Agreed. Details are important, such as what happens if you call a function twice in the same row.
The jira description currently says "See more details in Presto doc”. But I would like to see an actual specification, with examples. Julian > On Aug 3, 2026, at 2:36 PM, Mihai Budiu <[email protected]> wrote: > > One thing that would be nice to settle is how to mark such function as > "nondeterministic", similar to RAND(). As discussed elsewhere, these are > non-deterministic in a different way from functions such as NOW(), which > return the same value within a transaction. > > Mihai > > > ________________________________ > From: Jerome Haltom <[email protected]> > Sent: Monday, August 3, 2026 2:33 PM > To: Ian Bertolacci via dev <[email protected]> > Subject: [DISCUSS] Naming for UUID generation functions (CALCITE-5898) > > Subject: [DISCUSS] Naming for UUID generation functions (CALCITE-5898) > CALCITE-5898 asks for a uuid() function. I have an implementation ready, but > it raises a naming question I would like input on before I open a PR. > Rather than a single uuid(), I have implemented UUIDV4() and UUIDV7(), > following the names PostgreSQL 17 uses. A plain uuid() does not say which > UUID version the caller gets, and versions 4 and 7 have quite different > properties - version 7 is time-ordered and sorts in creation order, version 4 > is purely random. > Details and the full specification are on the JIRA: > https://issues.apache.org/jira/browse/CALCITE-5898 > Replies on the JIRA are fine if that is easier. > >
