(Score next);
public Score getBest( );
}
If the select method "likes" the value of next which is given to it, it can store the value for later. If it doesn't like it, it can simply discard it, and keep whatever value it already has. After all the data is processed, the best value will be left over, and can be requested by calling getBest. Of course, you can alter the interface to suit your particular needs (e.g. it seems like you might be expecting more than one value to be retrieved. Also, generics might help a lot as well).
The reason I like this idea is that now your function is very general purpose. In order to add new functionality, you don't need to add functions, and you don't need to modify any functions you already have. Instead, the user of your code can simply define their own implementation of Selector as they see fit. This allows your code to be far more compositional, which makes it easier to use. The only inconvenience is the need to define implementations of Selector, though, you could also provide several default ones.
The approach you have used would also work. But if you want to add some new functionality like "get lowest scores on Friday evening", you will need to add one more function, which kinda not so good thing to do.
As you have already have the data in a database you can generate database queries which would fetch the required results and display. So you need not modify your code every time.
For example, suppose I this:
class Gundam00 extends Gundam implements MobileSuit {
...
public void fight(final List<MobileSuit> mobiruSuitso, final List<Gundam> theOtherDudes, final List<Person> casualities) {
....
}
}
Suppose theOtherDudes and casualities parameters are optional. How can I make this method as clean as possible? I thought about having booleans indicating if they're null, and then checking them as needed.
I could also have different versions of the method for each combination of parameters but there would be a lot of code duplication I think.
Any suggestions?
I find that past 2-3 arguments, the ability to remember what all the arguments to a function are suffers. And comprehensibility along with it.
Passing named arguments can help. Languages with a convenient hash-like literal syntax make this really easy. Take JavaScript:
g = new Gundam00();
g.fight({opponent: enemy, casualties: 'numerous'});
You can also take advantage of variable length argument features to work this in (treat odd arguments as names, even arguments as the actual parameters).
g.fight('opponent',enemy,'casualties', 'numerous');
And some languages actually support named arguments straight-out (see: http://en.wikipedia.org/wiki/Named_parameter#Use_in_programming_languages ).
Finally, you might want to consider adding other methods for this using what some call a Fluent Interface (http://en.wikipedia.org/wiki/Fluent_interface ). Basically, you've got method call which return the object itself, so you can chain calls together:
g.opponent(enemy).casualties('numerous').fight();
This might be the easiest option if you're working in a manifestly/statically-typed class-focused language.
Update
Responding to Setsuna's comment... in that last example, if you've got the luxury, you can make methods like opponent and casualties simple setters that don't affect any internal state or computation in any other way than setting a parameter for which they're named. They simply set internal properties up, and then all of the real work happens inside action methods like fight.
If you can't do that (or if you don't like writing methods whose operations are sub-atomic), you could stake out a half-way spot between this idea with the hash-like literal idea, and create your own collection class specifically for invoking named arguments:
n = new NArgs();
g.fight(n.arg('opponent',enemy).arg('casualties','numerous').arg('motion','slow'));
A little more unwieldy, but it separates out the named arguments problem and lets you keep your methods a bit more atomic, and NArgs is probably something you could implement pretty easily just wrapping some methods around one type of Collection (HashTable?) or another that's available in your language.
Add the methods. Overloading methods is generally an antipattern and a refactoring opportunity for someone else.
http://www.codinghorror.com/blog/2007/03/curlys-law-do-one-thing.html
I thought about having booleans indicating if they're null, and then checking them inside and reacting accordingly.
Or ... you could just check if they're null.
if(theOtherDudes == null)
...
If there is only one "main method" in your class, then you can implement the optional arguments as getter/setter functions. Example:
public void setOtherDudes(final List<Gundam> theOtherDudes) {} // for input arguments
public List<Person> getCasualities() {} // for output arguments
And then, in your documentation, mention that if the caller has any optional input arguments it has to be passed in before calling fight(), and the optional output values will be available when fight() has been called.
This is worthwhile if there are dozens of optional arguments. Otherwise, I suggest overloading the method as the simplest way.
