函数式编程会取代 GoF 设计模式吗？

Question

Since I started learningF# and OCaml last year, I've read a huge number of articles which insist that design patterns (especially in Java) are workarounds for the missing features in imperative languages. One article I found makes a fairly strong claim :

Most people I've met have read the Design Patterns book by the Gang of Four (GoF). Any self respecting programmer will tell you that the book is language agnostic and the patterns apply to software engineering in general, regardless of which language you use. This is a noble claim. Unfortunately it is far removed from the truth.

Functional languages are extremely expressive. In a functional language one does not need design patterns because the language is likely so high level, you end up programming in concepts that eliminate design patterns all together.

The main features of functional programming (FP) include functions as first-class values, currying, immutable values, etc. It doesn't seem obvious to me that OO design patterns are approximating any of those features.

Additionally, in functional languages which support OOP (such as F# and OCaml), it seems obvious to me that programmers using these languages would use the same design patterns found available to every other OOP language. In fact, right now I use F# and OCaml every day, and there are no striking differences between the patterns I use in these languages vs. the patterns I use when I write in Java.

Is there any truth to the claim that functional programming eliminates the need for OOP design patterns? If so, could you post or link to an example of a typical OOP design pattern and its functional equivalent?

Answer 1

The blog post you quoted overstates its claim a bit. FP doesn't eliminate the need for design patterns. The term "design patterns" just isn't widely used to describe the same thing in FP languages. But they exist. Functional languages have plenty of best practice rules of the form "when you encounter problem X, use code that looks like Y", which is basically what a design pattern is.

However, it's correct that most OOP-specific design patterns are pretty much irrelevant in functional languages.

I don't think it should be particularly controversial to say that design patterns in general only exist to patch up shortcomings in the language. And if another language can solve the same problem trivially, that other language won't have need of a design pattern for it. Users of that language may not even be aware that the problem exists , because, well, it's not a problem in that language.

Here is what the Gang of Four has to say about this issue:

The choice of programming language is important because it influences one's point of view. Our patterns assume Smalltalk/C++-level language features, and that choice determines what can and cannot be implemented easily. If we assumed procedural languages, we might have included design patterns called "Inheritance", "Encapsulation," and "Polymorphism". Similarly, some of our patterns are supported directly by the less common object-oriented languages. CLOS has multi-methods, for example, which lessen the need for a pattern such as Visitor. In fact, there are enough differences between Smalltalk and C++ to mean that some patterns can be expressed more easily in one language than the other. (See Iterator for example.)

(The above is a quote from the Introduction to the Design Patterns book, page 4, paragraph 3)

The main features of functional programming include functions as first-class values, currying, immutable values, etc. It doesn't seem obvious to me that OO design patterns are approximating any of those features.

What is the command pattern, if not an approximation of first-class functions?:) In an FP language, you'd simply pass a function as the argument to another function. In an OOP language, you have to wrap up the function in a class, which you can instantiate and then pass that object to the other function. The effect is the same, but in OOP it's called a design pattern, and it takes a whole lot more code. And what is the abstract factory pattern, if not currying? Pass parameters to a function a bit at a time, to configure what kind of value it spits out when you finally call it.

So yes, several GoF design patterns are rendered redundant in FP languages, because more powerful and easier to use alternatives exist.

But of course there are still design patterns which are not solved by FP languages. What is the FP equivalent of a singleton? (Disregarding for a moment that singletons are generally a terrible pattern to use.)

And it works both ways too. As I said, FP has its design patterns too; people just don't usually think of them as such.

But you may have run across monads. What are they, if not a design pattern for "dealing with global state"? That's a problem that's so simple in OOP languages that no equivalent design pattern exists there.

We don't need a design pattern for "increment a static variable", or "read from that socket", because it's just what you do .

Saying a monad is a design pattern is as absurd as saying the Integers with their usual operations and zero element is a design pattern. No, a monad is a mathematical pattern , not a design pattern.

In (pure) functional languages, side effects and mutable state are impossible, unless you work around it with the monad "design pattern", or any of the other methods for allowing the same thing.

Additionally, in functional languages which support OOP (such as F# and OCaml), it seems obvious to me that programmers using these languages would use the same design patterns found available to every other OOP language. In fact, right now I use F# and OCaml everyday, and there are no striking differences between the patterns I use in these languages vs the patterns I use when I write in Java.

Perhaps because you're still thinking imperatively? A lot of people, after dealing with imperative languages all their lives, have a hard time giving up on that habit when they try a functional language. (I've seen some pretty funny attempts at F#, where literally every function was just a string of 'let' statements, basically as if you'd taken a C program, and replaced all semicolons with 'let'.:))

But another possibility might be that you just haven't realized that you're solving problems trivially which would require design patterns in an OOP language.

