SQL select all items of an owner from an item-to-owner table

Question

I have a table bike_to_owner. I would like to select current items owned by a specific user.

Table structure is

CREATE TABLE IF NOT EXISTS `bike_to_owner` (
  `bike_id` int(10) unsigned NOT NULL,
  `user_id` int(10) unsigned NOT NULL,
  `last_change_date` date NOT NULL,
  PRIMARY KEY  (`bike_id`,`user_id`,`last_change_date`)
) ENGINE=MyISAM DEFAULT CHARSET=utf8 COLLATE=utf8_unicode_ci;

In the profile page of the user I would like to display all his/her current possessions.

I wrote this statement:

SELECT `bike_id`,`user_id`,max(last_change_date) FROM `bike_to_owner` 
WHERE `user_id` = 3 group by `last_change_date`

but i'm not quite sure it works correctly in all cases. Can you please verify this is correct and if not suggest me something better. Using php/mysql.

Thanks in advance!

Answer 1

Almost there, but you should be grouping by bike_id and user_id rather than last change date:

SELECT `bike_id`,`user_id`,max(last_change_date) FROM `bike_to_owner` 
WHERE `user_id` = 3 group by `bike_id`, `user_id`

ETA:

As an aside, I'm curious why you have the last change date in that table as part of the key. Would it be possible for a bike to be associated with a user multiple times with different change dates? If so, why? I would think that a bike is either associated with a user or not associated with a user - there's really nothing that could "change" there aside from the relationship being added or deleted. Unless there are other fields on that table that you aren't showing?

Answer 2

In light of your comments, I'm going to just give a new answer here.

Since you could have a user taking ownership of a bike, then giving it up, then getting it back again and you want to be able to track that history, I would suggest a slight change to your schema:

CREATE TABLE IF NOT EXISTS `bike_to_owner` (
  `bike_id` int(10) unsigned NOT NULL,
  `user_id` int(10) unsigned NOT NULL,
  `acquired_date` date NOT NULL,
  `sold_date` date NULL,
  PRIMARY KEY  (`bike_id`,`user_id`,`acquired_date`)
) ENGINE=MyISAM DEFAULT CHARSET=utf8 COLLATE=utf8_unicode_ci;

Whenever a bike changes hands, you would update the sold_date for the previous owner to the current date and create a new record for the new owner with a sold_date of NULL and an acquired_date of the current date.

Then your query to determine current ownership would look like this:

SELECT `bike_id`,`user_id` FROM `bike_to_owner` 
WHERE `user_id` = 3 AND sold_date IS NULL

Answer 3

Here is the statement that works:

SELECT t1.*
FROM bike_to_owner t1
LEFT JOIN bike_to_owner AS t2 ON t1.bike_id = t2.bike_id
AND t1.last_change_date < t2.last_change_date
WHERE t2.last_change_date IS NULL
AND t1.user_id = 3

Answer 4

I think a lot depends on what information you are trying to store and be able to retrieve from the database. Attempting to second-guess this, I'd propose the following set of tables (and note that only the one column in any of these tables allow nulls):

--  User  -----
user_id  (primary key)
name

--  Bike  ------
bike_id  (primary key)
description
current_owner  nullable  (foreign key to User)

This presumes that “bike ownership” is a criticial and frequently desired piece of information, and is worth adding denormalized data to the structure to make finding a bike's current user or a user's current bikes trivial. (A key reason for adding denormalized data is to simplify data retrieval and/or improve performance. Skipping working out sales history on every query you make does both.) If duration of ownership is important, a column for “purchased_on” could be added, or you delve into the transaction tables.

To track when bikes were purchased:

--  Bike_Purchase   ------
user_id  (foreign key to User)
bike_id  (foreign key to Bike)
purchased_on
(primary key on all three columns)

would track every time a user purchased a bike. If instead you'd rather track when a bike was sold, you could have

--  Bike_Sale   ------
user_id  (foreign key to User)
bike_id  (foreign key to Bike)
sold_on
(primary key on all three columns)

To track both purchases and sales, include either both tables or a simple conglomerate:

--  Bike_Transaction   ------
user_id  (foreign key to User)
bike_id  (foreign key to Bike)
transaction_type  **
transaction_date
(primary key on all four columns)

This allows you to accurately track purchases and sales for your users, and disregard transactions by non-users. Sold a bike? Make an entry. Don't know or care who they sold it to? Don't make an entry.

Whichever time-tracking table you use, determining duration of ownership is as simple as joining on (user_id, bike_id) and getting max(date). Piecing together the “ownership chain” of a given bike, or what bikes a user owned and when/for how long would be tricky, but entirely doable.

** For transaction_type, you may need to set up a Transaction_Type table to track the different transactions (sold, purchased, traded, found, lost/stolen, etc. etc.) Alternatively, you could make it a varchar containing a descriptive string, or put a CHECK constraint to limit it to selected values.