PG GraphQL详细介绍与基本使用

自己的九又四分之三站台2026-01-03 14:01
sql 复制代码 
SELECT
  (SELECT COUNT(*) FROM imdb.title_basics)     AS titles,
  (SELECT COUNT(*) FROM imdb.name_basics)      AS names,
  (SELECT COUNT(*) FROM imdb.title_principals) AS principals;

![[2_1查询数据入库结果.png]]

表结构说明

1️⃣ 作品表（title_basics）

sql 复制代码 
\d imdb.title_basics;

重点字段解释：

字段 含义
tconst 作品唯一 ID（tt 开头）
titleType movie / tvSeries / short
primaryTitle 常用标题
startYear 上映年份
runtimeMinutes 时长
genres 逗号分隔

2️⃣ 人物表（name_basics）

复制代码 
\d imdb.name_basics;
字段 含义
nconst 人物唯一 ID（nm 开头）
primaryName 人名
birthYear 出生年
primaryProfession actor, director 等

3️⃣ 关系表（title_principals）

复制代码 
\d imdb.title_principals;

这是最关键的一张表，也是图模型的核心。

字段 含义
tconst 作品 ID
nconst 人物 ID
category actor / actress / director
characters 饰演角色

2. 准备小量数据

sql 复制代码 
DROP TABLE IF EXISTS imdb.small_name_basics;
DROP TABLE IF EXISTS imdb.small_title_basics;
DROP TABLE IF EXISTS imdb.small_title_principals;


CREATE TABLE imdb.small_title_basics AS
SELECT *
FROM imdb.title_basics
WHERE startyear BETWEEN 2015 AND 2018
  AND isadult = 0
  AND titletype IN ('movie','tvSeries','tvMiniSeries')
ORDER BY tconst
LIMIT 2000;     -- ⭐ 控制规模

ALTER TABLE imdb.small_title_basics
  ADD PRIMARY KEY (tconst);


CREATE TABLE imdb.small_title_principals AS
SELECT p.*
FROM imdb.title_principals p
JOIN imdb.small_title_basics t
  ON t.tconst = p.tconst
WHERE p.category IN ('actor','actress');

CREATE INDEX ON imdb.small_title_principals(tconst);
CREATE INDEX ON imdb.small_title_principals(nconst);
CREATE INDEX ON imdb.small_title_principals(category);


CREATE TABLE imdb.small_name_basics AS
SELECT n.*
FROM imdb.name_basics n
JOIN (
  SELECT DISTINCT nconst
  FROM imdb.small_title_principals
) x ON x.nconst = n.nconst;

ALTER TABLE imdb.small_name_basics
  ADD PRIMARY KEY (nconst);


SELECT
  (SELECT count(*) FROM imdb.small_name_basics)      AS people,
  (SELECT count(*) FROM imdb.small_title_basics)     AS titles,
  (SELECT count(*) FROM imdb.small_title_principals) AS edges;

3. 效率比对

3.1. 两人是否合作过

3.1.1. Cypher

sql 复制代码 
EXPLAIN (ANALYZE, BUFFERS)
SELECT *
FROM cypher('imdb_graph', $$
  MATCH (a:Person {primaryname:'Melissa Peters'})
        -[:WORKED_ON]->(t:Title)<-[:WORKED_ON]-
        (b:Person {primaryname:'Jason Hopley'})
  RETURN count(DISTINCT t) AS co_titles
$$) AS (co_titles agtype);
sql 复制代码 
Aggregate  (cost=304.13..304.14 rows=1 width=32) (actual time=1.353..1.355 rows=1.00 loops=1)
  Buffers: shared hit=297
  ->  Sort  (cost=304.11..304.12 rows=1 width=246) (actual time=1.349..1.350 rows=1.00 loops=1)
        Sort Key: (_agtype_build_vertex(t.id, _label_name('25188'::oid, t.id), t.properties))
        Sort Method: quicksort  Memory: 25kB
        Buffers: shared hit=297
        ->  Nested Loop  (cost=1.14..304.10 rows=1 width=246) (actual time=1.206..1.346 rows=1.00 loops=1)
              Buffers: shared hit=297
              ->  Nested Loop  (cost=0.84..302.61 rows=2 width=254) (actual time=1.197..1.314 rows=32.00 loops=1)
                    Buffers: shared hit=264
                    ->  Nested Loop  (cost=0.56..302.05 rows=1 width=262) (actual time=1.193..1.296 rows=7.00 loops=1)
                          Buffers: shared hit=241
                          ->  Nested Loop  (cost=0.28..301.67 rows=1 width=16) (actual time=1.189..1.285 rows=7.00 loops=1)
                                Buffers: shared hit=220
                                ->  Seq Scan on "Person" a  (cost=0.00..293.36 rows=1 width=8) (actual time=1.180..1.272 rows=1.00 loops=1)
                                      Filter: (properties @> '{"primaryname": "Melissa Peters"}'::agtype)
                                      Rows Removed by Filter: 6348
                                      Buffers: shared hit=214
                                ->  Index Scan using "WORKED_ON_start_id_idx" on "WORKED_ON" _age_default_alias_0  (cost=0.28..8.30 rows=1 width=24) (actual time=0.007..0.010 rows=7.00 loops=1)
                                      Index Cond: (start_id = a.id)
                                      Index Searches: 1
                                      Buffers: shared hit=6
                          ->  Index Scan using "Title_pkey" on "Title" t  (cost=0.28..0.38 rows=1 width=246) (actual time=0.001..0.001 rows=1.00 loops=7)
                                Index Cond: (id = _age_default_alias_0.end_id)
                                Index Searches: 7
                                Buffers: shared hit=21
                    ->  Index Scan using "WORKED_ON_end_id_idx" on "WORKED_ON" _age_default_alias_1  (cost=0.28..0.54 rows=2 width=24) (actual time=0.001..0.002 rows=4.57 loops=7)
                          Index Cond: (end_id = _age_default_alias_0.end_id)
                          Filter: _ag_enforce_edge_uniqueness2(_age_default_alias_0.id, id)
                          Rows Removed by Filter: 1
                          Index Searches: 7
                          Buffers: shared hit=23
              ->  Memoize  (cost=0.29..0.73 rows=1 width=8) (actual time=0.001..0.001 rows=0.03 loops=32)
                    Cache Key: _age_default_alias_1.start_id
                    Cache Mode: logical
                    Hits: 21  Misses: 11  Evictions: 0  Overflows: 0  Memory Usage: 1kB
                    Buffers: shared hit=33
                    ->  Index Scan using "Person_pkey" on "Person" b  (cost=0.28..0.72 rows=1 width=8) (actual time=0.002..0.002 rows=0.09 loops=11)
                          Index Cond: (id = _age_default_alias_1.start_id)
                          Filter: (properties @> '{"primaryname": "Jason Hopley"}'::agtype)
                          Rows Removed by Filter: 1
                          Index Searches: 11
                          Buffers: shared hit=33
Planning:
  Buffers: shared hit=60
Planning Time: 0.535 ms
Execution Time: 1.388 ms

