How and When to Use Databricks Identity Column

[@artus198]

yep, I come from SQL , we used identity all the time on PKs ... ! Thanks for the video.

[@AccountingatCHDS]

Partitioning by identity coliumn would be indeed a bad idea but if we had the partition by range feature, it could be interesting so that you partition by million or rows as an example...

[@felipeporto4396]

This is pure gold

[@mateen161]

Useful video. Thank you.

[@lunatyck05]

Should also mention if you try to insert a id value you will get an error when using GENERATED ALWAYS AS IDENTITY but you are able to use GENERATED BY DEFAULT and override identity or insert nulls into the column

[@BryanCafferky]

Hmmm.. I thought I did mention that but guess I was not clear. Thanks

[@NoahPitts713]

Great and timeless info as always! qq - how has your experience been using the IDENTITY Column since making this video?

[@BryanCafferky]

I have not dared to try it again. lol I don't think it is the best idea for a scaled out service to be honest. Ok if updates are limited but I would still try to find an alternative.

[@Ali-q4d4c]

Why delta table GENERATED AS ALWAYS IDENTITY column doesn’t generate sequential continuous number? It starts from 1,2,3,6,8,9 ,11 etc..

[@BryanCafferky]

Interesting. Not sure but if you insert and later delete, the previously used identity value will never be used again so gaps can occur over time.

[@MinPro-ws7ib]

Is the table managed or external?

[@BryanCafferky]

It's actually not relevant for this.