如何按ndc_description和year分组计算均值并赋值到原数据数组？

问题背景 #

给定如下数据：

const data = [
  {ndc_description: "VIAGRA 50 MG TABLET", new_percent_change: 12.9, year: "2017"}, 
  {ndc_description: "VIAGRA 50 MG TABLET", new_percent_change: 12.9, year: "2019"}, 
  {ndc_description: "VIAGRA 50 MG TABLET", new_percent_change: 12.9, year: "2017"}, 
  {ndc_description: "JANUVIA 100 MG TABLET", new_percent_change: 4.41, year: "2017"},
  {ndc_description: "JANUVIA 100 MG TABLET", new_percent_change: 4.41, year: "2019"},
  {ndc_description: "JANUVIA 100 MG TABLET", new_percent_change: 4.41, year: "2017"}
];

需求是按ndc_description和year分组，计算每组new_percent_change的均值，并在原数组的每个对象中新增一列存储该均值。

已通过以下代码计算出分组均值：

const roll = d3.rollups(data, v => d3.mean(v, d => d.new_percent_change), d => d.year, d => d.ndc_description);

解决方案 #

d3.rollups返回嵌套数组结构（[year, [ndc_description, mean]]），我们可以先把它转换成便于查询的映射结构，再遍历原数据数组给每个元素添加均值字段。

方法1：转换为嵌套Map #

// 将rollups结果转为嵌套Map：year -> ndc_description -> mean
const meanMap = new Map(roll.map(([year, ndcEntries]) => 
  [year, new Map(ndcEntries)]
));

// 遍历原数据，新增group_mean字段存储均值
data.forEach(d => {
  d.group_mean = meanMap.get(d.year).get(d.ndc_description);
});

方法2：转换为平面对象 #

// 将rollups结果转为平面对象，用`year-ndc_description`作为键
const meanObj = {};
roll.forEach(([year, ndcEntries]) => {
  ndcEntries.forEach(([ndc, mean]) => {
    meanObj[`${year}-${ndc}`] = mean;
  });
});

// 遍历原数据添加字段
data.forEach(d => {
  d.group_mean = meanObj[`${d.year}-${d.ndc_description}`];
});

执行后，原data数组中的每个对象都会新增group_mean字段，对应分组的均值会被赋值进去。比如所有VIAGRA 50 MG TABLET且year=2017的对象，group_mean值为12.9；JANUVIA 100 MG TABLET且year=2019的对象，group_mean值为4.41。

内容的提问来源于stack exchange，提问作者Fateme Nazari