3.1.2. SQL

sql 复制代码 
EXPLAIN (ANALYZE, BUFFERS)
WITH
a AS (SELECT nconst FROM imdb.small_name_basics WHERE primaryname='Melissa Peters' LIMIT 1),
b AS (SELECT nconst FROM imdb.small_name_basics WHERE primaryname='Jason Hopley' LIMIT 1)
SELECT count(DISTINCT p1.tconst) AS co_titles
FROM imdb.small_title_principals p1
JOIN a ON a.nconst = p1.nconst
JOIN imdb.small_title_principals p2 ON p2.tconst = p1.tconst
JOIN b ON b.nconst = p2.nconst;
sql 复制代码 
Aggregate  (cost=25.51..25.52 rows=1 width=8) (actual time=0.952..0.953 rows=1.00 loops=1)
  Buffers: shared hit=27 read=4
  ->  Sort  (cost=25.50..25.50 rows=1 width=10) (actual time=0.948..0.949 rows=1.00 loops=1)
        Sort Key: p1.tconst
        Sort Method: quicksort  Memory: 25kB
        Buffers: shared hit=27 read=4
        ->  Nested Loop  (cost=1.13..25.49 rows=1 width=10) (actual time=0.674..0.943 rows=1.00 loops=1)
              Join Filter: (p2.nconst = small_name_basics.nconst)
              Rows Removed by Join Filter: 38
              Buffers: shared hit=27 read=4
              ->  Limit  (cost=0.28..8.30 rows=1 width=10) (actual time=0.250..0.250 rows=1.00 loops=1)
                    Buffers: shared hit=2 read=1
                    ->  Index Scan using idx_small_name_primaryname on small_name_basics  (cost=0.28..8.30 rows=1 width=10) (actual time=0.249..0.249 rows=1.00 loops=1)
                          Index Cond: (primaryname = 'Jason Hopley'::text)
                          Index Searches: 1
                          Buffers: shared hit=2 read=1
              ->  Nested Loop  (cost=0.85..17.10 rows=7 width=20) (actual time=0.421..0.688 rows=39.00 loops=1)
                    Buffers: shared hit=25 read=3
                    ->  Nested Loop  (cost=0.56..16.61 rows=1 width=10) (actual time=0.225..0.228 rows=7.00 loops=1)
                          Buffers: shared hit=6 read=1
                          ->  Limit  (cost=0.28..8.30 rows=1 width=10) (actual time=0.007..0.007 rows=1.00 loops=1)
                                Buffers: shared hit=3
                                ->  Index Scan using idx_small_name_primaryname on small_name_basics small_name_basics_1  (cost=0.28..8.30 rows=1 width=10) (actual time=0.007..0.007 rows=1.00 loops=1)
                                      Index Cond: (primaryname = 'Melissa Peters'::text)
                                      Index Searches: 1
                                      Buffers: shared hit=3
                          ->  Index Scan using idx_small_prin_nconst on small_title_principals p1  (cost=0.28..8.30 rows=1 width=20) (actual time=0.216..0.218 rows=7.00 loops=1)
                                Index Cond: (nconst = small_name_basics_1.nconst)
                                Index Searches: 1
                                Buffers: shared hit=3 read=1
                    ->  Index Scan using idx_small_prin_tconst on small_title_principals p2  (cost=0.28..0.43 rows=6 width=20) (actual time=0.064..0.065 rows=5.57 loops=7)
                          Index Cond: (tconst = p1.tconst)
                          Index Searches: 7
                          Buffers: shared hit=19 read=2
Planning:
  Buffers: shared hit=57 read=9 dirtied=2
Planning Time: 3.234 ms
Execution Time: 0.983 ms

