Functional programming leads to deep insights into the nature of computation. -- Martin Odersky
形式化
FizzBuzzWhizz詳細描述請自行查閱相關(guān)資料。此處以3, 5, 7為例，形式化地描述一下問題。
r1
- times(3) -> Fizz
- times(5) -> Buzz
- times(7) -> Whizz
r2
- times(3) && times(5) && times(7) -> FizzBuzzWhizz
- times(3) && times(5) -> FizzBuzz
- times(3) && times(7) -> FizzWhizz
- times(5) && times(7) -> BuzzWhizz
r3
- contains(3) -> Fizz
- the priority of contains(3) is highest
rd
- others -> others
接下來我將使用Scala嘗試FizzBuzzWhizz問題的設(shè)計和實現(xiàn)。
語義模型
從上面的形式化描述，可以很容易地得到FizzBuzzWhizz問題的語義模型。
Rule: (Int) -> String
Matcher: (Int) -> Boolean
Action: (Int) -> String
其中，Rule存在三種基本的類型：
Rule ::= atom | allof | anyof
三者之間構(gòu)成了「樹型」結(jié)構(gòu)。
atom: (Matcher, Action) -> String
allof: rule1 && rule2 ...
anyof: rule1 || rule2 ...
測試用例
借助Scala強大的「類型系統(tǒng)」能力，可拋棄掉很多重復(fù)的「樣板代碼」，使得設(shè)計更加簡單、漂亮。此外，Scala構(gòu)造DSL的能力也相當(dāng)值得稱贊，非常直接，簡單。
import org.scalatest._
import prop._
class RuleSpec extends PropSpec with TableDrivenPropertyChecks with Matchers {
val spec = {
val r1_3 = atom(times(3), to("Fizz"))
val r1_5 = atom(times(5), to("Buzz"))
val r1_7 = atom(times(7), to("Whizz"))
val r1 = anyof(r1_3, r1_5, r1_7)
val r2 = anyof(
allof(r1_3, r1_5, r1_7),
allof(r1_3, r1_5),
allof(r1_3, r1_7),
allof(r1_5, r1_7))
val r3 = atom(contains(3), to("Fizz"))
val rd = atom(always(true), nop);
anyof(r3, r2, r1, rd)
}
val specs = Table(
("n", "expect"),
(3, "Fizz"),
(5, "Buzz"),
(7, "Whizz"),
(3 * 5, "FizzBuzz"),
(3 * 7, "FizzWhizz"),
((5 * 7) * 2, "BuzzWhizz"),
(3 * 5 * 7, "FizzBuzzWhizz"),
(13, "Fizz"),
(35, "Fizz"), // 35 > 5*7
(2, "2")
)
property("fizz buzz whizz") {
forAll(specs) { spec(_) should be (_) }
}
}
匹配器：Matcher
Matcher是一個「一元函數(shù)」，入?yún)?code>Int，返回值為Boolean，是一種典型的「謂詞」。從OO的角度看，always是一種典型的Null Object。
object Matchers {
type Matcher = Int => Boolean
def times(n: Int): Matcher = _ % n == 0
def contains(n: Int): Matcher = _.toString.contains(n.toString)
def always(bool: Boolean): Matcher = _ => bool
}
執(zhí)行器：Action
Action也是一個「一元函數(shù)」，入?yún)?code>Int，返回值為String，其本質(zhì)就是定制常見的map操作，將定義域映射到值域。
object Actions {
type Action = Int => String
def to(str: String): Action = _ => str
def nop: Action = _.toString
}
規(guī)則：Rule
Composition Everywhere
Rule是FizzBuzzWhizz最核心的抽象，也是設(shè)計的靈魂所在。從語義上Rule分為2種基本類型，并且兩者之間形成了優(yōu)美的、隱式的「樹型」結(jié)構(gòu)，體現(xiàn)了「組合式設(shè)計」的強大威力。
AtomicCompositions: anyof, allof
Rule是一個「一元函數(shù)」，入?yún)?code>Int，返回值為String。其中，def atom(matcher: => Matcher, action: => Action)的入?yún)⑹褂?code>Pass By Name的惰性求值的特性。
object Rules {
type Rule = (Int) => String
def atom(matcher: => Matcher, action: => Action): Rule =
n => if (matcher(n)) action(n) else ""
def anyof(rules: Rule*): Rule =
n => rules.map(_(n))
.filterNot(_.isEmpty)
.headOption
.getOrElse("")
def allof(rules: Rule*): Rule =
n => rules.foldLeft("") { _ + _(n) }
}
源代碼
- Github: https://github.com/horance-liu/fizz-buzz-whizz
- C++11參考實現(xiàn)： https://codingstyle.cn/topics/97
- Java參考實現(xiàn)： https://codingstyle.cn/topics/100
- Ruby參考實現(xiàn)： https://codingstyle.cn/topics/114
