如何获得包含某些值的重复字段的频率？

Question

假设我有一个像这样的数据集

{"id":15,"classification":"goth","categories":["blackLipstick","hotTopic"]}
{"id":14,"classification":"goth","categories":["drinking","girls","hotTopic"]}
{"id":13,"classification":"jock","categories":["basketball","chicharones","fooball","girls","pop","pregnant","sports","starTrek","tortilla","tostada"]}
{"id":12,"classification":"geek","categories":["academics","cacahuates","computers","glasses","papas","physics","programming","ps4","science"]}
{"id":11,"classification":"geek","categories":["cacahuates","fajitas","math","pregnant","raves","xbox"]}
{"id":10,"classification":"goth","categories":["cutting"]}
{"id":9,"classification":"geek","categories":["cafe","chalupa","chimichangas","manson","physics","pollo","tostada"]}
{"id":8,"classification":"jock","categories":["basketball","chalupa","enchurrito","piercings","running","sports"]}
{"id":7,"classification":"geek","categories":["aguacate","blackLipstick","computers","fajitas","fooball","glasses","lifting","outdoors","physics","pollo","pregnant","ps4"]}
{"id":6,"classification":"none","categories":["brocode","girls","raves","tacos"]}
{"id":5,"classification":"goth","categories":["blackLipstick","blackShirts","drugs","mole","piercings","tattoos","tortilla"]}
{"id":4,"classification":"jock","categories":["girls","tattoos"]}
{"id":3,"classification":"goth","categories":["girls"]}
{"id":2,"classification":"none","categories":["cutting","enchurrito","fooball","pastel","pregnant","tattoos","vampires"]}
{"id":1,"classification":"goth","categories":["cacahuates","cutting","drugs","empanadas","frijoles","manson","nachos","outdoors","piercings","tattoos"]}
{"id":0,"classification":"geek","categories":["pollo","pop","programming","science"]}

我如何写一个查询，在哪里我可以说：“如果某人有类别‘数学’，他们通常有哪些其他类别？”

对于这个数据集，我可以写一些类似这样的东西来告诉我什么是哥特人、极客和骑手最喜欢的。

SELECT classification, categories, count(categories) C 
FROM [xx.stereotypes] group by classification
, categories ORDER BY C DESC LIMIT 1000

但是在我的真实数据集中，我没有分类字段。我想要一个查询，它可以帮助我创建像"goth“、"jock”或"geek“这样的分类。

例如，如何选择类别中包含“数学”的类别的所有计数，这只能选择数学。

SELECT categories, count(categories) C FROM [xx.stereotypes]
where categories CONTAINS "math" group by categories ORDER
BY C DESC LIMIT 1000

Answer 1

如何选择类别中包含“数学”的类别的所有计数？

SELECT categories, COUNT(1) AS weight
FROM [xx.stereotypes]
OMIT RECORD IF NOT SOME(categories = 'math')
GROUP BY categories
ORDER BY weight DESC

我如何写一个查询，在哪里我可以说：“如果某人有类别‘数学’，他们通常有哪些其他类别？”

SELECT category, related_category, weight 
FROM (
  SELECT category, related_category, COUNT(1) AS weight 
  FROM (
    SELECT a.id AS id, a.categories AS category, b.categories AS related_category 
    FROM (FLATTEN([xx.stereotypes], categories)) AS a
    JOIN (FLATTEN([xx.stereotypes], categories)) AS b
    ON a.id = b.id
    HAVING category != related_category
  )
  GROUP BY category, related_category
) 
WHERE category = 'math'
ORDER BY category, weight DESC, related_category

我想要一个可以帮助我创建分类的查询

下面是为每个id分配分类的简化方法

SELECT id, category AS classification 
FROM (
  SELECT 
    x.id AS id, y.category AS category, SUM(weight) AS rate,
    ROW_NUMBER() OVER(PARTITION BY id ORDER BY rate DESC) AS pos
  FROM (FLATTEN(xx.stereotypes, categories)) AS x
  JOIN (
    SELECT category, related_category, COUNT(1) AS weight 
    FROM (
      SELECT a.id AS id, a.categories AS category, b.categories AS related_category 
      FROM (FLATTEN([xx.stereotypes], categories)) AS a
      JOIN (FLATTEN([xx.stereotypes], categories)) AS b
      ON a.id = b.id
    )
    GROUP BY category, related_category 
  ) AS y
  ON x.categories = y.related_category
  GROUP BY 1, 2 
)
WHERE pos = 1
ORDER BY id DESC