3eb7988a1663 23 hours ago

The Go community sold me on the necessity of programming mascots. The gopher is adorable, and I love seeing the creative tweaks for how people incorporate it into their marketing.

Ferris the crab also seems like it has been around since the early days, but I do not see nearly as many projects with crab heavy media. Guess it also has to be a mammal.

  • alexpadula 18 hours ago

    The mascot is indeed cute af!

    • sitzkrieg 14 hours ago

      Renée French is an excellent artist!

  • noir_lord 18 hours ago

    PHP has the elephants, lots and lots of different elephants.

    https://afieldguidetoelephpants.net/#variations

    I think they are cute as well but despite programming PHP (among other things) for decades, never had an elephant though I know a bunch of PHP devs who have dozens (or more).

    • Insanity 17 hours ago

      Java has “Duke”, arguably not as memorable though (even had to look up his name just now).

philippemnoel 15 hours ago

More people using and extending Postgres is a huge win for the ecosystem, but I am somewhat uncertain of the appeal of Go for this task. It seems to really shine at "platform"-level tasks (e.g. K8s, network proxies, etc. etc.), while Postgres extensions are more at the data plane level.

WanderZil 4 hours ago

Nice that errors reported by Go are raised only after Go returns, avoiding a `longjmp` through Go stack frames.

sooraj_sj 8 hours ago

anyone has good experice with vecor database ? is postgresql good ?

ltbarcly3 22 hours ago

LLMs make "I need to shoehorn a language I already know well into an environment where it doesn't really belong" an unnecessary extravagance.

Just have an AI write plpgsql. If what you are doing is too complicated for plpgsql you are doing something stupid.

  • dapperdrake 20 hours ago

    Then you will love plv8.

    Load an entire Chrome V8 or NodeJS V8 into Postgres to have JS pg modules.

    JS now runs on phones, on web servers, and inside your postgres relational database. Add Mongo and vibe code JS for redis or memcached for bonus internet points.

    If Rust is rewritten in JS, then Bun can finally return to its roots.

    https://github.com/plv8/plv8

    END half satire (plv8 is real, though)

    • jerrysievert 19 hours ago

      Give pljs a chance, fast startup, very fast execution, and a lot lighter weight than plv8.

  • tehlike 20 hours ago

    PLV8 is amazing. plpgsql is overly verbose. It's also now possible to write modules in rust. They are all great.

    • jerrysievert 15 hours ago

      thanks! but unless you need wasm, or very long running procedures, have you considered pljs instead? the startup time is much lower, which for fast running procedures ends up a bit faster than plv8. of course, for longer running functions or wasm (which pljs doesn’t support) plv8 is of course the winner.

  • cannonpalms 19 hours ago

    What about the full storage engine extensions, like ClickHouse, Timescale, etc.?

  • sgarland 19 hours ago

    Depends what the extension is doing; C (or Rust) is much faster for CPU-heavy operations than PL/PgSQL.

  • sebazzz 9 hours ago

    Well... SQL Server actually supports .NET (Framework).

angrydingo 22 hours ago

for code snippets on darkmode i can't see bracket

  • jttnr 21 hours ago

    Same for me on my mobile phone with dark theme.

Icathian 20 hours ago

Okay now this is cool. Anything to make postgres extensions nice and easy to write.

Still gonna go with pgrx though.

Kinrany 20 hours ago

Now that reimplementing old ideas is a commodity, can't we have a new database without carrying the historical baggage and bad decisions of old?

  • gjvc 20 hours ago

    which particular historical baggage and bad decisions of old were you thinking of here?

    • Kinrany 17 hours ago

      The big three are SQL, unpredictable and overly complex query planning, and defaulting to isolation levels that are about as safe as C.

collabspherestu 21 hours ago

The trickiest part with writing Postgres extensions in Go usually isn't the Cgo bindings—it's the runtime friction with Postgres's process model:

Signal handling conflicts: Postgres relies heavily on signals (SIGINT, SIGTERM, SIGUSR1) for query cancellations, latch wakeups, and statement timeouts. The Go runtime aggressively installs its own signal handlers for preemption and stack management, which can lead to subtle deadlocks or unhandled interrupts inside a backend process. Memory contexts vs. Go GC: Postgres manages memory via hierarchical memory contexts (CurrentMemoryContext, palloc), which automatically clean up on transaction aborts or errors (via longjmp). If Go code panics or a Postgres query gets cancelled mid-execution, reconciling Go’s GC-managed heap with Postgres memory lifecycles is notoriously tricky. Process-per-connection overhead: Postgres forks a new process per connection. Spinning up the Go runtime, GC threads, and runtime scheduler in every spawned backend worker can add noticeable memory and startup overhead under high connection counts.

Rust with pgrx has gained a lot of traction recently precisely because it has zero runtime and integrates directly into palloc and Postgres error handlers. Still, having pure Go tooling for teams with deep Go codebases is an interesting space to watch.

nonacosa 14 hours ago

Using TinyGo instead of standard cgo is what makes this practical inside a Postgres backend. With full Go, the runtime spawns background OS threads for scheduling and sysmon, which creates headaches around Postgres process-per-connection forks and signal handlers. TinyGo gives you a single-threaded C ABI binary that lives inside the backend without fighting the host lifecycle, even if it means giving up goroutines.