When you use currying, or pass a function as an argument to another, stop and think about how you'd do that in an OOP language.

Is there any truth to the claim that functional programming eliminates the need for OOP design patterns?

Yep.:) When you work in a FP language, you no longer need the OOP-specific design patterns. But you still need some general design patterns, like MVC or other non-OOP specific stuff, and you need a couple of new FP-specific "design patterns" instead. All languages have their shortcomings, and design patterns are usually how we work around them.

Anyway, you may find it interesting to try your hand at "cleaner" FP languages, like ML (my personal favorite, at least for learning purposes), orHaskell , where you don't have the OOP crutch to fall back on when you're faced with something new.

---

As expected, a few people objected to my definition of design patterns as "patching up shortcomings in a language", so here's my justification:

As already said, most design patterns are specific to one programming paradigm, or sometimes even one specific language. Often, they solve problems that only exist in that paradigm (see monads for FP, or abstract factories for OOP).

Why doesn't the abstract factory pattern exist in FP? Because the problem it tries to solve does not exist there.

So, if a problem exists in OOP languages, which does not exist in FP languages, then clearly that is a shortcoming of OOP languages. The problem can be solved, but your language does not do so, but requires a bunch of boilerplate code from you to work around it. Ideally, we'd like our programming language to magically make all problems go away. Any problem that is still there is in principle a shortcoming of the language.;)

Answer 2

Is there any truth to the claim that functional programming eliminates the need for OOP design patterns?

Functional programming is not the same as object-oriented programming. Object-oriented design patterns don't apply to functional programming. Instead, you have functional programming design patterns.

For functional programming, you won't read the OO design pattern books; you'll read other books on FP design patterns.

language agnostic

Not totally. Only language-agnostic with respect to OO languages. The design patterns don't apply to procedural languages at all. They barely make sense in a relational database design context. They don't apply when designing a spreadsheet.

a typical OOP design pattern and its functional equivalent?

The above shouldn't exist. That's like asking for a piece of procedural code rewritten as OO code. Ummm... If I translate the original Fortran (or C) into Java, I haven't done anything more than translate it. If I totally rewrite it into an OO paradigm, it will no longer look anything like the original Fortran or C -- it will be unrecognizable.

There's no simple mapping from OO design to functional design. They're very different ways of looking at the problem.

Functional programming (like all styles of programming) has design patterns. Relational databases have design patterns, OO has design patterns, and procedural programming has design patterns. Everything has design patterns, even the architecture of buildings.

Design patterns -- as a concept -- are a timeless way of building, irrespective of technology or problem domain. However, specific design patterns apply to specific problem domains and technologies.

Everyone who thinks about what they're doing will uncover design patterns.

Answer 3

Brian's comments on the tight linkage between language and pattern is to the point,

The missing part of this discussion is the concept of idiom. James O. Coplien's book, "Advanced C++" was a huge influence here. Long before he discovered Christopher Alexander and the Column Without a Name (and you can't talk sensibly about patterns without reading Alexander either), he talked about the importance of mastering idioms in truly learning a language. He used string copy in C as an example, while(*from++ = *to++); You can see this as a bandaid for a missing language feature (or library feature), but what really matters about it is that it's a larger unit of thought, or of expression, than any of its parts.

That is what patterns, and languages, are trying to do, to allow us to express our intentions more succinctly. The richer the units of thought the more complex the thoughts you can express. Having a rich, shared vocabulary at a range of scales - from system architecture down to bit twiddling - allows us to have more intelligent conversations, and thoughts about what we should be doing.

We can also, as individuals, learn. Which is the entire point of the exercise. We each can understand and use things we would never be able to think of ourselves. Languages, frameworks, libraries, patterns, idioms and so on all have their place in sharing the intellectual wealth.

Answer 4

The GoF book explicitly ties itself to OOP - the title is Design Patterns - Elements of Reusable Object-Oriented Software (emphasis mine).

Answer 5

Design Patterns in Dynamic Programming by Peter Norvig has thoughtful coverage of this general theme, though about 'dynamic' languages instead of 'functional' (there's overlap).

Answer 6

Here's another link, discussing this topic: http://blog.ezyang.com/2010/05/design-patterns-in-haskel/

In his blog post Edward describes all 23 original GoF patterns in terms of Haskell.

Answer 7

When you try to look at this at the level of "design patterns" (in general) and "FP versus OOP", the answers you'll find will be murky at best.

Go a level deeper on both axes, though, and consider specific design patterns and specific language features and things become clearer.

