检查 lon , lat 对数组中的重大变化答案

【问题标题】：Check for big changes in an array of lon , lat pair检查 lon , lat 对数组中的重大变化
【发布时间】：2021-08-10 18:02:51
【问题描述】：

我是 sql 新手，想看看是否有可能在 lon 和 lat 的数组中找到差异。如何检查从当前位置到下一个位置的值是否有很大变化例子： [-79.88523, 30.2566][-65.5523,25.256] 使用 BigQuery 并且仅限于选择

select geo from ( 
SELECT 
routeID ,  json_extract(st_asgeojson(st_makeline( 
array_agg(st_geogpoint(locs.lon, locs.lat) order by 
locs.date))),'$.coordinates') as geo,
FROM 
howardcounty.routebatches
cross join UNNEST(locations) as locs
where  locs.date between {{start_date}} and {{end_date}} 
group by routeID 
order by routeID
limit 1000
)where length(geo)-length(replace(geo,"]","")) > 1+4

【问题讨论】：

标签： google-bigquery gis

【解决方案1】：

使用lag 访问以前的数据行。但首先必须将字符串解析为浮点数：

select *,lon-lon_old as diff_lon, lat-lat_old as diff_lat
#RouteID , count(1)
from (
select *, 
    lag(lon) over (partition by RouteID order by row) as lon_old,
    lag(lat) over (partition by RouteID order by row) as lat_old
from (
select 
    row_number() over () as row ,RouteID,
    cast(split(data,",")[offset(0)] as float64) as lon, 
    cast(replace(split(data,",")[offset(1)],"]","") as float64) as lat
from(
select RouteID, split(substring(geo,3),"],[") as data from ( 
SELECT 
routeID ,  json_extract(st_asgeojson(st_makeline( 
array_agg(st_geogpoint(locs.lon, locs.lat) order by 
locs.date))),'$.coordinates') as geo,
FROM 

(select 1 as RouteID,  [struct(5 as lon,5 as lat, current_date() as date), struct(1,2,"2020-01-01") ] as locations UNION ALL 
select 2 as RouteID,  [struct(15 as lon,15 as lat, current_date() as date), struct(1,2,"2020-01-01") ]
)


cross join UNNEST(locations) as locs
#where  locs.date between {{start_date}} and {{end_date}} 
group by routeID 
order by routeID
limit 1000
)where length(geo)-length(replace(geo,"]","")) > 1+4
), unnest(data) data  
)) 
#where lon-lon_old>0.5 or lat-lat_old>0.5
#group by 1

一个更简单的例子是

With dat as  (select [struct(-79.88523 as A, 30.2566 as B), struct(-65.5523,25.256)] as data )

select *,value.A-old.A as diff_A
from (
select *, lag(value) over (order by row) as old
from (
select row_number() over () as row, value
from dat, unnest (dat.data) as value
)
)

【讨论】：

感谢您的回复，我不能使用 with 并且仅限于 SELECT 仅在数据集上，如何仅使用 select 重写 bigquery 的代码？
感谢分配，成功了。然而，这仅显示来自一条路线的一组坐标，我如何制作它以便它对所有其他 routeID 执行相同的操作？我还需要集成一个 if 语句来过滤掉任何比 0.5 有更大变化的东西
where 需要过滤而不是 if。在第一行注释* 并删除其他 cmets 行。该脚本适用于多个 routeID。