Big Query - 背靠背计算开始和结束日期答案

【问题标题】：Big Query - Calculate start and end date back to backBig Query - 背靠背计算开始和结束日期
【发布时间】：2021-02-27 20:41:57
【问题描述】：

我有一个问题，我需要一些建议，我需要计算大查询中背靠背的休假日历天数。（例如，07-01-2020 到 10-01-2020 和 13-01-2020 到 15-01-2020 的 2 条休假记录，应返回 07-01-2020 到 15-01-2020）

但是，由于该周有公共假期，某些周的休假间隔为 3/4 天。谁能建议一个可能的解决方法？我为公共假期创建了一张表格，但我坚持认为我怎么可能将公共假期的几周视为背靠背。我考虑过窗口函数，但我不确定正确的逻辑是什么。

原始数据集

personnel_number	start_date	end_date	next_start_date	next_end_date	days_between_next_row	remarks
100100	16/1/2020	17/1/2020	20/1/2020	24/1/2020	3
100100	20/1/2020	24/1/2020	28/1/2020	31/1/2020	4	"public holiday on 27-Jan"
100100	28/1/2020	31/1/2020	10/2/2020	13/2/2020	10
100100	10/2/2020	13/2/2020	NULL	NULL

公众假期表

pub_start_date	pub_end_date	remarks
25/1/2020	27/1/2020	"CNY Holiday"

期望的结果

personnel_number	start_date	back_to_back_end_date
100100	16/1/2020	31/1/2020
100100	10/2/2020	13/2/2020

【问题讨论】：

标签： sql date google-bigquery calendar window-functions

【解决方案1】：

以下是 BigQuery 标准 SQL

#standardSQL
with temp as (
  -- all pto days from original table
  select personnel_number, day, '1' type from `project.dataset.table`, 
  unnest(generate_date_array(start_date, end_date)) day
  
  union distinct -- add weekend days if last pto day is friday
  select personnel_number, day, '0' type from `project.dataset.table`, 
  unnest([] || if(extract(dayofweek from end_date) = 6, [end_date + 1, end_date + 2], [])) day
  
  union distinct -- all holiday days from holidays table 
  select personnel_number, day, '0' from (select distinct personnel_number from `project.dataset.table`), 
  (select day from holidays, unnest(generate_date_array(pub_start_date, pub_end_date)) day)
  
  union distinct -- add weekend days to holidays if last day of hliday is friday 
  select personnel_number, day, '0' from (select distinct personnel_number from `project.dataset.table`), 
  (select day from holidays, unnest([] || if(extract(dayofweek from pub_end_date) = 6, [pub_end_date + 1, pub_end_date + 2], [])) day) 
)
select personnel_number,
  start_date + start_tail as start_date,                     -- removing leading non pto days
  back_to_back_end_date - end_tail as back_to_back_end_date  -- removing trailing non pto days
from (
  select personnel_number, 
    min(day) start_date, 
    max(day) back_to_back_end_date, 
    length(regexp_extract(string_agg(type, '' order by day), r'^0*')) start_tail, -- detect number of leading non pto days (holidays or weekend days)
    length(regexp_extract(string_agg(type, '' order by day), r'0*$')) end_tail,   -- detect number of leading non pto days (holidays or weekend days)
    regexp_contains(string_agg(type, '' order by day), r'1') valid
  from (
    select personnel_number, day, type, countif(flag) over(partition by personnel_number order by day) grp
    from (
      select *, day != 1 + ifnull(lag(day) over(partition by personnel_number order by day), day) flag 
      from temp
    )
  )
  group by personnel_number, grp
)
where valid

如果适用于您问题中的样本数据

with `project.dataset.table` as (
  select 100100 personnel_number, date '2020-01-16' start_date, date '2020-01-17' end_date union all
  select 100100, '2020-01-20', '2020-01-24' union all
  select 100100, '2020-01-28', '2020-01-31' union all
  select 100101, '2020-02-10', '2020-02-13'
), holidays as (
  select date '2020-01-25' pub_start_date, date '2020-01-27' pub_end_date, 'CNY Holiday' remarks 
)

输出是

【讨论】：

嗨，是的，解决方案完美运行，感谢您的投入