Meetup Beleza na Web - Funções analíticas com SQL

Funções Analíticas com SQL
1

● 13 anos com desenvolvimento de software
● 2 certificações Java
● Formação DBA Oracle
● Responsável pelo BI da Beleza na Web
● Entusiasta pelo mercado financeiro e economia
● Formação em 2 estilos de Kung Fu
Bruno Paulino
2

● Paginação
● Sumarização
● Cálculos
● Análise posicional entre linhas
● Análise de valor específico entre linhas
Roadmap - Necessidades técnicas e de negócio
3

SELECT
o.id, o.customer_id, o.status, o.total
FROM raw_order.orders o
WHERE o.status != 'cancelado'
AND o.created_at BETWEEN '2018-08-01'::date AND now()
ORDER BY o.created_at DESC;
4
Estudo de Caso

5
Paginação - Antes
SELECT
o.id, o.customer_id, o.status, o.total
ORDER BY o.created_at DESC
LIMIT 50 OFFSET 0;
SELECT COUNT(*)
AND o.created_at BETWEEN '2018-08-01'::date AND now();

6
Problemas
● Custo de Execução
● Nova query = Novo I/O
● Os dados são os mesmos!

SELECT
o.id, o.customer_id, o.status, o.total,
(
SELECT COUNT(*)
) total_rows
LIMIT 50 OFFSET 150;
7
Eu sei resolver!

SELECT
(
SELECT COUNT(*)
) total_rows
8
Eu sei resolver!

9
Paginação - Depois
SELECT
COUNT(*) OVER() total_rows
LIMIT 50 OFFSET 0;
blz:meetup ANTES +1s
blz:meetup DEPOIS +284ms

10
Window Functions
● Cálculos
● Relação entre linhas
● De agregação ou Analíticas
● Cláusula OVER()
● Espera o critério de relação para análise

11
Sumarização
SELECT
COUNT(*) OVER() total_rows,
SUM(o.total) OVER() sum_total

12
Dúvida de Negócio - Proporção de Frete
SELECT
i.order_id, i.sku, o.freight, i.price, o.sub_total,
i.quantity * i.price
/ SUM(i.quantity * i.price)
OVER(PARTITION BY i.order_id) proporcao_item,
o.freight * (
OVER(PARTITION BY i.order_id)
) frete_proporcional
FROM raw_order.item i
JOIN raw_order.orders o ON i.order_id = o.id
WHERE i.order_id IN(1, 2)
ORDER BY 1

13
Dúvida de Negócio - Proporção de Frete
SELECT
i.order_id, i.sku, o.freight, i.price, o.sub_total,
OVER(PARTITION BY i.order_id) proporcao_item,
o.freight * (
OVER(PARTITION BY i.order_id)
) frete_proporcional
FROM raw_order.item i
JOIN raw_order.orders o ON i.order_id = o.id
WHERE i.order_id IN(1, 2)
ORDER BY 1

14
Relação entre as linhas
OVER(PARTITION BY critério)
SELECT
SUM(o.total)
OVER(PARTITION BY o.customer_id) customer_total
AND o.created_at BETWEEN
'2018-01-01'::date AND now()
AND o.customer_id IN(1, 2)
ORDER BY o.customer_id, o.created_at DESC;
id | customer | status | total | customer_total
----+----------+------------------+----------+---------------
17 | 1 | aguardando envio | 573.9 | 3409.36
6 | 1 | enviado | 31.80 | 3409.36
5 | 1 | enviado | 186.70 | 3409.36
4 | 1 | devolucao total | 120.40 | 3409.36
3 | 1 | enviado | 410.96 | 3409.36
8 | 2 | enviado | 108.70 | 689.13
2 | 2 | enviado | 51.92 | 689.13
1 | 2 | enviado | 246.72 | 689.13

15
SELECT
SUM(o.total)
'2018-01-01'::date AND now()
----+----------+------------------+----------+---------------
6 | 1 | enviado | 31.80 | 4098.49
5 | 1 | enviado | 186.70 | 4098.49
4 | 1 | devolucao total | 120.40 | 4098.49
3 | 1 | enviado | 410.96 | 4098.49
8 | 2 | enviado | 108.70 | 4098.49
2 | 2 | enviado | 51.92 | 4098.49
1 | 2 | enviado | 246.72 | 4098.49