So, for example, some specific patterns, like Visitor , Strategy , Command , and Observer definitely change or disappear when using a language with algebraic data types and pattern matching , closures , first class functions , etc. Some other patterns from the GoF book still 'stick around', though.

In general, I would say that, over time, specific patterns are being eliminated by new (or just rising-in-popularity) language features. This is the natural course of language design; as languages become more high-level, abstractions that could previously only be called out in a book using examples now become applications of a particular language feature or library.

(Aside: here's a recent blog I wrote, which has other links to more discussion on FP and design patterns.)

Answer 8

I would say that when you have a language like Lisp with its support for macros, then you can build you own domain-specific abstractions, abstractions which often are much better than the general idiom solutions.

Answer 9

Norvig's presentation alludes to an analysis they did of all the GoF patterns, and they say that 16 of the 23 patterns had simpler implementations in functional languages, or were simply part of the language. So presumably at least seven of them either were a) equally complicated or b) not present in the language. Unfortunately for us, they are not enumerated

I think it's clear that most of the "creational" or "structural" patterns in GoF are merely tricks to get the primitive type systems in Java or C++ to do what you want. But the rest are worthy of consideration no matter what language you program in.

One might be Prototype; while it is a fundamental notion of JavaScript, it has to be implemented from scratch in other languages.

One of my favorite patterns is the Null Object pattern: represent the absence of something as an object that does an appropriate kind of nothing. This may be easier to model in a functional language. However, the real achievement is the shift in perspective.

Answer 10

And even the OO design pattern solutions are language specific.

Design patterns are solutions to common problems that your programming language doesn't solve for you. In Java, the Singleton pattern solves the one-of-something (simplified) problem.

In Scala, you have a top level construct called Object in addition to Class. It's lazily instantiated and there is only one.You don't have to use the Singleton pattern to get a Singleton. It's part of the language.

Answer 11

Patterns are ways of solving similar problems that get seen again and again, and then get described and documented. So no, FP is not going to replace patterns; however, FP might create new patterns, and make some current "best practices" patterns "obsolete".

Answer 12

The GoF Design Patterns is coding workaround recipes for OO languages that are descendants of Simula 67 , like Java and C++.

Most of the "ills" treated by the design patterns are caused by:

• statically typed classes, which specify objects, but are not themselves objects;
• restriction to single dispatch (only the leftmost argument is used to select a method, the remaining arguments are considered as static types only: if they have dynamic types, it's up to the method to sort that out with ad-hoc approaches);
• distinction between regular function calls and object-oriented function calls, meaning that object-oriented functions cannot be passed as functional arguments where regular functions are expected and vice versa; and
• distinction between "base types" and "class types".

There isn't a single one of these design patterns that doesn't disappear in the Common Lisp Object System, even though the solution is structured in essentially the same way as in the corresponding design pattern. (Moreover, that object system precedes the GoF book by well over a decade. Common Lisp became an ANSI standard the same year that that book was first published.)

As far as functional programming is concerned, whether or not the patterns apply to it depends on whether the given functional programming language has some kind of object system, and whether it is modeled after the object systems which benefit from the patterns. That type of object-orientation does not mix well with functional programming, because the mutation of state is at the front and centre.

Construction and non-mutating access are compatible with functional programming, and so patterns which have to do with abstracting access or construction could be applicable: patterns like Factory, Facade, Proxy, Decorator, and Visitor.

On the other hand, the behavioral patterns like State and Strategy probably do not directly apply in functional OOP because mutation of state is at their core. This doesn't mean they don't apply; perhaps they somehow apply in combination with whatever tricks are available for simulating mutable state.

Answer 13

As others have said, there are patterns specific to functional programming. I think the issue of getting rid of design patterns is not so much a matter of switching to functional, but a matter of language features .

Take a look at how Scala does away with the "singleton pattern": you simply declare an object instead of a class. Another feature, pattern matching, helps avoiding the clunkiness of the visitor pattern. See the comparison here: Scala's Pattern Matching = Visitor Pattern on Steroids

And Scala, like F#, is a fusion of OO-functional. I don't know about F#, but it probably has these kind of features.

Closures are present in functional language, but they need not be restricted to them. They help with the delegator pattern.

One more observation. This piece of code implements a pattern: it's such a classic and it's so elemental that we don't usually think of it as a "pattern", but it sure is:

 for(int i = 0; i < myList.size(); i++) { doWhatever(myList.get(i)); }

Imperative languages like Java and C# have adopted what is essentially a functional construct to deal with this: "foreach".