3.2. 两人共同合作过哪些作品（清单）

3.2.1. Cypher

sql 复制代码 
EXPLAIN (ANALYZE, BUFFERS)
SELECT *
FROM cypher('imdb_graph', $$
  MATCH (a:Person {primaryname:'Melissa Peters'})
        -[:WORKED_ON]->(t:Title)<-[:WORKED_ON]-
        (b:Person {primaryname:'Daniel F.K. Fernandes'})
  RETURN DISTINCT
         t.primarytitle AS title,
         t.startyear    AS year,
         t.titletype    AS type
  ORDER BY year
$$) AS (title agtype, year agtype, type agtype);
sql 复制代码 
Unique  (cost=304.14..304.15 rows=1 width=96) (actual time=1.391..1.396 rows=6.00 loops=1)
  Buffers: shared hit=297
  ->  Sort  (cost=304.14..304.14 rows=1 width=96) (actual time=1.390..1.392 rows=11.00 loops=1)
        Sort Key: (agtype_access_operator(VARIADIC ARRAY[_agtype_build_vertex(t.id, _label_name('25188'::oid, t.id), t.properties), '"startyear"'::agtype])), (agtype_access_operator(VARIADIC ARRAY[_agtype_build_vertex(t.id, _label_name('25188'::oid, t.id), t.properties), '"primarytitle"'::agtype])), (agtype_access_operator(VARIADIC ARRAY[_agtype_build_vertex(t.id, _label_name('25188'::oid, t.id), t.properties), '"titletype"'::agtype]))
        Sort Method: quicksort  Memory: 25kB
        Buffers: shared hit=297
        ->  Nested Loop  (cost=1.14..304.13 rows=1 width=96) (actual time=1.214..1.374 rows=11.00 loops=1)
              Buffers: shared hit=297
              ->  Nested Loop  (cost=0.84..302.61 rows=2 width=254) (actual time=1.177..1.301 rows=32.00 loops=1)
                    Buffers: shared hit=264
                    ->  Nested Loop  (cost=0.56..302.05 rows=1 width=262) (actual time=1.171..1.282 rows=7.00 loops=1)
                          Buffers: shared hit=241
                          ->  Nested Loop  (cost=0.28..301.67 rows=1 width=16) (actual time=1.163..1.268 rows=7.00 loops=1)
                                Buffers: shared hit=220
                                ->  Seq Scan on "Person" a  (cost=0.00..293.36 rows=1 width=8) (actual time=1.094..1.194 rows=1.00 loops=1)
                                      Filter: (properties @> '{"primaryname": "Melissa Peters"}'::agtype)
                                      Rows Removed by Filter: 6348
                                      Buffers: shared hit=214
                                ->  Index Scan using "WORKED_ON_start_id_idx" on "WORKED_ON" _age_default_alias_0  (cost=0.28..8.30 rows=1 width=24) (actual time=0.065..0.069 rows=7.00 loops=1)
                                      Index Cond: (start_id = a.id)
                                      Index Searches: 1
                                      Buffers: shared hit=6
                          ->  Index Scan using "Title_pkey" on "Title" t  (cost=0.28..0.38 rows=1 width=246) (actual time=0.002..0.002 rows=1.00 loops=7)
                                Index Cond: (id = _age_default_alias_0.end_id)
                                Index Searches: 7
                                Buffers: shared hit=21
                    ->  Index Scan using "WORKED_ON_end_id_idx" on "WORKED_ON" _age_default_alias_1  (cost=0.28..0.54 rows=2 width=24) (actual time=0.001..0.002 rows=4.57 loops=7)
                          Index Cond: (end_id = _age_default_alias_0.end_id)
                          Filter: _ag_enforce_edge_uniqueness2(_age_default_alias_0.id, id)
                          Rows Removed by Filter: 1
                          Index Searches: 7
                          Buffers: shared hit=23
              ->  Memoize  (cost=0.29..0.73 rows=1 width=8) (actual time=0.001..0.001 rows=0.34 loops=32)
                    Cache Key: _age_default_alias_1.start_id
                    Cache Mode: logical
                    Hits: 21  Misses: 11  Evictions: 0  Overflows: 0  Memory Usage: 1kB
                    Buffers: shared hit=33
                    ->  Index Scan using "Person_pkey" on "Person" b  (cost=0.28..0.72 rows=1 width=8) (actual time=0.002..0.002 rows=0.09 loops=11)
                          Index Cond: (id = _age_default_alias_1.start_id)
                          Filter: (properties @> '{"primaryname": "Daniel F.K. Fernandes"}'::agtype)
                          Rows Removed by Filter: 1
                          Index Searches: 11
                          Buffers: shared hit=33
Planning:
  Buffers: shared hit=60
Planning Time: 0.515 ms
Execution Time: 1.426 ms

