问题
I am designing DATABASE for a Sales and Purchase application like ERP and using MYSQL as RDBMS, I have doubt on creating table for sales and purchase entities to go with single table for each module(Sale/Purchase) or multiple tables for each entities(Sales order, Sale invoice, Sale return, Purchase order, Purchase invoice, Purchase return) in longer run. Below is my use case.
My application will have Sale Order, Sale Delivery, Sale Invoice, Sale Return and Credit Note and same entity for Purchase module also and all these entity may be enter linked in there module. Like Sale Order can be converted into Sale Delivery or Sale Order and Sale Delivery can be converted into Sale Invoice. So there need to maintain reference b/w each entity of a module.
Now, I am little confuse to keep all this in one table for each module say "sale_entity" and "purchase_entity" having entity type Or should I create separate table for each entity type say sale_order, sale_invoice, sale_return, purchase_order, purchase_invoice, purchase_return etc.
Below is what running in my mind for both the cases:
Single Table: I really want to keep this in single table for each module but I am worry about performance in longer run, It will increase the table size quickly and may slow down the performance.
Multiple table: It will be difficult to manage, maintain relationships and fetching data in reports for all entity types of records at once, requires union and all.
My understanding is that large size of table performs slower than the small size of table, Please correct if I am wrong.
Please put some light on it, and suggest me how should I proceed.
Thank You
回答1:
Go for a single table! When you don't you will end up with a lot of duplicate data that is going to be a pain to maintain.
Using UNION AND JOIN etc... are the slowest methods of querying. Try to avoid them as much as you can.
To update a sale_order to a sale_delivery with a single table requires a query like this: UPDATE orders SET orderType = 3 WHERE orderId = 1533
To update a sale_order to a sale_delivery with multiple tables requires a query like this:
INSERT INTO sale_delivery
(orderId, clientId, ....)
SELECT orderId, clientId, ...
FROM sale_order;
You can see that this method is going to be a lot slower and you are going to have to delete the original sale_order. Hoping the insert was succesfull.
How are you going to handle orderId's? I suppose you are going to use an autoincrement, are you going to copy this to the other table? How do you know that id is still available in the other table?
What are you going to do with order_details? You have to match them with the order. Are you going to update the relation when you 'copy' the order to the new table?
CONCLUSION: Please stick to a single table. Be sure to create indexes.
What you can do however is split the database per end-user company. When you are going to have data that is never going to be used together you can create a new database for the new data.
回答2:
Rule of thumb: If two table have identical (or nearly identical) columns, make it one table, not two. You may need to add a column to distinguish the two types of data. You may need to have a column be NULL if it applies to one use but not the other. Too many NULLable columns --> don't combine the tables.
Rule of thumb: One-to-many and many-to-many relationships require having two or three tables. (The third is for many-to-many.)
"Orders" usually involved "Order Items". That is one row in Orders maps to one or more OrderItems. Those should be separate tables.
"Purchases" versus "Returns"? Maybe they could be in the same table, and distinguish by the sign of the amount?
回答3:
Use separate tables.
I agree with the notion that you will have similar data in both tables but there is more to a Sales Order header and Sales Orderline than just the design of the table.
The concept of a sales order is very different logically than say the concept of a purchase order.
On it's most basic level one is directly related to a debtor account and one to a creditor account.
Both also have very different procedures when it comes to stock movements. E.g. One is goods inwards, which has a whole set of concepts like costing associated. Whereas the other is a goods outwards which has associated delivery information such as couriers, cost of delivery, etc.
I am assuming if you are trying to replicate an ERP then you will also have to have some sort of manufacturing works orders tables, which when considered together with a sales order and purchase order, interact very differently with stock transactions and stock locations. When you consider the differences between 1. receiving goods to a location during putaway, to 2. consuming raw materials receieved and generating new output items to different stock locations, to 3. sales orders for items that are generated from a works order and not part of the goods inwards receiving process, then you start to see that trying to dump all that logic into a single table would be complete madness.
You also need to consider the affects of having a single table in terms of programmability. What happens when you need to lock the table for a transaction? You have now locked all users from all data related to orders incl. Sales orders, purchase orders, and works orders.
So, in short, you need to separate the data with the logic.
I have never seen an ERP database that has a single table for all order types, or all stock transactions, or all debitor transactions, etc.
Anyone who says different is a bluffer of monumental proportions.
If in doubt remember that RELATIONAL database systems were designed so that you could separate data across multiple tables reducing the issues associated with single table designs, like performance, table locking, separation of logic (stored procedures, triggers, etc). Use sales order header numbers (or equivelant) as the relational key. Its a far tidier design principle.
来源：https://stackoverflow.com/questions/32646014/keeping-one-table-or-multiple-table-for-similar-type-of-data-which-one-is-best-w