From now on I will think of LLMs as "Liquid Language".

Are there established packages for "wide logs" in #golang?

I wanted to accumulate fields and emit 1 slog record for each request. I set up a map in a middleware and then adding to it during flight. But that feels a bit janky.

After all, why shouldn't I begin #golang tests with

ctx, l := setup()

where l is a handle to things running "locally". Like a cli for my service api, the db, localstack s3.

Stumbled across this amazing jazz channel on youtube.
youtu.be/foHfPB6H19c?...

🌊⛵

TIL gopls has code actions for putting func params/struct/fields/etc. on separate lines.

In tests I usually `t.Logf("%#v", got)` then copy paste that in as the `want` variable. Now I can throw out my dodgy treesitter scripting and use gopls proper instead 🚀

Just finished The Overstory by Richard Powers. What a great read 🌲🌳🌴

When logging in #golang, do you prefer errors as fields that can be queried, or as parts of the message for glanceability. Sort of:

slog.Error("something wrong", "error", err)

or

slog.Errorf(fmt.Errorf("something wrong: %w", err))

Finally figured out what kind of audio a bot needs to send to Discord for smooth(-ish) playback. There are lots of tools & helpers but little documentation. The magic sauce was opus audio segments with variable bitrate off.

So ffmpeg flags: -c:a libopus -vbr off

I've picked up Caddy for some hobby projects and so far its great. Empty-VPS-to-working-HTTPS-service in no time and very little cruft:

foo.bar.baz {
reverse_proxy localhost:8080
}

Just remember to open those firewalls... all of them...

ClickHouse is doing some cool stuff with observability, ingesting data from Otel and so on. I didn't realize until now but there is a Grafana datasource plugin to hook up and visualize that data:
github.com/grafana/clic...

Would be interesting to compare this against other self-hosted obs.

I set up the LGTM (#grafana et al.) stack on my server to monitor some hobby projects. There were more components than I anticipated, but in the end it sort of works: github.com/vikblom/obsl...

I decided to bundle up my thoughts and starter code for monitoring a Go application before I forget: blog.aniara.dev/posts/go-otel/

Favoring Open Telemetry mostly because I like the idea and want it to be used widely. Even if the API is sometimes awkward. #golang

I really like the least #golang logger "slog". Makes so much sense to pass in a context when logging, passing that through all the way to the handler, which can grab whatever it needs (typically trace_id).

Only wish slog.LogContext was shorter...

Fumbling with a .zip sent by a friend with a Windows computer.

utf-8? of course not lol
is it utf-16? nonono
could it be latin-1? dream on
so windows-1252? get baited sucker

Code Page 865 for "nordic languages", Microsoft what are you smoking.

This weeks Go lesson learned: database/sql might (depending on which driver is used) close any and all connections that are used to run a rollback. This causes major connection churn if transactions are over-/mis-used. There is no counter for how many conns are closed like this in DBStats.

Hej! 👋