使用Guids维护数据库中的键与使用整数的另一个键之间的数据库关系

Question

I'm trying to write an application that will take the information out of a SQL Server database, load up the object models via a webservice, and then pass that information into an Access DB, which is based on a different schema than the SQL Server database, but the information contained is the same (so I assume this would be akin to an ETL process). The problem that I'm running into is that the Primary Keys in the database I'm pulling the information (SQL Server) are GUIDs whereas the database I'm putting them into (Access) are INTs. So for example:

Table 1, Persons Table in SQL Server:

╔══════════════════════════════════════╦══════════╦════════════╗
║                  ID                  ║   Name   ║ CreatedOn  ║
╠══════════════════════════════════════╬══════════╬════════════╣
║ 3648F6BB-F743-4952-9C69-93336667F3B1 ║ Person 1 ║ 01/01/2012 ║
║ DE44577A-CAE7-4101-B962-C052214E723B ║ Person 2 ║ 02/01/2012 ║
║ 94115C5E-9C7E-40CF-8A87-D4D837BD5DB3 ║ Person 3 ║ 03/01/2012 ║
║ F93E77D9-C344-420C-9499-BA3B4DD0F0BA ║ Person 4 ║ 04/01/2012 ║
╚══════════════════════════════════════╩══════════╩════════════╝

Table 2, Persons Table in Access:

╔════╦══════════╦════════════╗
║ ID ║   Name   ║ CreatedOn  ║
╠════╬══════════╬════════════╣
║  1 ║ Person 1 ║ 01/01/2012 ║
║  2 ║ Person 2 ║ 02/01/2012 ║
║  3 ║ Person 3 ║ 03/01/2012 ║
║  4 ║ Person 4 ║ 04/01/2012 ║
╚════╩══════════╩════════════╝

So Table 1 is how the data is returned in the SQL Server database, and Table 2 is how the information should appear in the Access database. So all GUIDs should be integers, but the relationships between data should remain the same. So for example, if I run a query to get the person's address (the address table would also be setup similarly) in SQL Server and Access, the query should return the same results regardless of whether or not a GUID or integer is used.

What I was thinking was using ROW_NUMBER() in SQL Server, ordered by the CreatedOn date (which is a date time type in the database and as such, should be unique across all records):

SELECT 
  (ROW_NUMBER() OVER (ORDER BY CreatedOn)) AS ID,
  Name,
  CreatedOn
FROM Table2;

The only thing is that I'm seeing duplicate integer IDs being returned from the query. So for example Table 1 above would look like this:

╔════╦══════════╦════════════╗
║ ID ║   Name   ║ CreatedOn  ║
╠════╬══════════╬════════════╣
║  1 ║ Person 1 ║ 01/01/2012 ║
║  2 ║ Person 2 ║ 02/01/2012 ║
║  1 ║ Person 3 ║ 03/01/2012 ║
║  1 ║ Person 4 ║ 04/01/2012 ║
╚════╩══════════╩════════════╝

When each ID should be unique. Can anyone think of a good way to go about doing what I'm trying to accomplish? Is there anything wrong with the way I'm currently trying to do it?

Any help would be greatly appreciated.

Answer 1

I wouldn't rely on ROW_NUMBER() if I were you, because you probably can't be sure that the order is always the same:

First, in your example you show only date values (without time) for the CreatedOn column. If more than one person is created at the same day, you can't be sure which comes first if you order by that date.
Even if you actually have a time in the CreatedOn column as well - if a person is deleted from the table, the ROW_NUMBER of all subsequent persons will change.

---

The easiest solution would be to change one of the tables, like already suggested by webturner in his answer .
If you can't do this for any reason (for example if you aren't allowed to change the schema of either database at all, or if legacy stuff would break if you changed the schema of the tables), you can create a mapping table where you store the relation between the two tables:

╔══════════════════════════════════════╦══════════╗
║             SqlServerID              ║ AccessID ║
╠══════════════════════════════════════╬══════════╣
║ 3648F6BB-F743-4952-9C69-93336667F3B1 ║    1     ║
║ DE44577A-CAE7-4101-B962-C052214E723B ║    2     ║
║ 94115C5E-9C7E-40CF-8A87-D4D837BD5DB3 ║    3     ║
║ F93E77D9-C344-420C-9499-BA3B4DD0F0BA ║    4     ║
╚══════════════════════════════════════╩══════════╝

You can even put it into a third database if you are not allowed to change the existing ones.

Answer 2

The simplest solution is to change the schema at one end or the other so they are the same. If you are adding records in both the Access and SQL ends then I'd use the GUID to prevent the same ID being added for two different records. Then all you are doing is rolling your own 'replication' system. Both Access and SQL support GUIDs.

Failing that you are going to require some sort of lookup for each table. This will give you the Integer equivalent for each GUID and vice versa. If you add the GUID as an additional column in the Access table (Table 2) then you can use this as the lookup table.

ROW_NUMBER() will be returning unique numbers, but will start again at 1 each time it is used so all the inserts for each table would need to be done in one set. If you use an Autonumber field in access this will give you unique values across separate inserts.

Given that you now you have SQL's GUID and unique IDs in the Access table you just need to do a lookup whenever you insert into a table with a foreign key, such as the address table. So when you are inserting addresses from SQL with GUIDs to identify the person Join access's Person table on the GUID and insert the Person's integer ID instead.