3.2.2. SQL

sql 复制代码 
EXPLAIN (ANALYZE, BUFFERS)
WITH
a AS (SELECT nconst FROM imdb.small_name_basics WHERE primaryname='Melissa Peters' LIMIT 1),
b AS (SELECT nconst FROM imdb.small_name_basics WHERE primaryname='Daniel F.K. Fernandes' LIMIT 1)
SELECT DISTINCT
  t.primarytitle AS title,
  t.startyear    AS year,
  t.titletype    AS type
FROM imdb.small_title_principals p1
JOIN a ON a.nconst = p1.nconst
JOIN imdb.small_title_principals p2 ON p2.tconst = p1.tconst
JOIN b ON b.nconst = p2.nconst
JOIN imdb.small_title_basics t ON t.tconst = p1.tconst
ORDER BY year;
sql 复制代码 
Unique  (cost=25.82..25.83 rows=1 width=30) (actual time=0.527..0.531 rows=6.00 loops=1)
  Buffers: shared hit=63 read=1
  ->  Sort  (cost=25.82..25.82 rows=1 width=30) (actual time=0.527..0.528 rows=11.00 loops=1)
        Sort Key: t.startyear, t.primarytitle, t.titletype
        Sort Method: quicksort  Memory: 25kB
        Buffers: shared hit=63 read=1
        ->  Nested Loop  (cost=1.41..25.81 rows=1 width=30) (actual time=0.459..0.505 rows=11.00 loops=1)
              Buffers: shared hit=63 read=1
              ->  Nested Loop  (cost=1.13..25.49 rows=1 width=20) (actual time=0.453..0.479 rows=11.00 loops=1)
                    Join Filter: (p2.nconst = small_name_basics.nconst)
                    Rows Removed by Join Filter: 28
                    Buffers: shared hit=30 read=1
                    ->  Limit  (cost=0.28..8.30 rows=1 width=10) (actual time=0.431..0.431 rows=1.00 loops=1)
                          Buffers: shared hit=2 read=1
                          ->  Index Scan using idx_small_name_primaryname on small_name_basics  (cost=0.28..8.30 rows=1 width=10) (actual time=0.430..0.430 rows=1.00 loops=1)
                                Index Cond: (primaryname = 'Daniel F.K. Fernandes'::text)
                                Index Searches: 1
                                Buffers: shared hit=2 read=1
                    ->  Nested Loop  (cost=0.85..17.10 rows=7 width=30) (actual time=0.018..0.043 rows=39.00 loops=1)
                          Buffers: shared hit=28
                          ->  Nested Loop  (cost=0.56..16.61 rows=1 width=10) (actual time=0.013..0.017 rows=7.00 loops=1)
                                Buffers: shared hit=7
                                ->  Limit  (cost=0.28..8.30 rows=1 width=10) (actual time=0.007..0.007 rows=1.00 loops=1)
                                      Buffers: shared hit=3
                                      ->  Index Scan using idx_small_name_primaryname on small_name_basics small_name_basics_1  (cost=0.28..8.30 rows=1 width=10) (actual time=0.007..0.007 rows=1.00 loops=1)
                                            Index Cond: (primaryname = 'Melissa Peters'::text)
                                            Index Searches: 1
                                            Buffers: shared hit=3
                                ->  Index Scan using idx_small_prin_nconst on small_title_principals p1  (cost=0.28..8.30 rows=1 width=20) (actual time=0.005..0.008 rows=7.00 loops=1)
                                      Index Cond: (nconst = small_name_basics_1.nconst)
                                      Index Searches: 1
                                      Buffers: shared hit=4
                          ->  Index Scan using idx_small_prin_tconst on small_title_principals p2  (cost=0.28..0.43 rows=6 width=20) (actual time=0.002..0.003 rows=5.57 loops=7)
                                Index Cond: (tconst = p1.tconst)
                                Index Searches: 7
                                Buffers: shared hit=21
              ->  Index Scan using small_title_basics_pkey on small_title_basics t  (cost=0.28..0.32 rows=1 width=40) (actual time=0.002..0.002 rows=1.00 loops=11)
                    Index Cond: (tconst = p2.tconst)
                    Index Searches: 11
                    Buffers: shared hit=33
Planning:
  Buffers: shared hit=48
Planning Time: 0.727 ms
Execution Time: 0.556 ms