Answer 14

I'd like to plug a couple of excellent but somewhat dense papers by Jeremy Gibbons: "Design patterns as higher-order datatype-generic programs" and "The essence of the Iterator pattern" (both available here: http://www.comlab.ox.ac.uk/jeremy.gibbons/publications/ ).

These both describe how idiomatic functional constructs cover the terrain that is covered by specific design patterns in other (object-oriented) settings.

Answer 15

You can't have this discussion without bringing up type systems.

The main features of functional programming include functions as first-class values, currying, immutable values, etc. It doesn't seem obvious to me that OO design patterns are approximating any of those features.

That's because these features don't address the same issues that OOP does... they are alternatives to imperative programming. The FP answer to OOP lies in the type systems of ML and Haskell... specifically sum types, abstract data types, ML modules, and Haskell typeclasses.

But of course there are still design patterns which are not solved by FP languages. What is the FP equivalent of a singleton? (Disregarding for a moment that singletons are generally a terrible pattern to use)

The first thing typeclasses do is eliminate the need for singletons.

You could go through the list of 23 and eliminate more, but I don't have time to right now.

Answer 16

I think only two GoF Design Patterns are designed to introduce the functional programming logic into natural OO language. I think about Strategy and Command. Some of the other GoF design patterns can be modified by functional programming to simplify the design and keep the purpose.

Answer 17

Essentially, yes !

• When a pattern circumvents the missing features (high order functions, stream handling...) that ultimalty facilitate composition .
• The need to re-write patterns' implementation again and again can itself be seen as a language smell .

Besides, thispage (AreDesignPatternsMissingLanguageFeatures) provides a "pattern/feature" translation table and some nice discussions, if you are willing to dig.

Answer 18

In the new 2013 book named "Functional Programming Patterns- in Scala and Clojure" the author Michael.B. Linn does a decent job comparing and providing replacements in many cases for the GoF patterns and also discusses the newer functional patterns like 'tail recursion', 'memoization', 'lazy sequence', etc.

This book is available on Amazon. I found it very informative and encouraging when coming from an OO background of a couple of decades.

Answer 19

Functional programming does not replace design patterns. Design patterns can not be replaced.

Patterns simply exist; they emerged over time. The GoF book formalized some of them. If new patterns are coming to light as developers use functional programming languages that is exciting stuff, and perhaps there will be books written about them as well.

Answer 20

OOP and the GoF patterns deal with states. OOP models reality to keep the code base as near as possible to the given requirements of reality. GoF design patterns are patterns that were identified to solve atomic real world problems. They handle the problem of state in a semantic way.

As in real functional programming no state exists, it does not make sense to apply the GoF patterns. There are not functional design patterns in the same way there are GoF design patterns. Every functional design pattern is artifical in contrast to reality as functions are constructs of math and not reality.

Functions lack the concept of time as they are always return the same value whatever the current time is unless time is part of the function parameters what makes it really hard to prrocess "future requests". Hybrid languages mix those concepts make the languages not real functional programming languages.

Functional languages are rising only because of one thing: the current natural restrictions of physics. Todays processors are limited in their speed of processing instructions due to physical laws. You see a stagnation in clock frequency but an expansion in processing cores. That's why parallelism of instructions becomes more and more important to increase speed of modern applications. As functional programming by definition has no state and therefore has no side effects it is safe to process functions safely in parallel.

GoF patterns are not obsolete. They are at least necessary to model the real world requirements. But if you use a functional programming language you have to transform them into their hybrid equivalents. Finally you have no chance to make only functional programs if you use persistence. For the hybrid elements of your program there remains the necessity to use GoF patterns. For any other element that is purely functional there is no necessity to use GoF patterns because there is no state.

Because the GoF patterns are not necessary for real functional programming, it doesn't mean that the SOLID principles should not be applied. The SOLID principles are beyond any language paradigm.

Answer 21

As the accepted answer said, OOP and FP all have their specific patterns.

However, there are some patterns which are so common that all programming platforms I can think of should have. Here is an (incomplete) list:

• Adapter. I can hardly think of a useful programming platform which is so comprehensive (and self-fulfilled) that it does not need to talk to the world. If it is going to do so, an adapter is definitely needed.

• Façade. Any programming platforms that can handle big source code should be able to modularise. If you were to create a module for other parts of the program, you will want to hide the "dirty" parts of the code and give it a nice interface.

• Interpreter. In general, any program is just doing two things: parse input and print output. Mouse inputs need to be parsed, and window widgets need to be printed out. Therefore, having an embedded interpreter gives the program additional power to customise things.

Also, I noticed in a typical FP language, Haskell, there is something similar to GoF patterns, but with different names. In my opinion this suggest they were there because there are some common problems to solve in both FP and OOP languages.

