+1
Having a unique constraint on a table is there a way to avoid FME stoping the bulk copy or insert simply ignoring the duplicate values and keep on with the translation, just reporting the warning ?
Or would be there any other alternative to UNIQUE constraints to avoid duplicates entering database without stopping the translation ?
Cheers
+17
If your goal is to filter out features having duplicate value for the unique field, the DuplicateRemover transformer might help you.
Takashi
+1
No, the DuplicateRemover will not work in this case.
Its a tile process running tile by tile and the duplicates are created at tiles shared borders. (point dups)
So tile A runs and creates feature ID1 at east border of tile, then tile B runs and creates the same feature ID1 at the west border of tile. At the end there are 2 ID's 1 at the shared border between tiles.
I was searching for some logic to be applied at DB side to prevent duplicate inserts , tried to apply a UNIQUE Constraint by ID but that will stop the translations whenever the constraint is checked.
Hope i was clear
Jorge
+17
Not sure the situation and requirement.
+2
It seems you want to run the workbench without stopping because of writer is having an unique constraint and failing the translation.
In these cases normally I would suggest to translate to stage user without unique constraint and verify the data and make necessary editing to the data / work bench (based on requirement) and then copy the data back to required user when data is as required.
This will help us to identify the loop holes / errors of data or work bench.
Hope this helps you.
Pratap
+25
+1
Thanks
+25
Thanks
That's an interesting one. My first thought is that you set a flag in the database when you start to process a tile. Then you have a SQLExecutor in the original workspace that checks if the flag is set for any neighboring tiles. If you put this in a custom transformer you could send it in a loop, using a Decelerator, to check every 10 seconds (for example) and keep looping until all the neighbor flags are unset. The other thought is you use a checkerboard pattern to process data, and do it in two processes (ie you process A1, A3, A5, B2, B4, C1, C3, etc, and once they are done you process A2, A4, B1, B3, B5, C2, C4, etc). That seems the most efficient to ensure no two neighboring sets of tiles are being processed simultaneously.
Enter your username or e-mail address. We'll send you an e-mail with instructions to reset your password.