When your in a situation where you need to return two things in a single method, what is the best approach?
I understand the philosophy that a method should do one thing only, but say you have a method that runs a database select and you need to pull two columns. I'm assuming you only want to traverse through the database result set once, but you want to return two columns worth of data.
The options I have come up with:
Use global variables to hold returns. I personally try and avoid globals where I can.
Pass in two empty variables as parameters then assign the variables inside the method, which now is a void. I don't like the idea of methods that have a side effects.
Return a collection that contains two variables. This can lead to confusing code.
Build a container class to hold the double return. This is more self-documenting then a collection containing other collections, but it seems like it might be confusing to create a class just for the purpose of a return.
This is not entirely language-agnostic: in Lisp, you can actually return any number of values from a function, including (but not limited to) none, one, two, ...
(defun returns-two-values ()
(values 1 2))
The same thing holds for Scheme and Dylan. In Python, I would actually use a tuple containing 2 values like
def returns_two_values():
return (1, 2)
As others have pointed out, you can return multiple values using the out parameters in C#. In C++, you would use references.
void
returns_two_values(int& v1, int& v2)
{
v1 = 1; v2 = 2;
}
In C, your method would take pointers to locations, where your function should store the result values.
void
returns_two_values(int* v1, int* v2)
{
*v1 = 1; *v2 = 2;
}
For Java, I usually use either a dedicated class, or a pretty generic little helper (currently, there are two in my private "commons" library: Pair<F,S> and Triple<F,S,T>, both nothing more than simple immutable containers for 2 resp. 3 values)
I would create data transfer objects. If it is a group of information (first and last name) I would make a Name class and return that. #4 is the way to go. It seems like more work up front (which it is), but makes it up in clarity later.
If it is a list of records (rows in a database) I would return a Collection of some sort.
I would never use globals unless the app is trivial.
Not my own thoughts (Uncle Bob's):
If there's cohesion between those two variables - I've heard him say, you're missing a class where those two are fields. (He said the same thing about functions with long parameter lists.)
On the other hand, if there is no cohesion, then the function does more than one thing.
I think the most preferred approach is to build a container (may it be a class or a struct - if you don't want to create a separate class for this, struct is the way to go) that will hold all the parameters to be returned.
In the C/C++ world it would actually be quite common to pass two variables by reference (an example, your no. 2).
I think it all depends on the scenario.
Thinking from a C# mentality:
1: I would avoid globals as a solution to this problem, as it is accepted as bad practice.
4: If the two return values are uniquely tied together in some way or form that it could exist as its own object, then you can return a single object that holds the two values. If this object is only being designed and used for this method's return type, then it likely isn't the best solution.
3: A collection is a great option if the returned values are the same type and can be thought of as a collection. However, if the specific example needs 2 items, and each item is it's 'own' thing -> maybe one represents the beginning of something, and the other represents the end, and the returned items are not being used interchangably, then this may not be the best option.
2: I like this option the best, if 4, and 3 do not make sense for your scenario. As stated in 3, if you wanted to get two objects that represent the beginning and end items of something. Then I would use parameters by reference (or out parameters, again, depending on how it's all being used). This way your parameters can explicitly define their purpose: MethodCall(ref object StartObject, ref object EndObject)
Personally I try to use languages that allow functions to return something more than a simple integer value.
First, you should distinguish what you want: an arbitrary-length return or fixed-length return.
If you want your method to return an arbitrary number of arguments, you should stick to collection returns. Because the collections--whatever your language is--are specifically tied to fulfill such a task.
But sometimes you just need to return two values. How does returning two values--when you're sure it's always two values--differ from returning one value? No way it differs, I say! And modern languages, including perl, ruby, C++, python, ocaml etc allow function to return tuples, either built-in or as a third-party syntactic sugar (yes, I'm talking about boost::tuple). It looks like that:
tuple<int, int, double> add_multiply_divide(int a, int b) {
return make_tuple(a+b, a*b, double(a)/double(b));
}
Specifying an "out parameter", in my opinion, is overused due to the limitations of older languages and paradigms learned those days. But there still are many cases when it's usable (if your method needs to modify an object passed as parameter, that object being not the class that contains a method).
