zoukankan      html  css  js  c++  java
  • [Functional Programming] From simple implementation to Currying to Partial Application

    Let's say we want to write a most simple implementation 'avg' function:

    const avg = list => {
        let sum = 0;
        for(let i = 0; i < list.length; i++) {
            sum += list[i]
        }
        return sum / list.length
    }

    Basiclly, the 'avg' function doing two things:

    • Calculate sum
    • Divide sum / length

    It works fine for tiny / small application, but for the large application, we need to think about reuseablitiy. We want to breakdown one function and think about any reuseable partten, which later can be reused.

    In the following examples, We want to bring in two libarays which are commonly used in FP. One is Ramda, another one is Crocks.

    Currying:

    First, we want to write 'sum' and 'devide' functions by ourselves:

    const { curry, reduce, compose } = require("crocks");
    const R = require("ramda");
    
    const sum = reduce(R.add, 0);
    // divideByLen :: [Number] -> Number -> Number
    const divideByLen = curry(
      compose(
        R.flip(R.divide),
        R.length
      )
    );

    'sum' is simple, using 'reduce' from Crocks, you can also write JS reduce, doesn't matter.

    What we need to explain is 'divideByLen' function. 

    • Why 'curry'?

    Basic we want to call divideByLen in two ways:

    divideByLen([1,2,3], sum([1,2,3]))
    divideByLen([1,2,3])(sum([1,2,3]))

    [Notice] You need to bring in 'curry' from Crocks, it is more flexable. 

    • Why 'flip'?

    Because R.divide(sum, length), we need to feed the divide function with sum as first argement, then length as second arguement. But when we write code, length will be feeded frist, sum will be partially applied, it will come second, therefore we need to call 'flip'.

    Bring all together:

    const avg = list =>
      compose(
        divideByLen(list),
        sum
      )(list);

    We notice that, we have to pass 'list' to both Sum(list) and divideByLen(list). The code looks not so good. Whenever you are facing the situation, you need to pass the same arguement to two functions in parallel. You can consider to using 'Partial Application'.

    Partial Application:

    // Ramda

    const avg = R.converge(R.divide, [R.sum, R.length]);

    We are using 'Ramda's converge' function, bascilly you have pass in a data, the data will be passed to R.sum(data) & R.length(data), the return results of those two functions, will be passed to R.divide(resOfSum, resOfLength). 

    //Crocks:

    const { curry, fanout, merge, compose } = require("crocks");
    
    const avg = compose(
      merge(R.divide),
      fanout(R.sum, R.length)
    );

    We are using the Pair ADT, the data will be passed to R.sum(data) & R.length(data) thought 'fanout' function, it returns Pair(resOfSum, resOfLength).

    Then we use 'merge', it works with Pair ADT, we merge two results by R.divide(resOfSum, resOfLength).

  • 相关阅读:
    ESXi创建磁盘命令
    TNS-12518,TNS-12536,TNS-00506,Linux Error: 11: Resource temporarily unavailable
    监听的instance status blocked分析
    Oracle 用户、对象权限、系统权限
    MIME详解
    11g等待事件之library cache: mutex X
    Latch Free
    PowerDesigner小技巧
    yum本地源配置
    内核参数SEMMSL SEMMNS SEMOPM SEMMNI参数的设置
  • 原文地址:https://www.cnblogs.com/Answer1215/p/10753304.html
Copyright © 2011-2022 走看看