How can I determine the best place to add indexes in mysql? - mysql

Question

I've tried doing some searching, but I am having a bit of trouble getting a grasp on this.
Once I explain a query, how can I look at the information and determine the best place to add an index to speed up the query.
For example I have this query:
SELECT a.app_id, DATE_FORMAT(app_datetime, '%c/%d/%y %H:%i') as app_datetime, app_language
, if((select count(w.app_id) from li_app_cnc as w where w.app_id = a.app_id) > 0 , concat('CNC: ',(select cnc_note from li_app_cnc as w where w.app_id = a.app_id)),
if((select count(x.app_id) from li_app_dnc as x where x.app_id = a.app_id) > 0, concat('DNC: ', (select cancel_note from li_app_dnc as x where x.app_id = a.app_id)),
if((select count(y.app_id) from li_app_canceled as y where y.app_id = a.app_id) > 0, concat('Canceled: ', (select cancel_note from li_app_canceled as y where y.app_id = a.app_id)),
concat(h.emp_firstname, ' ', h.emp_lastname)))) as int_id, app_notes, app_facility, app_department
, app_requesting_person, app_service_provider
, cast(AES_DECRYPT(les_name, '$privatekey') as char) as les_name
, les_dob, cast(AES_DECRYPT(les_medicaid_id, '$privatekey') as char) as les_medicaid_id
, billing_total_time, billing_workorder_received, billing_admin_fee
, billing_notes, created_by, created_on, modified_by, modified_on, wo_entered_by
, t.cancel_code, t.cancel_note
, u.cnc_code, u.cnc_note
, v.cancel_code as dnc_code, v.cancel_note as dnc_note
FROM li_appointments.li_appointments as a
left Join orangehrm_li.hs_hr_employee as h on a.terp_id = h.employee_id
left Join li_appointments.li_app_canceled as t on t.app_id = a.app_id
left Join li_appointments.li_app_cnc as u on u.app_id = a.app_id
left Join li_appointments.li_app_dnc as v on v.app_id = a.app_id
where (app_client_id in (select account_number from li_appointments.li_client_access
where id = $userid) or created_by = '$username')
and date(app_datetime) = date(now())
and a.app_id not in (select f.app_id from li_app_dnc as f)
Which takes about 14 secconds
It is explained as:
1 PRIMARY a ALL 37539 Using where
1 PRIMARY h ALL 1036
1 PRIMARY t eq_ref PRIMARY PRIMARY 4 li_appointments.a.app_id 1
1 PRIMARY u eq_ref PRIMARY PRIMARY 4 li_appointments.a.app_id 1
1 PRIMARY v eq_ref PRIMARY PRIMARY 4 li_appointments.a.app_id 1
9 DEPENDENT SUBQUERY f unique_subquery PRIMARY PRIMARY 4 func 1 Using index
8 DEPENDENT SUBQUERY li_client_access ALL 72 Using where
7 DEPENDENT SUBQUERY y eq_ref PRIMARY PRIMARY 4 li_appointments.a.app_id 1
6 DEPENDENT SUBQUERY y eq_ref PRIMARY PRIMARY 4 li_appointments.a.app_id 1 Using index
5 DEPENDENT SUBQUERY x eq_ref PRIMARY PRIMARY 4 li_appointments.a.app_id 1
4 DEPENDENT SUBQUERY x eq_ref PRIMARY PRIMARY 4 li_appointments.a.app_id 1 Using index
3 DEPENDENT SUBQUERY w eq_ref PRIMARY PRIMARY 4 li_appointments.a.app_id 1
2 DEPENDENT SUBQUERY w eq_ref PRIMARY PRIMARY 4 li_appointments.a.app_id 1 Using index

Answer 1

Some thumb rules:
Index all the fields on which you will define WHERE conditions.
Index all the fields on which you will define relationships.
Index all the fields on which you will define grouping criteria.
Avoid indexing everything in your table. Think before creating the indexes.
(Although it depends on your specific needs, I personally avoid creating indexes on floating point columns)
Obviously, the above rules imply that you must index all the fields that work as primary key or as foreign key (if your tables are normalized, you already must have created the appropriate primary keys; if your tables are not normalized, then normalize them).