koanf VS logrus

LINK: https://github.com/knadh/koanf

For configuration, I like Koanf.

Of course. You should always use a good config package. caarlos0/env is okay but https://github.com/knadh/koanf is a better all around solution imo.

You can also use Viper with urfave/cli, or you can replace Viper with knadh/koanf, which I personally prefer.

Use viper with care. It has some dangerous fundamental flaws [1]. We got bitten hard, submitted a PR [2] and followed up for a year and a half to no avail, before I went ahead and reinvented the wheel and wrote koanf (plug), specifically to avoid viper's flaws.

Most importantly:

- Breaks JSON, YAML, TOML etc. language specs by forcibly lowercasing all keys internally. Dangerous because it can silently merge differently cased config keys into lowercase.

- Hard codes big unnecessary dependencies into the core, significantly bloating build sizes. No separation or abstraction.

[1] https://github.com/spf13/viper/pull/635

[2] https://github.com/knadh/koanf#alternative-to-viper

It's not hard to a) look for a key in an environment variable b) accept a key as a parameter c) load a key from a file d) all of the above with something like koanf

Came across this the other day https://github.com/knadh/koanf as an alternative to Viper.

https://github.com/knadh/koanf for config management. Readme contains why it's a good alternative to viper.

It's some sort of logging system well explained by Alex Edwards in Let’s Go Further. As stated, we could have used logrus or any other popular logging system in Go.

Option 2, please. You may not have been around for the logrus debacle, but it was a giant pain.

For the example I’m using the very popular logrus library and then I’m setting the log formatter to be JSON

For choosing the candidates for the poll, I didn't do any thorough research. I was looking for a library to use in my project at work, and I ended up at sirupsen/logrus which was already being used by one of the dependencies in that project.

I contribute to OSS as part of my job on the regular. The company is good about contributing upstream, signing CLAs, and all that. We still work against private forks for two main reasons: 1. Some changes that we need are not accepted by maintainers based on philosophical or architectural reasons that we can’t otherwise work around. You’re beholden to then unless you publicly fork the repo which has other legal/PR overhead for the company and OSS political implications. 2. Maintainers in the past have taken down repos, renamed repos, or changed the licensing on repos that have left us in a lurch. We always build against our own private forks because we need predictability and can’t be beholden to some other party for business continuity. We sync them down from the public upstream at our leisure.

If they change the case on their username on the other hand, the Go ecosystem explodes: https://github.com/sirupsen/logrus/issues/570#issuecomment-3...

Like, for example, some projects importing logrus with a capital L and some with a lowercase L, and go modules having no way to reconcile the two: https://github.com/sirupsen/logrus/issues/553

Ugh. Wish people would stop using logrus. It’s in maintenance mode and slow, especially its stack tracing.