Ordinary taks - complex solution - sql

Question

I have some users (C1, C2, C3, etc.) who handles products (aa, bb, cc, dd, ee, ff, gg, hh, etc.) in differents stores (St1, St2, St3, St4, etc.).
Every user in the list wants to know in which store which products they can handle cheaper.
How looks out the tables and how looks out the queries if user want get at least the following 3 things (one at the time):
1- Get own list of products. Exemple:
Pr. St1 St2 St3 St4
aa $20 $12 $19 $22
bb $31 $44 $38 $44
cc $18 $12 $19 $22
dd $36 $44 $38 $44
ee $15 $12 $19 $22
2- Get a list of lowest prices (but greather than 0) and see how much he/she save if he/she handles the same products on others stores. Exemple:
Pr. St4 St1 St3 St2
bb $23 $27 $26 $28
ee $14 $15 $15 $20
hh $36 $38 $40 $37
Sum $73 $80 $81 $85
Count products 3.
Pr. St2 St1 St3 St4
aa $32 $33 $38 $36
cc $21 $29 $27 $25
ff $13 $14 $17 $20
Sum $66 $76 $82 $81
Count products 3.
3- Get a list of products which has cero price in each store. Exemple:
Pr. St1 St2 St3 St4
kk $00 $12 $19 $22
ii $00 $44 $38 $44
Pr. St2 St1 St3 St4
ll $00 $21 $52 $20
mm $00 $13 $17 $15
A primitive not good solution:
CREATE TABLE usrs (
idc INT NOT NULL,
name VARCHAR(50) NOT NULL,
stores VARCHAR(50) NOT NULL,
pwd VARBINARY(72) NOT NULL,
PRIMARY KEY (idc)
)
COMMENT='Customers'
COLLATE='utf8_general_ci'
ENGINE=MyISAM;
CREATE TABLE stores (
ids INT NOT NULL,
nm VARCHAR(50) NOT NULL,
PRIMARY KEY (ids)
)
COLLATE='utf8_general_ci'
ENGINE=MyISAM;
CREATE TABLE products (
idp INT(11) NOT NULL AUTO_INCREMENT,
prod VARCHAR(50) NOT NULL,
st1 MEDIUMINT(9) NULL DEFAULT '0',
st2 MEDIUMINT(9) NULL DEFAULT '0',
st3 MEDIUMINT(9) NULL DEFAULT '0',
st4 MEDIUMINT(9) NULL DEFAULT '0',
PRIMARY KEY (idp)
)
COLLATE='utf8_general_ci'
ENGINE=MyISAM;
INSERT INTO products (prod,st1,st2,st3,st4) VALUES (aa,14,20,13,17);
INSERT INTO products (prod,st1,st2,st3,st4) VALUES (bb,33,29,38,33);
INSERT INTO products (prod,st1,st2,st3,st4) VALUES (cc,19,20,00,21);
INSERT INTO products (prod,st1,st2,st3,st4) VALUES (dd,22,29,25,33);
INSERT INTO products (prod,st1,st2,st3,st4) VALUES (ee,30,00,35,29);
INSERT INTO products (prod,st1,st2,st3,st4) VALUES (ff,10,14,11,13);
INSERT INTO products (prod,st1,st2,st3,st4) VALUES (gg,00,00,00,00);
INSERT INTO products (prod,st1,st2,st3,st4) VALUES (hh,16,22,30,10);
INSERT INTO products (prod,st1,st2,st3,st4) VALUES (ii,23,34,34,26);
INSERT INTO products (prod,st1,st2,st3,st4) VALUES (jj,41,32,39,41);
INSERT INTO products (prod,st1,st2,st3,st4) VALUES (kk,25,29,26,19);
INSERT INTO products (prod,st1,st2,st3,st4) VALUES (ll,24,27,10,24);
INSERT INTO products (prod,st1,st2,st3,st4) VALUES (mm,29,41,37,36);
I don't know how all the tables can looks like when count of stores are more than 2 and count of users are more than 1.
In my solution I have to manually change the tables when users or stores increases. I understand here needs a relationship between tables, but I don't find out it. Bad, very bad ...
Thanks in advance!

Answer 1

You could change your solution such that you had a list of prices connecting stores to products.
A price would be a tuple of Product, Store, and Price.
Then, you could select with a left join on this table. The where clause could have a specified store id, a price requirement, and/or product id.
Example:
I have modified the names of your tables for better clarity. Examples tested on MariaDB.
First, the models I will use:
CREATE TABLE `stores` (
`id` INT NOT NULL,
`name` VARCHAR(50) NOT NULL,
PRIMARY KEY (`id`)
);
CREATE TABLE `products` (
`id` INT(11) NOT NULL AUTO_INCREMENT,
`name` VARCHAR(50) NOT NULL,
PRIMARY KEY (`id`)
);
CREATE TABLE `prices` (
`product_id` INT(11) NOT NULL,
`store_id` INT NOT NULL,
`price` DOUBLE NOT NULL,
CONSTRAINT productIdToStoreID UNIQUE (product_id, store_id)
);
Here is how I would populate the database
INSERT INTO stores (id, name) VALUES (10, "Foo's Fun");
INSERT INTO stores (id, name) VALUES (20, "Bar's Barn");
INSERT INTO products (id, name) VALUES (10, "0xC0C0C01A");
INSERT INTO products (id, name) VALUES (20, "0xDEADBEEF");
INSERT INTO prices (product_id, store_id, price) VALUES (10, 10, 1.50);
INSERT INTO prices (product_id, store_id, price) VALUES (10, 20, 1.60);
INSERT INTO prices (product_id, store_id, price) VALUES (20, 10, 40);
INSERT INTO prices (product_id, store_id, price) VALUES (20, 20, 35);
Here are some starting points are queries you could write:
SELECT store_id, price FROM prices WHERE product_id = 10; -- All prices for product with id 10
SELECT store_id, price FROM prices WHERE product_id = 10 ORDER BY price ASC LIMIT 1; -- Get Cheapest Price and Store for product 10
SELECT stores.name, price FROM prices LEFT JOIN stores ON (stores.id = store_id) WHERE product_id = 10 ORDER BY price ASC LIMIT 1; -- Get the Store name and price.
Motivation:
We need a way that can relate n stores, m products. (Not sure where users go here). We need a data structure that can hold:
If a store has a item (conversely, if a product is in a store). We can see that this is a many to many relationship. SQL databases tend to be really optimized for this type of relationship.
The price of a product at a store.
Thus, we create a table that links a product to a store uniquely, with the extra metadata of the price.
As you pointed out, the alternative to this is a table that contains all of the stores as columns. However, this does not scale well at all as in order to add, remove, or change the name of a store would be a O(n) update. In a production environment, this would be detrimental as this would likely lock the database. but I digress.
This solution solves this issue by allowing the store list and product list to be decoupled from the actual products. If a price exists at a store for a product, there is a row. This means that the space complexity is at worst O(n), but in reality would be much less as not all stores would have all products.