The conclusion is that there's no generic answer--each situation has its own solution. But one common thing there is: it's not violation of any paradigm that function returns several items. That's a language limitation later somehow transferred to human mind.
Python (like Lisp) also allows you to return any number of
values from a function, including (but not limited to)
none, one, two
def quadcube (x):
return x**2, x**3
a, b = quadcube(3)
Some languages make doing #3 native and easy. Example: Perl. "return ($a, $b);". Ditto Lisp.
Barring that, check if your language has a collection suited to the task, ala pair/tuple in C++
Barring that, create a pair/tuple class and/or collection and re-use it, especially if your language supports templating.
If your function has return value(s), it's presumably returning it/them for assignment to either a variable or an implied variable (to perform operations on, for instance.) Anything you can usefully express as a variable (or a testable value) should be fair game, and should dictate what you return.
Your example mentions a row or a set of rows from a SQL query. Then you reasonably should be ready to deal with those as objects or arrays, which suggests an appropriate answer to your question.
When your in a situation where you
need to return two things in a single
method, what is the best approach?
It depends on WHY you are returning two things.
Basically, as everyone here seems to agree, #2 and #4 are the two best answers...
I understand the philosophy that a
method should do one thing only, but
say you have a method that runs a
database select and you need to pull
two columns. I'm assuming you only
want to traverse through the database
result set once, but you want to
return two columns worth of data.
If the two pieces of data from the database are related, such as a customer's First Name and Last Name, I would indeed still consider this to be doing "one thing."
On the other hand, suppose you have come up with a strange SELECT statement that returns your company's gross sales total for a given date, and also reads the name of the customer that placed the first sale for today's date. Here you're doing two unrelated things!
If it's really true that performance of this strange SELECT statement is much better than doing two SELECT statements for the two different pieces of data, and both pieces of data really are needed on a frequent basis (so that the entire application would be slower if you didn't do it that way), then using this strange SELECT might be a good idea - but you better be prepared to demonstrate why your way really makes a difference in perceived response time.
The options I have come up with:
1 Use global variables to hold returns. I personally try and avoid
globals where I can.
There are some situations where creating a global is the right thing to do. But "returning two things from a function" is not one of those situations. Doing it for this purpose is just a Bad Idea.
2 Pass in two empty variables as parameters then assign the variables
inside the method, which now is a
void.
Yes, that's usually the best idea. This is exactly why "by reference" (or "output", depending on which language you're using) parameters exist.
I don't like the idea of methods that have a side effects.
Good theory, but you can take it too far. What would be the point of calling SaveCustomer() if that method didn't have a side-effect of saving the customer's data?
By Reference parameters are understood to be parameters that contain returned data.
3 Return a collection that contains two variables. This can lead to confusing code.
True. It wouldn't make sense, for instance, to return an array where element 0 was the first name and element 1 was the last name. This would be a Bad Idea.
4 Build a container class to hold the double return. This is more self-documenting then a collection containing other collections, but it seems like it might be confusing to create a class just for the purpose of a return.
Yes and no. As you say, I wouldn't want to create an object called FirstAndLastNames just to be used by one method. But if there was already an object which had basically this information, then it would make perfect sense to use it here.
If I was returning two of the exact same thing, a collection might be appropriate, but in general I would usually build a specialized class to hold exactly what I needed.
And if if you are returning two things today from those two columns, tomorrow you might want a third. Maintaining a custom object is going to be a lot easier than any of the other options.
Use var/out parameters or pass variables by reference, not by value. In Delphi:
function ReturnTwoValues(out Param1: Integer):Integer;
begin
Param1 := 10;
Result := 20;
end;
If you use var instead of out, you can pre-initialize the parameter.
With databases, you could have an out parameter per column and the result of the function would be a boolean indicating if the record is retrieved correctly or not. (Although I would use a single record class to hold the column values.)
