在 Postgres 中使用 COUNT 和聚合

Question

I am running the below query to retrieve my data:

SELECT
    e.id AS id, e.name AS name, e.description AS description,
    c.slug AS category,
    COUNT(t.id) AS sold,
    json_agg(json_build_object('id', b.id, 'title', b.title, 'description', b.description,'price', b.price, 'available', b.qty_available, 'qty_per_sale', b.qty_per_sale))::JSONB AS book,
    json_agg(json_build_object('id', s.id, 'startDate', s.start_date, 'endDate', s.end_date,'daysAhead', (s.start_date::DATE - NOW()::DATE), 'times', s.times))::JSONB as dates
FROM event e
    LEFT JOIN books b ON b.event_id = e.id
    LEFT JOIN shows s ON s.event_id = e.id
    LEFT JOIN category c ON e.category_id = c.id
    LEFT JOIN ticket t ON t.book_id = b.id
WHERE
    (status = 'PUBLISHED' OR status = 'PROMOTED')
    AND s.end_date >= DATE(NOW())
    AND e.is_private = FALSE
    AND s.id = t.show_id
    AND t.canceled = FALSE
GROUP BY  e.id,c.slug
ORDER BY  sold
LIMIT  30

This query works fine. I get event-level aggregate sold .

TABLE ticket (
    id SERIAL PRIMARY KEY,
    book_id SERIAL REFERENCES books(id),
    order_id SERIAL REFERENCES purchases(id),
    show_id SERIAL REFERENCES shows(id),
    showtime character varying(10) NOT NULL,
    canceled boolean DEFAULT false
);

I want to add another sold property to this report, inside dates. Currently, dates entry looks like:

[
  {
    "id": 46,
    "times": [
      {
        "end": "13:00",
        "start": "12:00"
      }
    ],
    "endDate": "2022-02-27",
    "daysAhead": 308,
    "startDate": "2022-02-27"
  },
  {
    "id": 46,
    "times": [
      {
        "end": "13:00",
        "start": "12:00"
      }
    ],
    "endDate": "2022-02-27",
    "daysAhead": 308,
    "startDate": "2022-02-27"
  }
]

I want to aggregate on ticket showtime being equal to time.start , so each times element would look have this new sold property in addition to start and end.

I tried using Postgres WITH to turn the current query into a set, then JOIN the ticket table on it again, without much success so far.

Basically, I wish to just take times apart and add a property to it like COUNT(t.id) per event, only filtered per show.

However, nested aggregations are not allowed, and COUNT() is one.

Any suggestions?

Answer 1

Rather than turn the current query into a with statement and joining to tickets again, it may be easier to go the other way, ie use the ticket table in a with clause and join that to the current query. I'm not sure from your description exactly what you want the output to look like, but perhaps something like:

WITH ticket_summary as(
   select book_id, count(1) as ticket_count
   from ticket
   group by book_id
)
SELECT
    e.id AS id, e.name AS name, e.description AS description,
    c.slug AS category,
    COUNT(t.id) AS sold,
    json_agg(json_build_object('id', b.id, 'title', b.title, 'description', b.description,'price', b.price, 'available', b.qty_available, 'qty_per_sale', b.qty_per_sale, 'sales', ts.ticket_count))::JSONB AS book, --can now use in aggregate
    json_agg(json_build_object('id', s.id, 'startDate', s.start_date, 'endDate', s.end_date,'daysAhead', (s.start_date::DATE - NOW()::DATE), 'times', s.times))::JSONB as dates
FROM event e
    LEFT JOIN books b ON b.event_id = e.id
    LEFT JOIN shows s ON s.event_id = e.id
    LEFT JOIN category c ON e.category_id = c.id
    LEFT JOIN ticket t ON t.book_id = b.id
    --pull in summary data
    LEFT JOIN ticket_summary ts ON ts.book_id = b.id
WHERE
    (status = 'PUBLISHED' OR status = 'PROMOTED')
    AND s.end_date >= DATE(NOW())
    AND e.is_private = FALSE
    AND s.id = t.show_id
    AND t.canceled = FALSE
GROUP BY  e.id,c.slug
ORDER BY  sold
LIMIT  30