Shantanu's Blog

This might be a simple problem but i am having some trouble with it. On the 4th line, SUM(d.event_cost) as 'Events Cost' the value of this filed is taken, for whatever reason and multiplied by the count of the b.prospect_id. Not sure why this happens. The result i am trying to get is just the sum of alll d.event_cost. If i use SUM(DISTINCT d.event_cost) as 'Events Cost' it gives me the right result but then i run into trouble if i have the exact same amount of d.event cost in 2 or more fields.
I also tried to do Sum(d.event_cost)/count(b.prospect_id) but that would only work if i was retreiving one d.event_cost.

Thanks for the help in advance

SELECT   Concat_ws(' ',a.advisor_fn,a.advisor_ln)  AS 'Advisor',

         a.advisor_ao                              AS 'AO',

         a.advisor_manager                         AS 'Manager',

         SUM(d.event_cost)                         AS 'Events Cost',

         COUNT(b.prospect_id)                      AS 'Prospects',

         COUNT(g.referral_id)                      AS 'Referrals',

         COUNT(f.fepsleads_id)                     AS 'Feps Leads',

         SUM(b.pmm_set)                            AS 'PMMs Set',

         SUM(b.pmm_seen)                           AS 'PMMs Seen',

         SUM(b.prospect_client)                    AS 'CAs',

         SUM(b.gdc)                                AS 'GDC'

FROM     advisor_tb a,

         marketing_location_tb c,

         marketing_location_event_tb d,

         marketing_event_prospect_tb b

         LEFT JOIN marketing_event_fepslead_tb f

           ON b.prospect_id = f.prospect_id

         LEFT JOIN marketing_event_referral_tb g

           ON b.prospect_id = g.prospect_id

WHERE    a.advisor_id = c.advisor_id

         AND c.location_id = d.location_id

         AND d.event_id = b.event_id

         AND a.advisor_ao != 10

         AND a.advisor_ao != 260

         AND a.advisor_ao != 'none'

         AND a.advisor_ao != 44

         AND d.event_date BETWEEN '2007-01-01'

                                  AND '2007-07-07'

GROUP BY a.advisor_id

ORDER BY a.advisor_fn


http://forums.devshed.com/mysql-help-4/a-simple-query-problem-somewhat-urgent-463669.html

The answer is to inner join a derived table to the group of 3 inner joined tables like this...

SELECT   Concat_ws(' ',a.advisor_fn,a.advisor_ln)  AS 'Advisor',

         a.advisor_ao                              AS 'AO',

         a.advisor_manager                         AS 'Manager',

         SUM(d.event_cost)                         AS 'Events Cost',

         SUM(b.count_prospect_id)                  AS 'Prospects',

         SUM(b.count_referral_id)                  AS 'Referrals',

         SUM(b.count_fepsleads_id)                 AS 'Feps Leads',

         SUM(b.pmm_set)                            AS 'PMMs Set',

         SUM(b.pmm_seen)                           AS 'PMMs Seen',

         SUM(b.prospect_client)                    AS 'CAs',

         SUM(b.gdc)                                AS 'GDC'

FROM     advisor_tb a,

         marketing_location_tb c,

         marketing_location_event_tb d,

         (SELECT   h.event_id,

                   h.pmm_set,

                   h.pmm_seen,

                   h.prospect_client,

                   h.gdc,

                   COUNT(h.prospect_id)   AS count_prospect_id,

                   COUNT(g.referral_id)   AS count_referral_id,

                   COUNT(f.fepsleads_id)  AS count_fepsleads_id

          FROM     marketing_event_prospect_tb h

                   LEFT JOIN marketing_event_fepslead_tb f

                     ON b.prospect_id = f.prospect_id

                   LEFT JOIN marketing_event_referral_tb g

                     ON b.prospect_id = g.prospect_id

          GROUP BY h.event_id) b

WHERE    a.advisor_id = c.advisor_id

         AND c.location_id = d.location_id

         AND d.event_id = b.event_id

         AND a.advisor_ao != 10

         AND a.advisor_ao != 260

         AND a.advisor_ao != 'none'

         AND a.advisor_ao != 44

         AND d.event_date BETWEEN '2007-01-01'

                                  AND '2007-07-07'

GROUP BY a.advisor_id

ORDER BY a.advisor_fn

Labels: mysql, mysql case study

Improve performance

Daffy Duck has asked a question about Mysql LOAD DATA takes too much time to insert records...

