How does the library handle records with identical timestamps?

[wesleyteixeira]

I'm studying this framework and have a question about how it handles this scenario, or if it's a known limitation.

Scenario

1. Bulk insert creates 10,000 records in the same millisecond
2. All records have UpdatedAt = '2024-10-20T10:30:45.123Z'
3. First sync pulls 100 records (default PageSize)
4. Client saves delta-token: '2024-10-20T10:30:45.123Z'
5. Next sync queries: WHERE UpdatedAt > '2024-10-20T10:30:45.123Z'
6. Wouldn't this skip the remaining 9,900 records?

Common Patterns for Handling This

1. Composite key approach - Use a secondary column (like Id) for ordering:

WHERE UpdatedAt > @lastUpdatedAt
   OR (UpdatedAt = @lastUpdatedAt AND Id > @lastId)
ORDER BY UpdatedAt ASC, Id ASC

2. SQL Server rowversion - When using SQL Server, you could store the last synchronized rowversion instead of a timestamp, since rowversion is guaranteed to be unique and monotonically increasing:

WHERE [rowversion] > @lastRowVersion
ORDER BY [rowversion] ASC

Does Datasync Community Toolkit handle this scenario in any of these ways, or is it a known limitation?

[adrianhall]

It's a known limitation. You need to give each record a slightly different UpdatedAt - doing them in groups of 10 per msec is possible.

Unfortunately, using ANY database, we can't rely on functionality within a single database.