Clojurians Log v2

Morning folks - we're really proud to have released beta7 this morning 🚀 ☺️ It's probably the biggest change to XT2 since the original public early access, and brings support for time-series shaped data. For so long (even throughout 1.x) we've had people asking us whether we were a time-series database, and we had to try to explain the difference between being a bitemporal system-of-record and the sorts of access/write patterns of time-series data - that eventually we decided that it was worth us making XT work for time-series data too 😅. I'll shortly be writing some lower-level blogs on what's changed under the hood, keep your eyes peeled. This release contains all of the changes that we've wanted to make to the on-disk format since that early access release (we've saved them up so as not to churn the format too much) - and, as such, there's a migration tool to run to upgrade your indices. Instructions in the release notes - let us know if anything's unclear or if you have any problems/questions. With this change, we're now very much in the final straight for the 2.0.0 launch - we're only expecting very minor changes between here and GA, getting the website and documentation in order etc. Thank you all so much for your help and support throughout this process, whether it's been trying it out, raising bugs, giving us feedback and suggestions - it's all very much appreciated 🙏 James

Thanks @markaddleman ☺️ Now that we have b7 released we're looking to return to more of a "normal" development pattern - smaller iterations, more frequent releases, compared to b7 which has been (openly) more of a big bang than we'd usually want due to the indexing changes. Window functions will form part of the ongoing roadmap discussions - likely post-GA at this point, if I'm honest - but definitely a valuable addition, and would move up the list the more we understand of the use cases it'd enable. cc @taylor.jeremydavid - he's usually the one keeping track of who's asking for what 🙂

> [support for window functions] would move up the list the more we understand of the use cases it'd enable Definitely this. Any concrete examples or prototypes (public or private) folks can share would really help for planning and prioritisation

My data is an event log. So, columns are timestamp, primary key, and then dozens or a hundred state columns. I have questions like: • How many times did a primary key transition from state A to state B? • On average, how many state transitions occur before reaching the final state? • What are the timestamps and new states for every state transition for a set of primary keys? • While a primary key is in a particular state, what are the latest values for other state columns?

ah man, dependency fun, do you think...? we have bumped a few as part of the release - most notably Java >= 21 but I can't see offhand how that'd affect Derby. anything in the deps tree?

Hmm, I matrix test against different jdk versions in CI so I'll investigate that as a possibility. That's the only difference really between CI and locally...

Okay, I modified CI to only add the xtdb deps and test against it for jdk21+ and everything is fine now. Thank you.