Performance Optimisation des requêtes

Question

J'ai cette requête mysql:

SELECT tbl.userid, tbl.name, tbl.courseid, tbl.fullname, IFNULL(tbl.instance, '-') instance, IFNULL(tbl.activityname, '-') activityname, 
    IFNULL(tbl.module, '-') module, IFNULL(tbl.attempt, '-') attempt, IFNULL(tbl.score, '-') score, 
    CASE tbl.module WHEN 'quiz' THEN (
     CASE WHEN (tbl.attempt IS NULL AND tbl.score IS NULL) THEN 'NOT YET' 
     WHEN (tbl.attempt IS NOT NULL AND tbl.score IS NULL) THEN 'ATTEMPT' 
     WHEN (tbl.attempt IS NOT NULL AND tbl.score IS NOT NULL) THEN 'FINISH' 
     END) 
    WHEN 'scorm' THEN (
     CASE WHEN tbl.attempt IS NULL THEN 'NOT YET' 
     ELSE (SELECT `status` FROM mdl_scorm_logs WHERE scormid = tbl.instance AND userid = tbl.userid) 
     END) 
    ELSE '-' 
    END `status` 
FROM (
    SELECT tbl1.userid, tbl1.name, tbl1.courseid, tbl1.fullname, tbl2.instance, 
     CASE tbl2.module WHEN 'quiz' THEN (SELECT `name` FROM mdl_quiz WHERE id = tbl2.instance AND course = tbl1.courseid) 
     WHEN 'scorm' THEN (SELECT `name` FROM mdl_scorm WHERE id = tbl2.instance AND course = tbl1.courseid) 
     END activityname 
     , tbl2.module, 
     CASE tbl2.module WHEN 'quiz' THEN (SELECT MAX(attempt) FROM mdl_quiz_attempts WHERE quiz = tbl2.instance AND userid = tbl1.userid) 
     WHEN 'scorm' THEN (SELECT MAX(attempt) FROM mdl_scorm_scoes_track WHERE scormid = tbl2.instance AND userid = tbl1.userid) 
     END attempt, 
     (SELECT CONCAT(TRUNCATE(ROUND(gg.finalgrade/gi.grademax * 100, 2), 2), ' %') 
     FROM mdl_grade_grades gg 
     JOIN mdl_grade_items gi ON gg.itemid = gi.id 
     WHERE gg.userid = tbl1.userid AND gi.courseid = tbl1.courseid AND gi.itemname = activityname AND gi.itemtype = 'mod' 
      AND gi.itemmodule = tbl2.module AND gi.iteminstance = tbl2.instance) score 
    FROM (
     SELECT u.id userid, CONCAT(u.firstname, ' ', u.lastname) `name`, c.id courseid, c.fullname 
     FROM mdl_user u 
     JOIN mdl_role_assignments ra ON u.id = ra.userid 
     JOIN mdl_context co ON ra.contextid = co.id 
     JOIN mdl_role r ON ra.roleid = r.id 
     JOIN mdl_course c ON co.instanceid = c.id 
     JOIN mdl_enrol e ON c.id = e.courseid 
     JOIN mdl_user_enrolments ue ON e.id = ue.enrolid AND u.id = ue.userid 
     WHERE r.archetype = 'student') tbl1 
    LEFT JOIN (
     SELECT cm.course, m.name module, cm.instance FROM mdl_course_modules cm 
     JOIN mdl_modules m ON cm.module = m.id 
     WHERE m.name IN ('quiz', 'scorm') 
     ) tbl2 
    ON tbl1.courseid = tbl2.course) tbl 
WHERE tbl.userid = '4' AND tbl.courseid = '8' 
GROUP BY tbl.courseid, tbl.instance 
ORDER BY tbl.name, tbl.fullname, tbl.activityname 