16
SELECT
SUM(o.total)
'2018-01-01'::date AND now()
----+----------+------------------+----------+---------------
6 | 1 | enviado | 31.80 | 3409.36
5 | 1 | enviado | 186.70 | 3409.36
4 | 1 | devolucao total | 120.40 | 3409.36
3 | 1 | enviado | 410.96 | 3409.36
8 | 2 | enviado | 108.70 | 689.13
2 | 2 | enviado | 51.92 | 689.13
1 | 2 | enviado | 246.72 | 689.13

17
OVER(ORDER BY critério)
SELECT
o.id, o.created_at::date, o.total,
SUM(o.total) OVER(ORDER BY o.created_at ASC) total_so_far,
SUM(o.total) OVER() total_geral
AND o.created_at
BETWEEN '2018-01-01'::date AND now()
AND o.customer_id = 1
ORDER BY o.created_at ASC

18
Combinando...
SELECT
o.id, o.customer_id, o.created_at::date, o.total,
SUM(o.total) OVER(PARTITION BY o.customer_id ORDER BY o.created_at ASC) total_so_far,
SUM(o.total) OVER(PARTITION BY o.customer_id) total_geral
AND o.created_at
BETWEEN '2018-01-01'::date AND now()
ORDER BY o.customer_id, o.created_at ASC

19
Outro exemplo...
SELECT
dt_event::date, nm_event, vl_event,
SUM(vl_event) OVER(ORDER BY dt_event ASC) vl_balance
FROM extrato
ORDER BY dt_event

Somente funcionam com cláusula OVER()
20
Funções Analíticas - Por Posição da Linha
SELECT
dt_event,
ROW_NUMBER() OVER(per_event_day) nu_row,
RANK() OVER(per_event_day) nu_rank,
DENSE_RANK() OVER(per_event_day) nu_dense_rank,
PERCENT_RANK() OVER(per_event_time) pc_rank
FROM extrato
WINDOW
per_event_time
AS(ORDER BY dt_event ASC),
per_event_day
AS(ORDER BY dt_event::date ASC)
ORDER BY dt_event
dt_event | nu_row | nu_rank | nu_dense_rank | pc_rank
---------------------+--------+---------+---------------+---------
2018-03-15 10:00:00 | 1 | 1 | 1 | 0
2018-03-15 12:00:00 | 2 | 1 | 1 | 0.25
2018-03-16 12:00:00 | 3 | 3 | 2 | 0.5
2018-03-16 20:00:00 | 4 | 3 | 2 | 0.75
2018-03-17 10:00:00 | 5 | 5 | 3 | 1

Recebem o campo que se deseja analisar
21
Funções Analíticas - Análise de Valor
SELECT
dt_event::date, vl_event, ROW_NUMBER() OVER(per_event_time) nu_row,
LAG(vl_event) OVER(per_event_time) vl_prev, LEAD(vl_event) OVER(per_event_time) vl_next,
FIRST_VALUE(vl_event) OVER(per_event_time) vl_first, LAST_VALUE(vl_event) OVER(per_event_time) vl_last
FROM extrato
WINDOW per_event_time AS(ORDER BY dt_event ASC)
ORDER BY dt_event
dt_event | vl_event | nu_row | vl_prev | vl_next | vl_first | vl_last
------------+----------+--------+---------+---------+----------+---------
2018-03-15 | 2000 | 1 | | -98 | 2000 | 2000
2018-03-15 | -98 | 2 | 2000 | -40 | 2000 | -98
2018-03-16 | -40 | 3 | -98 | -160 | 2000 | -40
2018-03-16 | -160 | 4 | -40 | 50 | 2000 | -160
2018-03-17 | 50 | 5 | -160 | | 2000 | 50

22
Na prática...
SELECT
dt_cotacao, vl_cotacao,
vl_cotacao
- LAG(vl_cotacao) OVER(per_date)
AS vl_variacao,
100 * (vl_cotacao
/ LAG(vl_cotacao) OVER(per_date) - 1
) AS pc_variacao
FROM petr3
WINDOW per_date
AS(ORDER BY dt_cotacao ASC)
ORDER BY 1 DESC

Still using Windows 3.1?
So why stick to SQL-92?
Markus Winand @ modern-sql.com
Conclusão
23

Obrigado
/paulinobruno
● https://modern-sql.com/
● https://use-the-index-luke.com/
24

Meetup Beleza na Web - Funções analíticas com SQL

More Related Content

Similar to Meetup Beleza na Web - Funções analíticas com SQL (18)

Recently uploaded (20)

Meetup Beleza na Web - Funções analíticas com SQL