3.3. 某演员合作次数最多的搭档 Top10

3.3.1. Cypher

sql 复制代码 
EXPLAIN (ANALYZE, BUFFERS)
SELECT *
FROM cypher('imdb_graph', $$
  MATCH (a:Person {primaryname:'Melissa Peters'})
        -[:WORKED_ON]->(t:Title)<-[:WORKED_ON]-(b:Person)
  WHERE a <> b
  WITH b, count(DISTINCT t) AS co_titles
  RETURN b.primaryname AS partner, co_titles
  ORDER BY co_titles DESC
  LIMIT 10
$$) AS (partner agtype, co_titles agtype);
sql 复制代码 
Limit  (cost=304.19..304.20 rows=2 width=64) (actual time=1.600..1.603 rows=10.00 loops=1)
  Buffers: shared hit=360
  ->  Sort  (cost=304.19..304.20 rows=2 width=64) (actual time=1.599..1.601 rows=10.00 loops=1)
        Sort Key: _age_default_alias_previous_cypher_clause.co_titles DESC
        Sort Method: quicksort  Memory: 25kB
        Buffers: shared hit=360
        ->  Subquery Scan on _age_default_alias_previous_cypher_clause  (cost=304.12..304.18 rows=2 width=64) (actual time=1.548..1.593 rows=10.00 loops=1)
              Buffers: shared hit=360
              ->  GroupAggregate  (cost=304.12..304.18 rows=2 width=64) (actual time=1.544..1.584 rows=10.00 loops=1)
                    Group Key: (_agtype_build_vertex(b.id, _label_name('25188'::oid, b.id), b.properties))
                    Buffers: shared hit=360
                    ->  Sort  (cost=304.12..304.12 rows=2 width=278) (actual time=1.535..1.537 rows=30.00 loops=1)
                          Sort Key: (_agtype_build_vertex(b.id, _label_name('25188'::oid, b.id), b.properties)), (_agtype_build_vertex(t.id, _label_name('25188'::oid, t.id), t.properties))
                          Sort Method: quicksort  Memory: 49kB
                          Buffers: shared hit=360
                          ->  Nested Loop  (cost=1.12..304.11 rows=2 width=278) (actual time=1.055..1.349 rows=30.00 loops=1)
                                Join Filter: (_agtype_build_vertex(a.id, _label_name('25188'::oid, a.id), a.properties) <> _agtype_build_vertex(b.id, _label_name('25188'::oid, b.id), b.properties))
                                Rows Removed by Join Filter: 2
                                Buffers: shared hit=360
                                ->  Nested Loop  (cost=0.84..302.61 rows=2 width=429) (actual time=1.042..1.254 rows=32.00 loops=1)
                                      Buffers: shared hit=264
                                      ->  Nested Loop  (cost=0.56..302.05 rows=1 width=437) (actual time=1.037..1.236 rows=7.00 loops=1)
                                            Buffers: shared hit=241
                                            ->  Nested Loop  (cost=0.28..301.67 rows=1 width=191) (actual time=1.032..1.225 rows=7.00 loops=1)
                                                  Buffers: shared hit=220
                                                  ->  Seq Scan on "Person" a  (cost=0.00..293.36 rows=1 width=175) (actual time=1.023..1.212 rows=1.00 loops=1)
                                                        Filter: (properties @> '{"primaryname": "Melissa Peters"}'::agtype)
                                                        Rows Removed by Filter: 6348
                                                        Buffers: shared hit=214
                                                  ->  Index Scan using "WORKED_ON_start_id_idx" on "WORKED_ON" _age_default_alias_0  (cost=0.28..8.30 rows=1 width=24) (actual time=0.006..0.009 rows=7.00 loops=1)
                                                        Index Cond: (start_id = a.id)
                                                        Index Searches: 1
                                                        Buffers: shared hit=6
                                            ->  Index Scan using "Title_pkey" on "Title" t  (cost=0.28..0.38 rows=1 width=246) (actual time=0.001..0.001 rows=1.00 loops=7)
                                                  Index Cond: (id = _age_default_alias_0.end_id)
                                                  Index Searches: 7
                                                  Buffers: shared hit=21
                                      ->  Index Scan using "WORKED_ON_end_id_idx" on "WORKED_ON" _age_default_alias_1  (cost=0.28..0.54 rows=2 width=24) (actual time=0.001..0.002 rows=4.57 loops=7)
                                            Index Cond: (end_id = _age_default_alias_0.end_id)
                                            Filter: _ag_enforce_edge_uniqueness2(_age_default_alias_0.id, id)
                                            Rows Removed by Filter: 1
                                            Index Searches: 7
                                            Buffers: shared hit=23
                                ->  Index Scan using "Person_pkey" on "Person" b  (cost=0.28..0.72 rows=1 width=175) (actual time=0.001..0.001 rows=1.00 loops=32)
                                      Index Cond: (id = _age_default_alias_1.start_id)
                                      Index Searches: 32
                                      Buffers: shared hit=96
Planning:
  Buffers: shared hit=60
Planning Time: 0.978 ms
Execution Time: 1.694 ms

3.3.2. SQL

sql 复制代码 
EXPLAIN (ANALYZE, BUFFERS)
WITH a AS (
  SELECT nconst FROM imdb.small_name_basics
  WHERE primaryname='Melissa Peters' LIMIT 1
),
pairs AS (
  SELECT p2.nconst AS partner_nconst, p1.tconst
  FROM imdb.small_title_principals p1
  JOIN a ON a.nconst = p1.nconst
  JOIN imdb.small_title_principals p2 ON p2.tconst = p1.tconst
  WHERE p2.nconst <> a.nconst
)
SELECT nb.primaryname AS partner,
       count(DISTINCT tconst) AS co_titles
FROM pairs
JOIN imdb.small_name_basics nb ON nb.nconst = pairs.partner_nconst
GROUP BY nb.primaryname
ORDER BY co_titles DESC
LIMIT 10;
sql 复制代码 
Limit  (cost=19.96..19.97 rows=7 width=22) (actual time=0.242..0.245 rows=10.00 loops=1)
  Buffers: shared hit=109 read=9
  ->  Sort  (cost=19.96..19.97 rows=7 width=22) (actual time=0.241..0.243 rows=10.00 loops=1)
        Sort Key: (count(DISTINCT p1.tconst)) DESC
        Sort Method: quicksort  Memory: 25kB
        Buffers: shared hit=109 read=9
        ->  GroupAggregate  (cost=19.74..19.86 rows=7 width=22) (actual time=0.230..0.237 rows=10.00 loops=1)
              Group Key: nb.primaryname
              Buffers: shared hit=109 read=9
              ->  Sort  (cost=19.74..19.75 rows=7 width=24) (actual time=0.225..0.227 rows=30.00 loops=1)
                    Sort Key: nb.primaryname, p1.tconst
                    Sort Method: quicksort  Memory: 26kB
                    Buffers: shared hit=109 read=9
                    ->  Nested Loop  (cost=1.13..19.64 rows=7 width=24) (actual time=0.039..0.207 rows=30.00 loops=1)
                          Buffers: shared hit=109 read=9
                          ->  Nested Loop  (cost=0.85..17.12 rows=7 width=20) (actual time=0.025..0.052 rows=30.00 loops=1)
                                Join Filter: (p2.nconst <> small_name_basics.nconst)
                                Rows Removed by Join Filter: 9
                                Buffers: shared hit=28
                                ->  Nested Loop  (cost=0.56..16.61 rows=1 width=20) (actual time=0.018..0.021 rows=7.00 loops=1)
                                      Buffers: shared hit=7
                                      ->  Limit  (cost=0.28..8.30 rows=1 width=10) (actual time=0.009..0.010 rows=1.00 loops=1)
                                            Buffers: shared hit=3
                                            ->  Index Scan using idx_small_name_primaryname on small_name_basics  (cost=0.28..8.30 rows=1 width=10) (actual time=0.009..0.009 rows=1.00 loops=1)
                                                  Index Cond: (primaryname = 'Melissa Peters'::text)
                                                  Index Searches: 1
                                                  Buffers: shared hit=3
                                      ->  Index Scan using idx_small_prin_nconst on small_title_principals p1  (cost=0.28..8.30 rows=1 width=20) (actual time=0.007..0.009 rows=7.00 loops=1)
                                            Index Cond: (nconst = small_name_basics.nconst)
                                            Index Searches: 1
                                            Buffers: shared hit=4
                                ->  Index Scan using idx_small_prin_tconst on small_title_principals p2  (cost=0.28..0.43 rows=6 width=20) (actual time=0.003..0.003 rows=5.57 loops=7)
                                      Index Cond: (tconst = p1.tconst)
                                      Index Searches: 7
                                      Buffers: shared hit=21
                          ->  Index Scan using small_name_basics_pkey on small_name_basics nb  (cost=0.28..0.36 rows=1 width=24) (actual time=0.005..0.005 rows=1.00 loops=30)
                                Index Cond: (nconst = p2.nconst)
                                Index Searches: 30
                                Buffers: shared hit=81 read=9
Planning:
  Buffers: shared hit=30 read=3
Planning Time: 0.762 ms
Execution Time: 0.275 ms

