总计计数跨越多天,逐个小时

时间:2016-12-18 18:53:52

标签: javascript node.js postgresql group-by knex.js

我正在尝试使用Knex在多天内按小时对记录进行分组。因此,例如,9AM将是:

{
  hour: 9AM, // for days 12/11, 12/12, 12/13
  count: 10 // 10 Ids total over those days for hour 9AM
}

鉴于此快照中的记录,如何在多天内将它们汇总到hour个桶中?

enter image description here

如果我输出查询结果,您可以看到19:0012/12的{​​{1}}的两个单独结果。这两天'需要将计数总计为一个12/13分组:

hour 19:00

我当前的查询:

ROWS [ anonymous {
    session_ids: [ 3200 ],
    hour: 2016-12-12T14:00:00.000Z,
    count: '1' },
  anonymous {
    session_ids: [ 3201 ],
    hour: 2016-12-12T15:00:00.000Z,
    count: '1' },
  anonymous {
    session_ids: [ 3203, 3202 ],
    hour: 2016-12-12T19:00:00.000Z,
    count: '2' },
  anonymous {
    session_ids: [ 3204, 3205 ],
    hour: 2016-12-13T19:00:00.000Z, // This count should be aggregated into the `19:00` grouping above
    count: '2' } ]

1 个答案:

答案 0 :(得分:4)

使用EXTRACT,而不是date_trunc

var qry = db.knex
  .select(db.knex.raw("array_agg(t2.id) as session_ids, extract('hour' from t2.start_timestamp) as hour"))
  .count('*')
  .from('sessions as t2')
  .groupByRaw("extract('hour' from t2.start_timestamp)")
  .orderBy(db.knex.raw("extract('hour' from t2.start_timestamp)"));

date_trunc将时间戳截断为指定的精度(意味着GROUP BY不起作用,因为具有相同“小时”字段的两个时间戳的天数可能仍然不同):

SELECT date_trunc('hour', NOW());
┌────────────────────────┐
│       date_trunc       │
├────────────────────────┤
│ 2016-12-18 19:00:00+01 │
└────────────────────────┘
(1 row)

EXTRACT获取您要求的特定字段:

SELECT extract('hour' from NOW());
┌───────────┐
│ date_part │
├───────────┤
│        19 │
└───────────┘
(1 row)