Résultat après que je l'utilise EXPLIQUER:

id select_type table type possible_keys key key_len ref rows Extra 
1 PRIMARY <derived3> ALL     122 Using where; Using temporary; Using filesort 
3 DERIVED <derived9> ALL     42 
3 DERIVED <derived10> ALL     23 
10 DERIVED m range PRIMARY,mdl_modu_nam_ix mdl_modu_nam_ix 62  2 Using where; Using index 
10 DERIVED cm ref mdl_courmodu_mod_ix mdl_courmodu_mod_ix 8 moodle.m.id 4 
9 DERIVED r ref PRIMARY,mdl_role_arc_ix mdl_role_arc_ix 92  1 Using where; Using index 
9 DERIVED ra ref mdl_roleassi_rol_ix,mdl_roleassi_con_ix,mdl_roleassi_use_ix mdl_roleassi_rol_ix 8 moodle.r.id 5 
9 DERIVED u eq_ref PRIMARY PRIMARY 8 moodle.ra.userid 1 
9 DERIVED co eq_ref PRIMARY,mdl_cont_ins_ix PRIMARY 8 moodle.ra.contextid 1 
9 DERIVED e ref PRIMARY,mdl_enro_cou_ix mdl_enro_cou_ix 8 moodle.co.instanceid 1 Using index 
9 DERIVED c eq_ref PRIMARY PRIMARY 8 moodle.co.instanceid 1 
9 DERIVED ue eq_ref mdl_userenro_enruse_uix,mdl_userenro_enr_ix,mdl_userenro_use_ix mdl_userenro_enruse_uix 16 moodle.e.id,moodle.ra.userid 1 Using index 
8 DEPENDENT SUBQUERY gi ref PRIMARY,mdl_graditem_itenee_ix,mdl_graditem_cou_ix mdl_graditem_cou_ix 9 tbl1.courseid 1 Using where 
8 DEPENDENT SUBQUERY gg eq_ref mdl_gradgrad_useite_uix,mdl_gradgrad_ite_ix,mdl_gradgrad_use_ix mdl_gradgrad_useite_uix 16 tbl1.userid,moodle.gi.id 1 Using where 
7 DEPENDENT SUBQUERY mdl_scorm_scoes_track ref mdl_scorscoetrac_usescosco_uix,mdl_scorscoetrac_use_ix,mdl_scorscoetrac_sco_ix mdl_scorscoetrac_usescosco_uix 16 tbl1.userid,tbl2.instance 6 Using where; Using index 
6 DEPENDENT SUBQUERY mdl_quiz_attempts ref mdl_quizatte_use_ix,mdl_quizatte_qui_ix mdl_quizatte_qui_ix 8 tbl2.instance 2 Using where 
5 DEPENDENT SUBQUERY mdl_scorm eq_ref PRIMARY,mdl_scor_cou_ix PRIMARY 8 tbl2.instance 1 Using where 
4 DEPENDENT SUBQUERY mdl_quiz eq_ref PRIMARY,mdl_quiz_cou_ix PRIMARY 8 tbl2.instance 1 Using where 
2 DEPENDENT SUBQUERY mdl_scorm_logs eq_ref PRIMARY PRIMARY 16 func,func 1 Using where 

Toute idée de l'optimiser? Merci.

Answer 1

Oui. Convertissez les sous-requêtes corrélées en jointures externes. Vous devriez probablement être en mesure de prendre un bon départ en interrogeant ces termes sur ce site et google.

Idem pour les "tables dérivées", j'imagine.

La syntaxe

SELECT ... FROM (SELECT FROM ...) peut être problématique. Je ne comprends peut-être pas le contexte, mais la dénormalisation me semble être un choix extrêmement prématuré.

Answer 2

OMG. La dénormalisation est le seul moyen de gérer cette situation. Ok, vous allez optimiser cette requête. Que ferez-vous la prochaine fois, quand vous devrez joindre 15 tables en une seule requête?

Votre requête ressemble à un rapport. Vous devriez concevoir votre base de données pour la création de rapports. Consultez les liens suivants, cela peut vous être utile.

http://en.wikipedia.org/wiki/Denormalization

http://en.wikipedia.org/wiki/Star_schema