I am trying to load a large file into mysql, it has over 60 million records. So I split it into 60 files of 1m records and im using this to load the files..

LOAD DATA LOCAL INFILE 'file.name' INTO TABLE dbname;

However its very slow taking upto an hour to insert 1m records and server loads go sky high. How could I optimize mysql, is there anything I can put in my.cnf ?

http://forums.devshed.com/mysql-help-4/mysql-load-data-loads-server-476564.html


There were 3 options suggested by the experts.

1) Disable keys before importing data.
ALTER TABLE tbl_name DISABLE KEYS;
LOAD DATA ....
ALTER TABLE tbl_name ENABLE KEYS;

2) Change the key_butter_size value in my.cnf file.
The rule is 25% of available RAM, _but_ if you are running apache etc on the same server then you should check how much memory is used when mysql is not started and set the key_buffer_size to 25% of that.

[mysqld]
key_buffer_size=xxM

3) Use sed command to remove duplicates and then import the data.
sed -n 'G; s/\n/&&/; /^\([ -~]*\n\).*\n\1/d; s/\n//; h; P' file1.txt > file2.txt

Labels: mysql, mysql case study

Table join with COALESCE

I'm trying to write a query that will compare a column from two tables and return the records that they do not have in common.
For example:

TableA
----------------------------------
id | Item
1 | Water
2 | Milk
3 | Bread
4 | Gum
5 | Butter
6 | Flour
7 | Juice
8 | Cake
9 | Cookies
10 | Hummus

TableB
----------------------------------
bid | id | count
300 | 2 | 1
301 | 4 | 1
302 | 5 | 1
303 | 7 | 2
304 | 9 | 2

I want to write a query that will select the records from TableA whose ids are not found in table b and that do not have the count value of 1. So, the query will ultimately produce:

1 | Water
3 | Bread
6 | Flour
7 | Juice <-------- returned because has a count of 2
8 | Cake
9 | Cookies <-------- returned because has a count of 2
10 | Hummus

I tried:
SELECT a.id FROM TableA a LEFT JOIN TableB b ON a.id = b.id WHERE b.count > 1

but that didn't yield what I thought it would.

http://forums.devshed.com/mysql-help-4/complicated-query-476616.html

Labels: mysql, mysql case study

Group by 20 minutes

I know how to group a result set by a specific period eg

select sum(value),date_format(date,'%H') from table group by date_format(date,'%H');

this groups my results by hours.

My question is, is there a way that we can modify this query to show the results but grouped by 20min periods. I have tried

select sum(value),date_format(date,'%H:%i') from table group by date_format(date,'%H:%i'/3);

but this does not work

Any thoughts or suggestions would be greatly appreciated.

http://forums.mysql.com/read.php?10,174757,175012#msg-175012

Labels: mysql, mysql case study

Count child/parent in same table

Is it possible to do a count in the same table.

I will explain it better.

I have a categorie table

catID
catParentID
catName

id1: parentID:0 >> europe
id2: parentID:0 >> asia
id3: parentID:1 >> france
id4: parentID:1 >> germany
id5: parentID:2 >> china
id6: parentID:3 >> paris
id7: parentID:3 >> bordeaux
id8: parentID:3 >> nice
id9: parentID:4 >> berlin
id10: parentID:4 >> munich
id11: parentID:5 >> bejing

So let's say I want a query which displays the amount of cities in the countries:

germany: 2
france: 3
china: 1

Or the total amount of cities in a part of the world

asia: 1
europe: 5

Can it be done within one query?

http://forums.devshed.com/mysql-help-4/count-child-parent-in-same-table-469604.html

Labels: mysql case study

1) How do I backup my tables using UNIX file system?