As much as it pains me to do it, I find the most readable way to return multiple values in PHP (which is what I work with, mostly) is using a (multi-dimensional) array, like this:
function doStuff($someThing)
{
// do stuff
$status = 1;
$message = 'it worked, good job';
return array('status' => $status, 'message' => $message);
}
Not pretty, but it works and it's not terribly difficult to figure out what's going on.
I generally use tuples. I mainly work in C# and its very easy to design generic tuple constructs. I assume it would be very similar for most languages which have generics. As an aside, 1 is a terrible idea, and 3 only works when you are getting two returns that are the same type unless you work in a language where everything derives from the same basic type (i.e. object). 2 and 4 are also good choices. 2 doesn't introduce any side effects a priori, its just unwieldy.
Use std::vector, QList, or some managed library container to hold however many X you want to return:
QList<X> getMultipleItems()
{
QList<X> returnValue;
for (int i = 0; i < countOfItems; ++i)
{
returnValue.push_back(<your data here>);
}
return returnValue;
}
For the situation you described, pulling two fields from a single table, the appropriate answer is #4 given that two properties (fields) of the same entity (table) will exhibit strong cohesion.
Your concern that "it might be confusing to create a class just for the purpose of a return" is probably not that realistic. If your application is non-trivial you are likely going to need to re-use that class/object elsewhere anyway.
You should also consider whether the design of your method is primarily returning a single value, and you are getting another value for reference along with it, or if you really have a single returnable thing like first name - last name.
For instance, you might have an inventory module that queries the number of widgets you have in inventory. The return value you want to give is the actual number of widgets.. However, you may also want to record how often someone is querying inventory and return the number of queries so far. In that case it can be tempting to return both values together. However, remember that you have class vars availabe for storing data, so you can store an internal query count, and not return it every time, then use a second method call to retrieve the related value. Only group the two values together if they are truly related. If they are not, use separate methods to retrieve them separately.
Haskell also allows multiple return values using built in tuples:
sumAndDifference :: Int -> Int -> (Int, Int)
sumAndDifference x y = (x + y, x - y)
> let (s, d) = sumAndDifference 3 5 in s * d
-16
Being a pure language, options 1 and 2 are not allowed.
Even using a state monad, the return value contains (at least conceptually) a bag of all relevant state, including any changes the function just made. It's just a fancy convention for passing that state through a sequence of operations.
I will usually opt for approach #4 as I prefer the clarity of knowing what the function produces or calculate is it's return value (rather than byref parameters). Also, it lends to a rather "functional" style in program flow.
The disadvantage of option #4 with generic tuple classes is it isn't much better than returning a collection (the only gain is type safety).
public IList CalculateStuffCollection(int arg1, int arg2)
public Tuple<int, int> CalculateStuffType(int arg1, int arg2)
var resultCollection = CalculateStuffCollection(1,2);
var resultTuple = CalculateStuffTuple(1,2);
resultCollection[0] // Was it index 0 or 1 I wanted?
resultTuple.A // Was it A or B I wanted?
I would like a language that allowed me to return an immutable tuple of named variables (similar to a dictionary, but immutable, typesafe and statically checked). But, sadly, such an option isn't available to me in the world of VB.NET, it may be elsewhere.
I dislike option #2 because it breaks that "functional" style and forces you back into a procedural world (when often I don't want to do that just to call a simple method like TryParse).
I have sometimes used continuation-passing style to work around this, passing a function value as an argument, and returning that function call passing the multiple values.
Objects in place of function values in languages without first-class functions.
My choice is #4. Define a reference parameter in your function. That pointer references to a Value Object.
In PHP:
class TwoValuesVO {
public $expectedOne;
public $expectedTwo;
}
/* parameter $_vo references to a TwoValuesVO instance */
function twoValues( & $_vo ) {
$vo->expectedOne = 1;
$vo->expectedTwo = 2;
}
In Java:
class TwoValuesVO {
public int expectedOne;
public int expectedTwo;
}
class TwoValuesTest {
void twoValues( TwoValuesVO vo ) {
vo.expectedOne = 1;
vo.expectedTwo = 2;
}
}