Postgres查询效率低下

Question

Have a Postgres Query that is too inefficient. Hoping someone could point me in the right direction to make this query reasonable to run. I know the sum line is the issue just not how to improve / fix it.

/* All Study Transfer Times */
WITH cte_images_duration AS
(
  SELECT
    ssi.study_id,
    min(ssi.created_dt)                       AS first_image_created_dt,
    --max(ssi.created_dt)                       AS last_image_created_dt,
    max(ssi.created_dt) - min(ssi.created_dt) AS images_recieve_duration,
    --count(ssi.study_id)                       AS image_count,
    sum(  cast(btrim(ssif.file_size, 'KB') as real)   ) as transferSize
  FROM  study_series_instances AS ssi, study_series_instance_files as ssif
  where ssi.created_dt >= '2017-1-6'
  GROUP BY ssi.study_id
)
SELECT
    s.institution                                                                           AS "Institution",
    s.accession_no                                                                          AS "Accession#",
    s.modalities,
    to_char(timezone(f.time_zone, cte1.first_image_created_dt), 'MM/DD/YYYY HH:MI:SS pm')   AS "Transfer Start Date/Time",
  --to_char(timezone(f.time_zone, cte1.last_image_created_dt), 'MM/DD/YYYY HH:MI:SS pm')    AS "Transfer Stop Date/Time",
    to_char(cte1.images_recieve_duration, 'MI:SS')                                          AS "Transfer Time",
    --cte1.image_count                                                                      AS "Image Count",
    cte1.transferSize,
    to_char(timezone(f.time_zone, s.study_unread_dt), 'MM/DD/YYYY HH:MI:SS pm')             AS "Unread Date/Time"
    --to_char(timezone(f.time_zone, s.approved_dt), 'MM/DD/YYYY HH:MI:SS pm')                   AS "Approved Date/Time"
FROM 
  cte_images_duration AS cte1
  INNER JOIN studies AS s on s.id = cte1.study_id
  INNER JOIN facilities f ON f.id = s.facility_id
  where s.study_unread_dt > '2017-1-6'
ORDER BY s.accession_no
limit 10

Answer 1

Probably you've missed a condition while joining study_series_instances and study_series_instance_files. Like:

FROM study_series_instances AS ssi
  JOIN study_series_instance_files as ssif ON ssif.some_field = ssi.some_field

Answer 2

Apart from @Dmitry and @a_horse_with_no_name considerations, I'd try it this way as well:

SELECT
    s.institution                                                                           AS "Institution",
    s.accession_no                                                                          AS "Accession#",
    s.modalities,
    to_char(timezone(f.time_zone, cte1.first_image_created_dt), 'MM/DD/YYYY HH:MI:SS pm')   AS "Transfer Start Date/Time",
    to_char(cte1.images_recieve_duration, 'MI:SS')                                          AS "Transfer Time",
    cte1.transferSize,
    to_char(timezone(f.time_zone, s.study_unread_dt), 'MM/DD/YYYY HH:MI:SS pm')             AS "Unread Date/Time"
FROM
  (SELECT
      ssi.study_id,
      min(ssi.created_dt)                       AS first_image_created_dt,
      max(ssi.created_dt) - min(ssi.created_dt) AS images_recieve_duration,
      sum(cast(btrim(ssif.file_size, 'KB') as real)) AS transferSize
    FROM  
          study_series_instances AS ssi, 
          /* Some JOIN condition is expected here */
          study_series_instance_files AS ssif
    WHERE ssi.created_dt >= '2017-1-6'
    GROUP BY ssi.study_id
  ) AS cte1
  INNER JOIN studies AS s on s.id = cte1.study_id
  INNER JOIN facilities f ON f.id = s.facility_id
WHERE s.study_unread_dt > '2017-1-6'
ORDER BY s.accession_no
LIMIT 10

CTE expressions in PostgreSQL are optimization fences , and sometimes that's not what you want.