On: The many faces of DISTINCT in Postgres (2017)

[paulddraper]

It's as resilient to race conditions as it could be. Does exactly what you'd expect, no?

That is annoying about sequences, I agree.

[richbell → paulddraper]

> It's as resilient to race conditions *as it could be*. Does exactly what you'd expect, no?

While that that may be true, based on the hours I've spent scouring Stack Overflow and GitHub issues it seems that many people don't realize that it's only 99.999% resilient.

[vore → richbell]

It is 100% resilient short of your hard drive going through a microwave. It is almost definitely an application logic error if you are observing non-atomicity. What would non-atomicity even look like here? Only some columns being written?

[paulddraper → richbell]

Could you describe what you mean? What phenomenon? For the default transaction isolation level, or a different one?