List:General Discussion« Previous MessageNext Message »
From:Chris W Date:March 30 2010 7:40pm
Subject:Re: Table Length Question...
View as plain text  
Your first table layout is horrible, the second one is only marginally 
better.  You should read up on database normalization.

I have no idea what id, id2 and type are but since they seem like they 
are the same for every 'f' and every day, I am pretty sure they all 
relate directly to the user so those should all go in the first table.  
I will assume 'name' is the field you use to identify users.    So your 
second table, let's call it dftable, should have the following fields.

ID --- generic auto increment id for each record.
name -- foreign key to link this table back to your user table.
Fnumber -- this would be an int or maybe a set that would be set to 
1,2,3, or 4
Date -- this would be the just like it says the date and you only have 
one table not one for every month.
DFValue -- not sure what type of data this is.

you probably want a unique key on name, FNumber and Date.  They could 
also be your primary key instead of using ID, but I find it easier to 
code my applications and deal with changes in the long run if you use a 
single field auto increment value for your primary key.


a simple join with a group concat will give you back your first output 
example. 

However for you example where you want d02f01 the query would simply be
Note: since in my table I am using date, instead of just 02 we need the 
whole date so I will assume it is Mar, 02 of 2010

SELECT u.`name`, u.`id`, u.`id2`, u.`type`, d.`DFValue`
FROM `user` u
JOIN `dftable` d USING(`name`)
WHERE `FNumber` = '1' AND `Date` = '2010-03-02'

If id, id2 and or type change over time and you need to keep track of 
that you could add those fields in this table.

Chris W

Steven Staples wrote:
> Hi there,
>
> I currently store some information about a users daily habits in a table.
> The table has 4 fields per day, and another 4 fields as the keys.  This
> table, depending on the month, can be from (4 keys + (28 days * 4 fields per
> day)) fields, to (4 keys + (31 days * 4 fields per day)) fields long... 
>
> The table layout is like such:
> +-----+---+----+-----+------+------+------+------+------+------+------------
> --
> |name |id |id2 |type |d01f1 |d01f2 |d01f3 |d01f4 |d02f1 |d02f2 |.....and so
> on
> +-----+---+----+-----+------+------+------+------+------+------+------------
> --
>
> Performance wise, would it be better to have it laid out in a manner such as
> +-----+---+----+-----+----+---+---+---+---+
> |name |id |id2 |type |day |f1 |f2 |f3 |f4 |
> +-----+---+----+-----+----+---+---+---+---+
> So that each row, contains a single days details, rather than have a single
> row, contain the entire months details?
>
> Also, when i would do a select, if i wanted say d02f1, would it load the
> entire row first, and then just give me that field?
> -Select `d02f01` from `mytable` where [where clause]
> Or would it jsut load that field... 
>
> Does these questions make sense?  (they do in my head)
>
> Steven Staples
>
>
>
>   
Thread
Table Length Question...Steven Staples30 Mar
  • RE: Table Length Question...Gavin Towey30 Mar
  • Re: Table Length Question...Chris W30 Mar