优化此慢速mysql查询

Question

SELECT SUM( a.situacao =2 OR a.situacao =0 ) AS cotacoes, 
       SUM( a.situacao =1 ) AS n_publicar, 
       SUM( a.inicio <= NOW( ) AND DATE_ADD( a.inicio, INTERVAL a.duracao HOUR ) >= NOW( ) AND a.situacao =2 ) AS disputa, 
       SUM( a.inicio > NOW( ) AND a.situacao =2 ) AS agendados, 
       SUM( DATE_ADD( a.inicio, INTERVAL a.duracao HOUR ) <= NOW( ) AND a.situacao =2 AND a.id_vencedor = -2 ) AS analise, 
       SUM( a.prazoi > NOW( ) AND a.situacao =2 AND a.id_vencedor >0 ) AS aguardando_execucao, 
       SUM( a.situacao =0 ) AS cancelados, 
       SUM( DATE_ADD( a.inicio, INTERVAL a.duracao HOUR ) <= NOW( ) AND a.situacao =2 AND a.id_vencedor = -3 ) AS fracassados, 
       SUM( a.prazot < NOW( )  AND a.situacao =2 AND a.id_vencedor >0 AND ( b.id_avaliacao IS NULL  OR b.id_avaliacao =  '' ) ) AS avaliacao, 
       SUM( a.situacao =2 AND a.id_vencedor >0 AND a.prazot > NOW( )  AND a.prazoi <= NOW( ) ) AS execucao 
FROM leilao_pregoes a 
LEFT JOIN leilao_avaliacao b ON a.id_pregao = b.id_pregao AND b.situacao = 1
INNER JOIN leilao_edital c ON a.id_edital = c.id_edital 
WHERE c.id_comprador =1

I know that this can be improved...

Answer 1

I wonder whether COUNT would be faster where feasible.

The main improvement would be to fetch the columns as records:

cotacoes    123
n_publicar   456
...

SELECT 'cotacoes', SUM(...)
...
UNION SELECT 'n_publicar', SUM(...)
...

The reason being that the conditions are partially exclusing and joins could then be better exploited.

Also you might then tackle the slowest SELECT separately. And you could do parallel queries, splitting in subqueries.

Answer 2

LEFT JOIN leilao_avaliacao b ON a.id_pregao = b.id_pregao AND b.situacao = 1

leilao_avaliacao needs INDEX(id_pregao, situacao) in either order.

INNER JOIN leilao_edital c ON a.id_edital = c.id_edital
WHERE c.id_comprador =1

leilao_edital needs INDEX(id_comprador, id_edital) in either order

For further discussion, please provide SHOW CREATE TABLE and give us some hints of how big the tables are.