这是我在python和sqlalchemy中完成的查询，但我不认为这是sqlalchemy很慢，只是我不知道如何快速查询。查询大约需要8秒，并返回45,000个结果。
games = s.query(Box_Score, Game).join(Game, Box_Score.espn_game_id==Game.espn_game_id)
.filter(Game.total!=999)
.filter(Game.a_line!=999)\
.order_by(Box_Score.date.desc()).all()
这是常规SQL中的查询
SELECT box_scores.date AS box_scores_date, box_scores.id AS box_scores_id, box_scores.player_name AS box_scores_player_name, box_scores.team_name AS box_scores_team_name, box_scores.espn_player_id AS box_scores_espn_player_id, box_scores.espn_game_id AS box_scores_espn_game_id, box_scores.pass_attempt AS box_scores_pass_attempt, box_scores.pass_made AS box_scores_pass_made, box_scores.pass_yards AS box_scores_pass_yards, box_scores.pass_td AS box_scores_pass_td, box_scores.pass_int AS box_scores_pass_int, box_scores.pass_longest AS box_scores_pass_longest, box_scores.run_carry AS box_scores_run_carry, box_scores.run_yards AS box_scores_run_yards, box_scores.run_td AS box_scores_run_td, box_scores.run_longest AS box_scores_run_longest, box_scores.reception AS box_scores_reception, box_scores.reception_yards AS box_scores_reception_yards, box_scores.reception_td AS box_scores_reception_td, box_scores.reception_longest AS box_scores_reception_longest, box_scores.interception_lost AS box_scores_interception_lost, box_scores.interception_won AS box_scores_interception_won, box_scores.fg_attempt AS box_scores_fg_attempt, box_scores.fg_made AS box_scores_fg_made, box_scores.fg_longest AS box_scores_fg_longest, box_scores.punt AS box_scores_punt, box_scores.first_down AS box_scores_first_down, box_scores.penalty AS box_scores_penalty, box_scores.penalty_yards AS box_scores_penalty_yards, box_scores.fumbles AS box_scores_fumbles, box_scores.possession AS box_scores_possession, games.id AS games_id, games.espn_game_id AS games_espn_game_id, games.date AS games_date, games.status AS games_status, games.time AS games_time, games.season AS games_season, games.h_name AS games_h_name, games.a_name AS games_a_name, games.league AS games_league, games.h_q1 AS games_h_q1, games.h_q2 AS games_h_q2, games.h_q3 AS games_h_q3, games.h_q4 AS games_h_q4, games.h_ot AS games_h_ot, games.h_score AS games_h_score, games.a_q1 AS games_a_q1, games.a_q2 AS games_a_q2, games.a_q3 AS games_a_q3, games.a_q4 AS games_a_q4, games.a_ot AS games_a_ot, games.a_score AS games_a_score, games.possession_h2 AS games_possession_h2, games.d_yards_h1 AS games_d_yards_h1, games.f_yards_h1 AS games_f_yards_h1, games.h_ml AS games_h_ml, games.a_ml AS games_a_ml, games.h_h1_ml AS games_h_h1_ml, games.a_h1_ml AS games_a_h1_ml, games.h_q1_ml AS games_h_q1_ml, games.a_q1_ml AS games_a_q1_ml, games.h_h2_ml AS games_h_h2_ml, games.a_h2_ml AS games_a_h2_ml, games.h_line AS games_h_line, games.h_price AS games_h_price, games.a_line AS games_a_line, games.a_price AS games_a_price, games.h_open_line AS games_h_open_line, games.h_open_price AS games_h_open_price, games.a_open_line AS games_a_open_line, games.a_open_price AS games_a_open_price, games.h_h1_line AS games_h_h1_line, games.h_h1_price AS games_h_h1_price, games.a_h1_line AS games_a_h1_line, games.a_h1_price AS games_a_h1_price, games.h_q1_line AS games_h_q1_line, games.h_q1_price AS games_h_q1_price, games.a_q1_line AS games_a_q1_line, games.a_q1_price AS games_a_q1_price, games.h_h2_line AS games_h_h2_line, games.h_h2_price AS games_h_h2_price, games.a_h2_line AS games_a_h2_line, games.a_h2_price AS games_a_h2_price, games.total AS games_total, games.o_price AS games_o_price, games.u_price AS games_u_price, games.total_h1 AS games_total_h1, games.o_h1_price AS games_o_h1_price, games.u_h1_price AS games_u_h1_price, games.total_q1 AS games_total_q1, games.o_q1_price AS games_o_q1_price, games.u_q1_price AS games_u_q1_price, games.total_h2 AS games_total_h2, games.o_h2_price AS games_o_h2_price, games.u_h2_price AS games_u_h2_price
FROM box_scores JOIN games ON box_scores.espn_game_id = games.espn_game_id
WHERE games.total != :total_1 AND games.a_line != :a_line_1 ORDER BY box_scores.date DESC
此查询需要3秒以上，并返回55,000个结果。
box_scores = s.query(Box_Score).all()
我一定是做错了。我知道人们经常使用包含数百万条目的数据库，所以我不明白为什么选择50,000行应该是一个大问题。我也尝试加入Box_Score而不是Game，并取出order_by（）部分，并且都没有加快性能。
UPDATE;我想知道什么是碎片来回答下面的问题。不明白它但我确实运行了命令PRAGMA page_count - ＆gt; 64,785，这似乎不是一个大数字。我也做了sqlite nfl.db“VACUUM”;然后再次运行查询，没有性能改进。