I use the MS Sync Framework to sync my SQL Server instance with a local SQL CE file to make it possible working offline with my Windows app.
I use GUIDs as keys. On my table I have a unique index on 2 columns: user_id and setting_id:
usersettings table ------------------ id PK -> I also tried it without this column. Same result user_id FK setting_id FK value
Now I do the following:
I create a new record in this table in both databases – SQL Server and SQL CE with the same user_id and setting_id.
This should work and merge the data together since this can happen in real life. But I get an error when syncing saying the unique key constraint led to an error. The key pair already exists in the table.
A duplicate value cannot be inserted into a unique index. [ Table name = user_settings,Constraint name = unique_userid_settingid ]
Why can’t MS sync handle that? It should not try to insert the key pair again. It should update the value if needed.
Advertisement
Answer
The issue is if you add the same key pair to different copies of the table, they get different IDs (GUIDs) as primary keys in this usersettings table.
As this is simply a many-to-many table between Users and Settings, there is no need to have that ID as a PK (or even a column at all).
Instead, just use a concatenated key of the two FKs e.g.,
CREATE TABLE [dbo].[usersettings](
[user_id] [UNIQUEIDENTIFIER] NOT NULL,
[setting_id] [UNIQUEIDENTIFIER] NOT NULL,
[value] [varchar](50) NOT NULL,
CONSTRAINT [PK_usersettings] PRIMARY KEY CLUSTERED ([user_id] ASC, [setting_id] ASC) );
Of course, include appropriate field settings (e.g., if you use VARCHARs to store the IDs) and relevant FKs.
As the rows inserted should now be identical on the two copies, it should merge fine.
If you must have a single column as a unique identifier for the table, you could make it meaningful e.g.,
- the PK (ID) becomes a varchar (72)
- it gets filled with
CONCAT(user_ID, setting_id)
As the User_ID and Setting_ID are FKs, you should already have them generated so concatenating them should be easy enough.