Tools for auto-formatting linkml models (yaml)

[dalito]

How do you format the yaml of your linkml models? I have recently started experimenting with yamlfix in pid4cat-model. Its GitHub repo contains a pair of configs for yamllint and yamlfix that are consistent if you want to try (.yamllint.yaml and .yamlfix.toml).

Other yaml-formatters like prettier, yamlfixer or yamlfmt did not work well for me. I am very interested to hear about your preferred tools for auto-formatting.

Moved here from slack as suggested.

Chris Mungall:
I was sure we had previous discussions about yaml syntax level linting, but I can’t find. But I recall that others had similar experiences.

On cursory examination yamlfix looks great, and the fact it can be installed with uv or pip is a big win. Do you want to make a /discussion out of this?

[matentzn]

My general position to questions like this is:

1. Most people (~67%), including most schema developers, wont care too much which formatter (unless its some crazy opinionated things which does a lot of unnecessary things)
2. However, many people would appreciate some formatting tool that they can use to quickly canonicalise (normalise) their schema rendering. We use these all the time in ontology development. I would even go as far as saying that we should create one for LinkML which is a bit cognisant about our keys, and sorts for example metadata, enums, classes etc in that order.

In any case, whatever you come up with, I just want to throw my head in the ring for a linkml normalize-schema command which does some version of what you are proposing!

[cmungall]

I think it's important that schemas are readable and editable. It's not hard for yaml serializers to completely mess this up

E.g.
https://linkml.io/linkml-owl/examples/#parts-collection-with-counts

I think users might be upset if that sort of thing happens to their schema

[matentzn]

Related issues:

• Missing linter for transformation specification files. linkml-map#140 (this is about schema instances rather than schemas)

[cthoyt]

I'm using prettier for all my projects for YAML and markdown linting (see https://github.com/cthoyt/cookiecutter-snekpack/blob/98fcef2184d26d07ec67bd74f19b984b8c2398e3/%7B%7Bcookiecutter.package_name%7D%7D/justfile#L52-L55 in my cookiecutter repo)

# for formatting
npx --yes prettier --write --prose-wrap always "**/*.yml"
npx --yes prettier --write --prose-wrap always "**/*.md"

# for checking, e.g., in CI
npx --yes prettier --check --prose-wrap always "**/*.yml"
npx --yes prettier --check --prose-wrap always "**/*.md"

i am really skeptical of the python ecosystem of formatters compared to the JS world, which is much bigger and more likely to get maintained. Any python code that used ruamel has been a big pain for me, since it's hard to get that installed and running

[cmungall]

Quick comparison here:
https://www.perplexity.ai/search/i-need-to-normalize-yaml-forma-6XZDC8_jSfGEcVrLfiNemw

But IMO doesn't help decide between the options.

Ideally there would be a standard but that doesn't seem to be the case, making tool lock-in more likely.

It would be useful to see examples of each format, committing to the wrong one could be terrible if it renders yamls semantically correct but unreadable