MySQL临时文件排序问题

Question

From my research, I'm not even sure it is possible to avoid the temporary / filesort with the following query due to the group by and order by. I'm also not one to ask without doing a ton of research on my own. But if anyone can help figure out how to avoid the filesort or point me in the correct direction - even if its completely re-writing the query or suggesting something code side, it would be much appreciated. I've completely hit a wall trying to figure this one out. There's a link to the output of the explain at the bottom.

explain select CONCAT(scfs.name, ' ',scfs.state,' ',scfs.zip_code) as scfname, scfs.zip_code, IF(date(s.scan_datetime) <= date(NOW()),date(s.scan_datetime),null) as scandate, count(*) as total,
                    sum(case when s.delivery_status = 1 then 1 else 0 end) as final

                     from order_addresses oa
                         left join pkg_data_unique s
                            on oa.trace_code = s.pkg_trace_code
                         inner join scf_zip_codes z
                            on SUBSTR(oa.zip,1,3) = z.zip_code
                         inner join scfs scfs
                            on z.scf_zip_code = scfs.zip_code



                    where oa.order_id = 160387

                    group by 1,2,3
                    order by scfs.zip_code, scandate

CREATE ORDER ADDRESSES

CREATE TABLE order_addresses ( id int(10) unsigned NOT NULL AUTO_INCREMENT, order_id int(11) NOT NULL, name varchar(100) COLLATE utf8_unicode_ci NOT NULL, address varchar(100) COLLATE utf8_unicode_ci NOT NULL, address2 varchar(100) COLLATE utf8_unicode_ci NOT NULL, city varchar(50) COLLATE utf8_unicode_ci NOT NULL, state varchar(15) COLLATE utf8_unicode_ci NOT NULL, zip char(5) COLLATE utf8_unicode_ci NOT NULL, zip4 int(11) NOT NULL, imb_digits char(31) COLLATE utf8_unicode_ci NOT NULL, trace_code char(20) COLLATE utf8_unicode_ci NOT NULL, created_at timestamp NOT NULL DEFAULT '0000-00-00 00:00:00', updated_at timestamp NOT NULL DEFAULT '0000-00-00 00:00:00', deleted_at timestamp NULL DEFAULT NULL, PRIMARY KEY (id), UNIQUE KEY order_address_unique (order_id,imb_digits,name,address,address2,city,state,zip), KEY order_addresses_order_id_index (order_id), KEY order_addresses_name_index (name), KEY order_addresses_address_index (address), KEY order_addresses_city_index (city), KEY order_addresses_state_index (state), KEY order_addresses_zip_index (zip), KEY order_addresses_imb_digits_index (imb_digits), KEY order_addresses_trace_code_index (trace_code), KEY order_id_trace_code (order_id,trace_code) USING BTREE ) ENGINE=InnoDB AUTO_INCREMENT=487714542 DEFAULT CHARSET=utf8 COLLATE=utf8_unicode_ci ROW_FORMAT=COMPACT

PKG DATA UNIQUE TABLE

CREATE TABLE pkg_data_unique ( id int(10) unsigned NOT NULL AUTO_INCREMENT, zip char(5) COLLATE utf8_unicode_ci NOT NULL, opcode int(11) NOT NULL, pkg_trace_code char(20) COLLATE utf8_unicode_ci NOT NULL, scan_datetime datetime NOT NULL, original_scan_datetime datetime NOT NULL, delivery_status int(11) NOT NULL, created_at timestamp NOT NULL DEFAULT '0000-00-00 00:00:00', updated_at timestamp NOT NULL DEFAULT '0000-00-00 00:00:00', PRIMARY KEY (id), UNIQUE KEY pkg_data_unique_pkg_trace_code_index (pkg_trace_code) USING BTREE, KEY pkg_data_unique_zip_index (zip), KEY pkg_data_unique_opcode_index (opcode), KEY pkg_data_unique_scan_datetime_index (scan_datetime), KEY pkg_data_unique_delivery_status_index (delivery_status), KEY pkg_data_unique_original_scan_datetime (original_scan_datetime) ) ENGINE=InnoDB AUTO_INCREMENT=490667214 DEFAULT CHARSET=utf8 COLLATE=utf8_unicode_ci ROW_FORMAT=COMPACT

SCF ZIP CODES

CREATE TABLE scf_zip_codes ( id int(10) unsigned NOT NULL AUTO_INCREMENT, scf_zip_code varchar(3) COLLATE utf8_unicode_ci NOT NULL, zip_code varchar(3) COLLATE utf8_unicode_ci NOT NULL, created_at timestamp NOT NULL DEFAULT '0000-00-00 00:00:00', updated_at timestamp NOT NULL DEFAULT '0000-00-00 00:00:00', PRIMARY KEY (id), KEY scf_zip_codes_scf_zip_code_index (scf_zip_code), KEY scf_zip_codes_zip_code_index (zip_code) ) ENGINE=InnoDB AUTO_INCREMENT=916 DEFAULT CHARSET=utf8 COLLATE=utf8_unicode_ci ROW_FORMAT=COMPACT

SCFS

CREATE TABLE scfs ( id int(10) unsigned NOT NULL AUTO_INCREMENT, label_id bigint(20) unsigned NOT NULL, zip_code varchar(5) COLLATE utf8_unicode_ci NOT NULL, name varchar(255) COLLATE utf8_unicode_ci NOT NULL, state varchar(255) COLLATE utf8_unicode_ci NOT NULL, locale_key varchar(255) COLLATE utf8_unicode_ci NOT NULL, created_at timestamp NOT NULL DEFAULT '0000-00-00 00:00:00', updated_at timestamp NOT NULL DEFAULT '0000-00-00 00:00:00', PRIMARY KEY (id), UNIQUE KEY scfs_label_id_unique (label_id), UNIQUE KEY scfs_zip_code_unique (zip_code), KEY scfs_name_index (name), KEY scfs_state_index (state) ) ENGINE=InnoDB AUTO_INCREMENT=196 DEFAULT CHARSET=utf8 COLLATE=utf8_unicode_ci ROW_FORMAT=COMPACT

Image of Explain Output

Answer 1

figure out how to avoid the filesort or point me in the correct direction

The alternative(s) involve changing the intent of your query.

    group by  1,2,3
    order by  scfs.zip_code, scandate

The GROUP BY may involve a filesort an temp table.

The ORDER BY will certainly involve (another) filesort and temp table because it does not match the GROUP BY . No, shuffling to GROUP BY 2,1,3 (since col 2 is the zip code) won't help.

If you can live with this, GROUP BY 2,1,3 ORDER BY 2,1,3 , then you can get rid of one filesort and have approximately the same output.

Is your query "slow"? 'Filesort' is not the main performance killer.

There are a few things that can speed up the query some:

• Zipcodes (and some other things) don't need utf8; change the column to CHARACTER SET ascii .
• Secondary keys, when used, need to drill down the PRIMARY KEY BTree to find the row. Consider which, if any, secondary indexes could be promoted to PRIMARY KEY .
• date(s.scan_datetime) <= date(NOW()) --> s.scan_datetime <= CURDATE()
• sum(case when s.delivery_status = 1 then 1 else 0 end) => sum(case when s.delivery_status = 1) since a boolean expression evaluates to 1 or 0.
• INDEX(a), INDEX(a,b) --> the first can be DROPped without loss of functionality.
• scfs has 3 unique keys; id seems useless. Promote zipcode to be the PK and make it CHAR(5) CHARACTER SET ascii . It will be 5 bytes instead of id's 4 bytes.
• Don't blindly use (255) ; pick a smaller (but safe) value.