.frm holds the definition of a table. With MyISAM, the .MYD and MYI files hold the data and index information for the table.
If you are using InnoDB or another non-MyISAM engine for a table, you will not have .MYD and .MYI files. Therefore backing up the MyISAM tables is easy and fast if you have access to these files. You can create table of the same structure without keys, load data into it to get correct .MYD (data file). Create table with all keys defined and copy over .frm and .MYI files from it, followed by FLUSH TABLES. Now you can use REPAIR TABLE to rebuild all keys by sort, including UNIQUE keys.

2) What is the difference in MyISAM and InnoDB tables?

MyISAM InnoDB
Backup files Can't backup files
Doesn't support crash recovery Does support crash recovery
Table level locking Row level locking
fast full table scan Slow full tablescan
Full text index No full text index possible
No Foreign Key support Foreign Key support
No Transaction support Supports transactions


InnoDB is a storage engine, and so is MyISAM. When you create a table, you specify one of the types. When you add data and indexes to that table, the type of table determines the storage engine used. InnoDB does row level locking. This means that when you are updating a row, only that one row gets locked (which means that another connection to the database cannot modify that row). MyISAM locks the entire table. Only one connection / session at a time can update / insert / delete.

InnoDB uses the concept of a tablespace; MyISAM doesn't. A tablespace is where you store your data, and is made up of datafiles. You don't know where your data is stored in those data files. When you create a table in MyISAM, it crates a file of the same name as your table. Some queries are really bad on InnoDB compared to MyISAM, the most notorious is probably SELECT COUNT(*) A full table scan is also much faster on MyISAM than on InnoDB.

InnoDB doesn't support full text indexes on text columns so if you need to search text fields you may want to use MyISAM.

In addition to foreign keys, InnoDB offers transaction support, which is absolutely critical when dealing with larger applications. Sped does suffer though because all this Foreign Key / Transaction stuff takes lots of overhead. With InnoDB it becomes particularly important that you use good keys.

Inserting 50,000 records is something MyISAM is very comfortable with. In some cases it is faster to drop the indexes, insert the records and recreate the indexes (especially with full text indexes).

Use InnoDB in a situation where there are lots of inserts, updates and selects. Tests show that MyISAM is a lot faster when there are very few records (<50k), but the average execution time for a query increases almost linearly with the number of records while InnoDB shows almost constant query execution times for very small and large tables.

MyISAM does not support crash recovery while InnoDB does.


3) What should be the correct column definition?
VARCHAR(25) and VARCHAR(200) are practically the same for the purpose of storing data on the disk. If the user types only 5 characters in either columns, the bytes used will be the same. Select the value liberally but remember that VARCHAR(250) for
all text columns is not good idea either because it will not allow you to create composite indexes.

4) How does Primary and other keys affect the programming logic?

Composite primary indexes are great to avoid duplicate entries. But sometimes it can be a problem. For e.g. take a look at the following tables PRIMARY KEY definition.

CREATE TABLE `student` (
`enroll` varchar(13) NOT NULL default '',
`other` int(2) NOT NULL default '0',
`year` int(4) NOT NULL default '0',
`month` tinyint(2) unsigned default NULL,
`mode_read` int(2) NOT NULL default '0',
`mode_direct` int(2) NOT NULL default '0',
`sent_flag` int(11) NOT NULL default '0',
`system_date` datetime NOT NULL default '0000-00-00 00:00:00',
PRIMARY KEY (`enroll_no`,`other`,`year`)
) ENGINE=InnoDB DEFAULT CHARSET=latin1 |

Adding year as a part of primary key but not adding month column in it means there can be only 1 prospectus entry per year for each student. Now we can not use replace command that will insert a record if it does not exist in the primary key definition or update if it already does.

Labels: mysql FAQ