ランキングのテーブルにインデックスを追加した

Seq Scanを避けられたらうれしい。

Before after

```SQL
-- ユーザーのRated Point Sumを取得する
EXPLAIN ANALYSE
SELECT
    point_sum
FROM
    rated_point_sum
WHERE
    LOWER(user_id) = 'hotate29';
```

張る前

```text
Gather  (cost=1000.00..6442.09 rows=1278 width=17) (actual time=42.603..43.865 rows=1 loops=1)
  Workers Planned: 1
  Workers Launched: 1
  ->  Parallel Seq Scan on rated_point_sum  (cost=0.00..5314.29 rows=752 width=17) (actual time=31.409..41.356 rows=0 loops=2)
        Filter: (lower((user_id)::text) = 'hotate29'::text)
        Rows Removed by Filter: 127800
Planning Time: 0.102 ms
Execution Time: 43.943 ms
```

張った後

```text
Bitmap Heap Scan on rated_point_sum  (cost=30.32..2414.86 rows=1278 width=17) (actual time=0.024..0.024 rows=1 loops=1)
  Recheck Cond: (lower((user_id)::text) = 'hotate29'::text)
  Heap Blocks: exact=1
  ->  Bitmap Index Scan on rated_point_sum_lower_idx  (cost=0.00..30.01 rows=1278 width=0) (actual time=0.021..0.021 rows=1 loops=1)
        Index Cond: (lower((user_id)::text) = 'hotate29'::text)
Planning Time: 0.168 ms
Execution Time: 0.039 ms
```

---

```SQL
-- C++でのAC数上位3人を取得
EXPLAIN ANALYSE
SELECT
    user_id, problem_count
FROM
    language_count
WHERE
    simplified_language = 'C++'
ORDER BY
    problem_count DESC, user_id ASC
OFFSET 0 LIMIT 3;
```

張る前

```text
Limit  (cost=6877.32..6877.33 rows=3 width=12) (actual time=5.972..5.973 rows=3 loops=1)
  ->  Sort  (cost=6877.32..6879.82 rows=1000 width=12) (actual time=5.971..5.971 rows=3 loops=1)
        Sort Key: problem_count DESC, user_id
        Sort Method: top-N heapsort  Memory: 25kB
        ->  Seq Scan on language_count  (cost=0.00..6864.40 rows=1000 width=12) (actual time=0.013..5.854 rows=1000 loops=1)
              Filter: ((simplified_language)::text = 'C++'::text)
              Rows Removed by Filter: 25432
Planning Time: 0.197 ms
Execution Time: 5.988 ms
```

張った後

```text
Limit  (cost=0.29..0.41 rows=3 width=12) (actual time=0.031..0.032 rows=3 loops=1)
  ->  Index Only Scan using language_count_simplified_language_problem_count_user_id_idx on language_count  (cost=0.29..41.79 rows=1000 width=12) (actual time=0.030..0.031 rows=3 loops=1)
        Index Cond: (simplified_language = 'C++'::text)
        Heap Fetches: 0
Planning Time: 0.280 ms
Execution Time: 0.045 ms
```

データはそれぞれ
<https://kenkoooo.com/atcoder/resources/sums.json>
と<https://kenkoooo.com/atcoder/resources/lang.json>
を使用。

Author: hotate29

config/database-definition.sql

@@ -79,6 +79,8 @@ CREATE TABLE accepted_count (
   problem_count INT           NOT NULL,
   PRIMARY KEY (user_id)
 );
+CREATE INDEX ON accepted_count (LOWER(user_id));
+CREATE INDEX ON accepted_count (problem_count DESC, user_id);
 
 DROP TABLE IF EXISTS points;
 CREATE TABLE points (
@@ -94,6 +96,8 @@ CREATE TABLE rated_point_sum (
   point_sum       BIGINT NOT NULL,
   PRIMARY KEY (user_id)
 );
+CREATE INDEX ON rated_point_sum (LOWER(user_id));
+CREATE INDEX ON rated_point_sum (point_sum DESC, user_id);
 
 DROP TABLE IF EXISTS language_count;
 CREATE TABLE language_count (
@@ -102,6 +106,8 @@ CREATE TABLE language_count (
   problem_count         INT NOT NULL,
   PRIMARY KEY (user_id, simplified_language)
 );
+CREATE INDEX ON language_count (LOWER(user_id));
+CREATE INDEX ON language_count (simplified_language, problem_count DESC, user_id);
 
 DROP TABLE IF EXISTS predicted_rating;
 CREATE TABLE predicted_rating (
@@ -124,6 +130,8 @@ CREATE TABLE max_streaks (
   streak                BIGINT NOT NULL,
   PRIMARY KEY (user_id)
 );
+CREATE INDEX ON max_streaks (LOWER(user_id));
+CREATE INDEX ON max_streaks (streak DESC, user_id);
 
 -- For internal services:
 DROP TABLE IF EXISTS internal_problem_list_items;