3.4. 全局合作次数最多的演员对 Top20

3.4.1. Cypher

sql 复制代码 
EXPLAIN (ANALYZE, BUFFERS)
SELECT *
FROM cypher('imdb_graph', $$
  MATCH (a:Person)-[:WORKED_ON]->(t:Title)<-[:WORKED_ON]-(b:Person)
  WHERE a.nconst < b.nconst
  WITH a, b, count(DISTINCT t) AS co_titles
  RETURN a.primaryname AS actor1,
         b.primaryname AS actor2,
         co_titles
  ORDER BY co_titles DESC
  LIMIT 20
$$) AS (actor1 agtype, actor2 agtype, co_titles agtype);
sql 复制代码 
Limit  (cost=2856.07..2856.12 rows=20 width=96) (actual time=808.293..808.302 rows=20.00 loops=1)
  Buffers: shared hit=974, temp read=3840 written=3847
  ->  Sort  (cost=2856.07..2870.58 rows=5805 width=96) (actual time=808.292..808.299 rows=20.00 loops=1)
        Sort Key: _age_default_alias_previous_cypher_clause.co_titles DESC
        Sort Method: top-N heapsort  Memory: 27kB
        Buffers: shared hit=974, temp read=3840 written=3847
        ->  Subquery Scan on _age_default_alias_previous_cypher_clause  (cost=2454.89..2701.60 rows=5805 width=96) (actual time=632.865..802.809 rows=23551.00 loops=1)
              Buffers: shared hit=974, temp read=3840 written=3847
              ->  GroupAggregate  (cost=2454.89..2672.58 rows=5805 width=96) (actual time=632.857..771.831 rows=23551.00 loops=1)
                    Group Key: (_agtype_build_vertex(a.id, _label_name('25188'::oid, a.id), a.properties)), (_agtype_build_vertex(b.id, _label_name('25188'::oid, b.id), b.properties))
                    Buffers: shared hit=974, temp read=3840 written=3847
                    ->  Sort  (cost=2454.89..2469.40 rows=5805 width=310) (actual time=632.838..729.113 rows=28200.00 loops=1)
                          Sort Key: (_agtype_build_vertex(a.id, _label_name('25188'::oid, a.id), a.properties)), (_agtype_build_vertex(b.id, _label_name('25188'::oid, b.id), b.properties)), (_agtype_build_vertex(t.id, _label_name('25188'::oid, t.id), t.properties))
                          Sort Method: external merge  Disk: 30720kB
                          Buffers: shared hit=974, temp read=3840 written=3847
                          ->  Hash Join  (cost=1261.83..2091.99 rows=5805 width=310) (actual time=5.279..84.794 rows=28200.00 loops=1)
                                Hash Cond: (_age_default_alias_1.start_id = b.id)
                                Join Filter: (agtype_access_operator(VARIADIC ARRAY[a.properties, '"nconst"'::agtype]) < agtype_access_operator(VARIADIC ARRAY[b.properties, '"nconst"'::agtype]))
                                Rows Removed by Join Filter: 29360
                                Buffers: shared hit=974
                                ->  Hash Join  (cost=904.98..1631.35 rows=17415 width=429) (actual time=3.771..14.847 rows=57560.00 loops=1)
                                      Hash Cond: (_age_default_alias_0.end_id = t.id)
                                      Join Filter: _ag_enforce_edge_uniqueness2(_age_default_alias_0.id, _age_default_alias_1.id)
                                      Rows Removed by Join Filter: 7446
                                      Buffers: shared hit=760
                                      ->  Hash Join  (cost=356.85..680.87 rows=7446 width=191) (actual time=0.957..4.194 rows=7446.00 loops=1)
                                            Hash Cond: (_age_default_alias_0.start_id = a.id)
                                            Buffers: shared hit=444
                                            ->  Seq Scan on "WORKED_ON" _age_default_alias_0  (cost=0.00..304.46 rows=7446 width=24) (actual time=0.007..0.714 rows=7446.00 loops=1)
                                                  Buffers: shared hit=230
                                            ->  Hash  (cost=277.49..277.49 rows=6349 width=175) (actual time=0.939..0.939 rows=6349.00 loops=1)
                                                  Buckets: 8192  Batches: 1  Memory Usage: 1349kB
                                                  Buffers: shared hit=214
                                                  ->  Seq Scan on "Person" a  (cost=0.00..277.49 rows=6349 width=175) (actual time=0.003..0.317 rows=6349.00 loops=1)
                                                        Buffers: shared hit=214
                                      ->  Hash  (cost=455.05..455.05 rows=7446 width=270) (actual time=2.803..2.805 rows=7446.00 loops=1)
                                            Buckets: 8192  Batches: 1  Memory Usage: 2287kB
                                            Buffers: shared hit=316
                                            ->  Hash Join  (cost=131.00..455.05 rows=7446 width=270) (actual time=0.506..2.003 rows=7446.00 loops=1)
                                                  Hash Cond: (_age_default_alias_1.end_id = t.id)
                                                  Buffers: shared hit=316
                                                  ->  Seq Scan on "WORKED_ON" _age_default_alias_1  (cost=0.00..304.46 rows=7446 width=24) (actual time=0.005..0.420 rows=7446.00 loops=1)
                                                        Buffers: shared hit=230
                                                  ->  Hash  (cost=106.00..106.00 rows=2000 width=246) (actual time=0.496..0.497 rows=2000.00 loops=1)
                                                        Buckets: 2048  Batches: 1  Memory Usage: 561kB
                                                        Buffers: shared hit=86
                                                        ->  Seq Scan on "Title" t  (cost=0.00..106.00 rows=2000 width=246) (actual time=0.003..0.135 rows=2000.00 loops=1)
                                                              Buffers: shared hit=86
                                ->  Hash  (cost=277.49..277.49 rows=6349 width=175) (actual time=1.475..1.476 rows=6349.00 loops=1)
                                      Buckets: 8192  Batches: 1  Memory Usage: 1349kB
                                      Buffers: shared hit=214
                                      ->  Seq Scan on "Person" b  (cost=0.00..277.49 rows=6349 width=175) (actual time=0.011..0.436 rows=6349.00 loops=1)
                                            Buffers: shared hit=214
Planning:
  Buffers: shared hit=60
Planning Time: 0.860 ms
Execution Time: 825.279 ms

