Java 8 Streams多个分组依据

Posted 2021-05-04

我的温度记录是这样的

dt        |AverageTemperature |AverageTemperatureUncertainty|City   |Country |Latitude|Longitude
----------+-------------------+-----------------------------+-------+--------+--------+---------
1963-01-01|-5.417000000000002 |0.5                          |Karachi|Pakistan|57.05N  |10.33E  
1963-02-01|-4.7650000000000015|0.328                        |Karachi|Pakistan|57.05N  |10.33E  
1964-01-01|-5.417000000000002 |0.5                          |Karachi|Pakistan|57.05N  |10.33E  
1964-02-01|-4.7650000000000015|0.328                        |Karachi|Pakistan|57.05N  |10.33E  
1965-01-01|11.417000000000002 |0.5                          |Karachi|Pakistan|57.05N  |10.33E 
1965-02-01|12.7650000000000015|0.328                        |Karachi|Pakistan|57.05N  |10.33E

我必须将其解析为POJO并根据以下问题陈述计算平均增量：

使用Streams API计算每个国家/地区的年平均温度增量。为了计算delta，1900年的平均温度将从1901年的平均温度中减去，以获得特定城市1900年至1901年的增量。所有这些增量的平均值是一个城市的年平均温度增量。一个国家所有城市的平均值是一个国家的平均值。

我的温和POJO看起来像是有吸气剂和二传手

public class Temperature {
    private java.util.Date date;
    private double averageTemperature;
    private double averageTemperatureUncertainty;
    private String city;
    private String country;
    private String latitude;
    private String longitude;
}

我已经保留了一份温度列表，因为要使用流来解决这个问题。

要计算delta我试图使用以下流但我仍然无法计算实际的delta，因为我必须计算平均国家/地区delta，我已经对国家/地区，城市和日期进行了分组。

Map<String, Map<String, Map<Integer, Double>>> countriesMap = this.getTemperatures().stream()
                .sorted(Comparator.comparing(Temperature::getDate))
                .collect(Collectors.groupingBy(Temperature::getCountry,
                        Collectors.groupingBy(Temperature::getCity,
                        Collectors.groupingBy
                                (t -> {
                                            Calendar calendar = Calendar.getInstance();
                                            calendar.setTime(t.getDate());
                                            return calendar.get(Calendar.YEAR);
                                        }, 
                        Collectors.averagingDouble(Temperature::getAverageTemperature)))));

为了计算delta，我们必须计算Map<Integer, Double>的差异。

为了计算差异，我提出了以下代码，但无法将以下代码与上面的代码连接起来

Stream.of(10d, 20d, 10d) //this is sample data that I that I get in `Map<Integer, Double>` of countriesMap
        .map(new Function<Double, Optional<Double>>() {
            Optional<Double> previousValue = Optional.empty();
            @Override
            public Optional<Double> apply(Double current) {
                Optional<Double> value = previousValue.map(previous -> current - previous);
                previousValue = Optional.of(current);
                return value;
            }
        })
        .filter(Optional::isPresent)
        .map(Optional::get)
        .forEach(System.out::println);

如何一次性使用流计算delta或如何在countriesMap上执行流操作以计算delta和实现上述问题的声明。

答案

为了将问题陈述减少到一个较小的块，你可以研究的另一种方法是解析yearly温度并计算它们的delta，进一步averageing它。对于你问题中内部Map<Integer, Double>中Map类型的所有值，都必须这样做。它看起来像：

Map<Integer, Double> unitOfWork = new HashMap<>(); // innermost map you've attained ('yearToAverageTemperature' map)
unitOfWork = unitOfWork.entrySet()
        .stream()
        .sorted(Map.Entry.comparingByKey())
        .collect(Collectors.toMap(Map.Entry::getKey, Map.Entry::getValue, (e1, e2) -> e1, LinkedHashMap::new));
// the values sorted based on the year from a sorted map
List<Double> srtedValPerYear = new ArrayList<>(unitOfWork.values());
// average of deltas from the complete list 
double avg = IntStream.range(0, srtedVal.size() - 1)
        .mapToDouble(i -> (srtedVal.get(i + 1) - srtedVal.get(i)))
        .average().orElse(Double.NaN);

进一步注意，这只是一个City的<Year, AverageTemperature>记录的平均值，你必须遍历所有City键集，并且类似地为你所有的Country键集详尽地找出这样的平均值。

进一步将这个工作单元移动到一个方法中，迭代完整的地图图，这可以完成为：

// The average of all cities in a country is the average of a country.
AtomicReference<Double> countryValAvg = new AtomicReference<>(0.0);
countriesMap.forEach((country, cityMap) -> {
    // The average of all these deltas is the average annual temperature delta for a city.
    AtomicReference<Double> cityAvgTemp = new AtomicReference<>((double) 0);
    cityMap.forEach((city, yearMap) -> cityAvgTemp.set(cityAvgTemp.get() + averagePerCity(yearMap)));
    double avgAnnualTempDeltaPerCity = cityAvgTemp.get() / cityMap.size();

    countryValAvg.set(countryValAvg.get() + avgAnnualTempDeltaPerCity);
});
System.out.println(countryValAvg.get() / countriesMap.size());

其中averagePerCity是遵循的方法

double averagePerCity(Map<Integer, Double> unitOfWork) {
    unitOfWork = unitOfWork.entrySet()
            .stream()
            .sorted(Map.Entry.comparingByKey())
            .collect(Collectors.toMap(Map.Entry::getKey, Map.Entry::getValue, (e1, e2) -> e1, LinkedHashMap::new));
    List<Double> srtedVal = new ArrayList<>(unitOfWork.values());
    return IntStream.range(0, srtedVal.size() - 1)
            .mapToDouble(i -> (srtedVal.get(i + 1) - srtedVal.get(i)))
            .average().orElse(Double.NaN);
}

注意：上面的代码可能缺少验证，它只是为了提供一个概念，即如何将完整的问题分解成更小的部分然后解决。

编辑1：哪个could be improved further as：

// The average of all cities in a country is the average of a country.
AtomicReference<Double> countryValAvg = new AtomicReference<>(0.0);
countriesMap.forEach((country, cityMap) -> {
    // The average of all these deltas is the average annual temperature delta for a city.
    double avgAnnualTempDeltaPerCity = cityMap.values()
            .stream()
            .mapToDouble(Quick::averagePerCity) // Quick is my class name
            .average()
            .orElse(Double.NaN);
    countryValAvg.set(countryValAvg.get() + avgAnnualTempDeltaPerCity);
});
System.out.println(countryValAvg.get() / countriesMap.size());

编辑2：进一步

double avgAnnualTempDeltaPerCity = countriesMap.values().stream()
        .mapToDouble(cityMap -> cityMap.values()
                .stream()
                .mapToDouble(Quick::averagePerCity) // Quick is my class name
                .average()
                .orElse(Double.NaN))
        .average().orElse(Double.NaN);