• Monad transformer and decorator. The former used to add additional ability into an existing monad, the latter add additional ability to an existing object.

Answer 22

I think that each paradigm serves a different purpose and as such cannot be compared in this way.

I have not heard that the GoF design patterns are applicable to every language. I have heard that they are applicable to all OOP languages . If you use functional programming then the domain of problems that you solve is different from OO languages.

I wouldn't use functional language to write a user interface, but one of the OO languages like C# or Java would make this job easier. If I were writing a functional language then I wouldn't consider using OO design patterns.

Answer 23

OOP and FP have different goals. OOP aims to encapsulate the complexities/moving parts of software components and FP aims to minimize the complexity and dependencies of software components.

However these two paradigms are not necessarily 100% contradicting and could be applied together to get the benefit from both worlds.

Even with a language that does not natively support functional programming like C#, you could write functional code if you understand the FP principles. Likewise you could apply OOP principles using F# if you understand OOP principles, patterns, and best practices. You would make the right choice based on the situation and problem that you try to solve, regardless of the programming language you use.

Answer 24

Some patterns are easier to implement in a language supporting FP. For example, Strategy can be implemented using nicely using closures. However depending on context, you may prefer to implement Strategy using a class-based approach, say where the strategies themselves are quite complicated and/or share structure that you want to model using Template Method.

In my experience developing in a multi-paradigm language (Ruby), the FP implementation works well in simple cases, but where the context is more complicated the GoF OOP based approach is a better fit.

The FP approach does not replace the OOP approach, it complements it.

Answer 25

It does, in that a high-level functional PL (like OCaml, with classes, modules, etc.) certainly supersedes OOP imperative languages in type versatility and power of expression. The abstractions do not leak, you can express most of your ideas directly in the program. Therefore, yes, it does replace design patterns, most of which are ridiculously simplistic compared to functional patterns anyhow.

Answer 26

In functional programming, design patterns have a different meaning. In fact, most of OOP design patterns are unnecessary in functional programming because of the higher level of abstraction and HOFs used as building blocks.

The principle of an HOF means that functions can be passed as arguments to other functions. and functions can return values.

Answer 27

The paramount characteristic of functional programming, IMHO, is that you are programming with nothing but expressions -- expressions within expressions within expressions that all evaluate to the last, final expression that "warms the machine when evaluated".

The paramount characteristic of object-oriented programming, IMHO is that you are programming with objects that have internal state. You cannot have internal state in pure functions -- object-oriented programming languages need statements to make things happen. (There are no statements in functional programming.)

You are comparing apples to oranges. The patterns of object-oriented programming do not apply to function programming, because functional programming is programming with expressions, and object-oriented programming is programming with internal state.

Answer 28

Brace yourselves.

It will aggravate many to hear me claim to have replaced design patterns and debunked SOLID and DRY. I'm nobody. Nevertheless, I correctly modeled collaborative (manufacturing) architecture and published the rules for building processes online along with the code and science behind it at my website http://www.powersemantics.com/ .

My argument is that design patterns attempt to achieve what manufacturing calls "mass customization", a process form in which every step can be reshaped, recomposed and extended. You might think of such processes as uncompiled scripts. I'm not going to repeat my (online) argument here. In short, my mass customization architecture replaces design patterns by achieving that flexibility without any of the messy semantics. I was surprised my model worked so well, but the way programmers write code simply doesn't hold a candle to how manufacturing organizes collaborative work.

• Manufacturing = each step interacts with one product
• OOP = each step interacts with itself and other modules, passing the product around from point to point like useless office workers

This architecture never needs refactoring. There are also rules concerning centralization and distribution which affect complexity. But to answer your question, functional programming is another set of processing semantics, not an architecture for mass custom processes where 1) the source routing exists as a (script) document which the wielder can rewrite before firing and 2) modules can be easily and dynamically added or removed.

We could say OOP is the "hardcoded process" paradigm and that design patterns are ways to avoid that paradigm. But that's what mass customization is all about. Design patterns embody dynamic processes as messy hardcode. There's just no point. The fact that F# allows passing functions as a parameter means functional and OOP languages alike attempt to accomplish mass customization itself.

How confusing is that to the reader, hardcode which represents script? Not at all if you think your compiler's consumers pay for such features, but to me such features are semantic waste. They are pointless, because the point of mass customization is to make processes themselves dynamic , not just dynamic to the programmer wielding Visual Studio.

Answer 29

Let give an example of the wrong premise you state.

The adapter pattern we have in OOP as usecase adapter such as in cleanarch and ddd can be implemented in Functional via the monad variation of Option.

You are not replacing them but transforming them.