3.4.2. SQL

sql 复制代码 
EXPLAIN (ANALYZE, BUFFERS)
WITH pair_titles AS (
  SELECT
    LEAST(p1.nconst, p2.nconst)    AS n1,
    GREATEST(p1.nconst, p2.nconst) AS n2,
    p1.tconst
  FROM imdb.small_title_principals p1
  JOIN imdb.small_title_principals p2
    ON p2.tconst = p1.tconst
   AND p2.nconst <> p1.nconst
  WHERE p1.category IN ('actor','actress')
    AND p2.category IN ('actor','actress')
),
pair_counts AS (
  SELECT n1, n2, count(DISTINCT tconst) AS co_titles
  FROM pair_titles
  GROUP BY n1, n2
)
SELECT
  n1_name.primaryname AS actor1,
  n2_name.primaryname AS actor2,
  pc.co_titles
FROM pair_counts pc
JOIN imdb.small_name_basics n1_name ON n1_name.nconst = pc.n1
JOIN imdb.small_name_basics n2_name ON n2_name.nconst = pc.n2
ORDER BY pc.co_titles DESC
LIMIT 20;
sql 复制代码 
Limit  (cost=11994.91..11994.96 rows=20 width=36) (actual time=102.020..102.026 rows=20.00 loops=1)
  Buffers: shared hit=308
  ->  Sort  (cost=11994.91..12125.50 rows=52235 width=36) (actual time=102.019..102.023 rows=20.00 loops=1)
        Sort Key: (count(DISTINCT p1.tconst)) DESC
        Sort Method: top-N heapsort  Memory: 26kB
        Buffers: shared hit=308
        ->  Hash Join  (cost=8502.32..10604.96 rows=52235 width=36) (actual time=79.821..99.637 rows=23551.00 loops=1)
              Hash Cond: ((GREATEST(p1.nconst, p2.nconst)) = n2_name.nconst)
              Buffers: shared hit=308
              ->  Hash Join  (cost=8274.47..10239.90 rows=52235 width=54) (actual time=79.033..94.666 rows=23551.00 loops=1)
                    Hash Cond: ((LEAST(p1.nconst, p2.nconst)) = n1_name.nconst)
                    Buffers: shared hit=223
                    ->  GroupAggregate  (cost=8046.62..9352.49 rows=52235 width=72) (actual time=78.107..90.157 rows=23551.00 loops=1)
                          Group Key: (LEAST(p1.nconst, p2.nconst)), (GREATEST(p1.nconst, p2.nconst))
                          Buffers: shared hit=138
                          ->  Sort  (cost=8046.62..8177.20 rows=52235 width=74) (actual time=78.095..80.458 rows=56400.00 loops=1)
                                Sort Key: (LEAST(p1.nconst, p2.nconst)), (GREATEST(p1.nconst, p2.nconst)), p1.tconst
                                Sort Method: quicksort  Memory: 4066kB
                                Buffers: shared hit=138
                                ->  Hash Join  (cost=255.15..1629.29 rows=52235 width=74) (actual time=1.304..15.193 rows=56400.00 loops=1)
                                      Hash Cond: (p1.tconst = p2.tconst)
                                      Join Filter: (p2.nconst <> p1.nconst)
                                      Rows Removed by Join Filter: 8606
                                      Buffers: shared hit=138
                                      ->  Seq Scan on small_title_principals p1  (cost=0.00..162.07 rows=7446 width=20) (actual time=0.007..0.708 rows=7446.00 loops=1)
                                            Filter: (category = ANY ('{actor,actress}'::text[]))
                                            Buffers: shared hit=69
                                      ->  Hash  (cost=162.07..162.07 rows=7446 width=20) (actual time=1.291..1.292 rows=7446.00 loops=1)
                                            Buckets: 8192  Batches: 1  Memory Usage: 451kB
                                            Buffers: shared hit=69
                                            ->  Seq Scan on small_title_principals p2  (cost=0.00..162.07 rows=7446 width=20) (actual time=0.002..0.675 rows=7446.00 loops=1)
                                                  Filter: (category = ANY ('{actor,actress}'::text[]))
                                                  Buffers: shared hit=69
                    ->  Hash  (cost=148.49..148.49 rows=6349 width=24) (actual time=0.919..0.920 rows=6349.00 loops=1)
                          Buckets: 8192  Batches: 1  Memory Usage: 418kB
                          Buffers: shared hit=85
                          ->  Seq Scan on small_name_basics n1_name  (cost=0.00..148.49 rows=6349 width=24) (actual time=0.002..0.357 rows=6349.00 loops=1)
                                Buffers: shared hit=85
              ->  Hash  (cost=148.49..148.49 rows=6349 width=24) (actual time=0.782..0.782 rows=6349.00 loops=1)
                    Buckets: 8192  Batches: 1  Memory Usage: 418kB
                    Buffers: shared hit=85
                    ->  Seq Scan on small_name_basics n2_name  (cost=0.00..148.49 rows=6349 width=24) (actual time=0.009..0.352 rows=6349.00 loops=1)
                          Buffers: shared hit=85
Planning:
  Buffers: shared hit=12
Planning Time: 0.448 ms
Execution Time: 102.072 ms

