[linsomniac]

You might be able to get away with staying with MySQL or a variant. Years ago I did an "interview problem" to generate a report of some accounting data of a sample data set in MySQL. The details are foggy now, but I ended up doing it all using advanced SQL features that I thought only existed in PostgreSQL (you know, or Oracle), but whatever engine I ended up using had it as well. It was probably the latest Percona or Maria?

[Conlectus]

MySQL has worse guarantees in several cases, though. For example, using `SERIALIZABLE` isolation means your queries can deadlock when run in parallel.

Likewise, selecting a column not present in a GROUP BY leads to random values being returned.

[donatj]

> Likewise, selecting a column not present in a GROUP BY leads to random values being returned.

That hasn't been the default behavior for something like 5 years.

Beyond that the values aren't "random" as much as "one of the values from the set" which can be a useful feature in cares where you want any of the values and don't want the performance hit that comes with using an aggregate function particularly on very very large sets.

[johannes1234321]

MySQL improved a lot in those places.