Combine selected data into same row based on column id (order)

Question

I have a combination of tables that are used to described location flows of people. The main 'travel_flows' tables has a structure as seen in the first table below. The order of the flows is described by the 'order_id' column.

+-----------+------------+---------+-------------+----------+
| travel_id | purpose_id | name_id | location_id | order_id |
+-----------+------------+---------+-------------+----------+
|       434 |         23 |      55 |          85 |        1 |
|       212 |         43 |      55 |          45 |        2 |
|       411 |         41 |      55 |          17 |        3 |
|       148 |         23 |      32 |          32 |        1 |
|       153 |         11 |      32 |          19 |        2 |
+-----------+------------+---------+-------------+----------+

What I am trying to achieve using PostgreSQL 9.6 is to group the returned rows by 'name_id' in a binary origin(from)/destination(to) format based on 'location_id' and 'order_id' value, similar to the following table:

+-----------------+--------------+------------------+---------------+------------+----------------+
| from_purpose_id | from_name_id | from_location_id | to_purpose_id | to_name_id | to_location_id |
+-----------------+--------------+------------------+---------------+------------+----------------+
|              23 |           55 |               85 |            43 |         55 |             45 |
|              43 |           55 |               45 |            41 |         55 |             17 |
|              23 |           32 |               32 |            11 |         32 |             19 |
+-----------------+--------------+------------------+---------------+------------+----------------+

Is there any way to achieve this with a select statement?

Answer 1

You can do this with lead window function.

select * from (
select purpose_id,name_id,location_id,
lead(purpose_id) over(partition by name_id order by order_id) as to_purpose_id,
lead(name_id) over(partition by name_id order by order_id) as to_name_id,
lead(location_id) over(partition by name_id order by order_id) as to_location_id
from tbl
) t
where to_purpose_id is not null and to_name_id is not null and to_location_id is not null

Answer 2

You can easily achieve this with Window Functions (see https://www.postgresql.org/docs/current/static/tutorial-window.html ), specifically, function lead() :

with flow as (
  select
    name_id,
    purpose_id as from_purpose_id,
    lead(purpose_id) over w1 as next_purpose_id,
    location_id as from_location_id,
    lead(location_id) over w1 as next_location_id,
    order_id
  from travel_flows
  window w1 as (partition by name_id order by order_id)
)
select
  name_id, from_purpose_id, next_purpose_id, from_location_id, next_location_id
from flow
where next_purpose_id is not null
order by name_id, order_id
;