4. 效率比对表

4.1. 效率比对表

Demo 场景 Cypher Execution (ms) SQL Execution (ms) 速度比 (Cypher/SQL) Cypher Buffers SQL Buffers Cypher temp 主要瓶颈（从 Plan 看）
3.1 两人是否合作过（count distinct） 1.388 0.983 1.41× shared hit=297 shared hit=27 read=4 否 Cypher：Seq Scan Person + NL；SQL：索引驱动 NL
3.2 两人共同作品清单（distinct + order） 1.426 0.556 2.56× shared hit=297 shared hit=63 read=1 否 Cypher：Unique+Sort，前置仍有 Seq Scan Person；SQL：索引 + NL，排序数据量小
3.3 某演员搭档 Top10（聚合+排序） 1.694 0.275 6.16× shared hit=360 shared hit=109 read=9 否 Cypher：GroupAggregate 前置仍 Seq Scan Person；SQL：索引 + GroupAgg + TopN
3.4 全局合作最多演员对 Top20（全局统计） 825.279 102.072 8.09× shared hit=974 + temp shared hit=308 是（temp r/w） Cypher：external merge + GroupAggregate + temp spill；SQL：HashJoin + GroupAgg（内存 sort 4MB，无 temp）

4.1. 关键补充表

Demo Cypher Planning (ms) SQL Planning (ms) 备注
3.1 0.535 3.234 SQL 计划时间更高，但执行更快（索引路径+CTE）
3.2 0.515 0.727 接近
3.3 0.978 0.762 接近
3.4 0.860 0.448 接近

4.2. 一句话结论

  1. 点查/小范围 join（3.1、3.2） ：Cypher 和 SQL 都是毫秒级，但 SQL 更快（约 1.4--2.6× ），原因是 关系表索引路径更成熟 ，而 Cypher 里 Person 的属性过滤走了 Seq Scan。
  2. 单源聚合 TopN（3.3） ：SQL 优势开始放大（约 6×），因为聚合/排序对 Postgres 执行器更友好。
  3. 全局统计（3.4） ：SQL 明显领先（约 8× ），AGE 发生 external merge + temp spill （temp read/written），这是性能分水岭。
相关推荐
DarkAthena6 小时前
【DuckDB】探索函数调用新范式：点操作符链式调用
数据库·sql·duckdb
大模型RAG和Agent技术实践6 小时前
SQL Agent从“黑盒“到“全透明“：基于LangGraph+Phoenix的可观测性实战指南
数据库·人工智能·sql·agent·langgraph
rchmin6 小时前
Redis BitMap介绍及使用场景示例
数据库·redis·缓存
Dxy12393102167 小时前
MySQL 日志全解析
数据库·mysql
思成Codes7 小时前
MySQL——最左前缀法则
数据库·mysql
杨云龙UP7 小时前
Windows环境下安装SQL Server 2016企业版+SP3补丁+SSMS连接操作手册_20251230
运维·服务器·数据库·sql·算法·sqlserver·哈希算法
雪碧聊技术7 小时前
基于Redis的分布式锁
数据库·redis·分布式
weixin_446260857 小时前
提升PostgreSQL编码效率的利器：pg-aiguide✨
数据库·postgresql
酸菜牛肉汤面7 小时前
24、SQL的生命周期？
数据库
热门推荐
01GitHub 镜像站点02jdk21下载、安装（Windows、Linux、macOS）03Linux下V2Ray安装配置指南04手把手教你通过Gemini3 pro 学生认证，白用一年，手慢无！052025 最新教程：注册并切换到美区 Apple ID06安娜的档案(Anna’s Archive) 镜像网站/国内最新可访问入口（持续更新）07Claude Code Skills 实用使用手册08KGG转MP3工具|非KGM文件|解密音频09UV安装并设置国内源10Arduino IDE更新 / 安装库时，提示Error: 4 DEADLINE超时的解决方法