Please supported mixed mode values

[xnox]

If one specifies that something should be a string, please encode it in the most natural and deterministic way.

Example for something desired to be formatted as a list of strings please ensure that all form of canonical integer and floating point numbers are encoded as strings.

https://yaml.org/spec/1.2.2/#24-tags

Lists

canonical: 12345
decimal: +12345
octal: 0o14
hexadecimal: 0xC
canonical: 1.23015e+3
exponential: 12.3015e+02
fixed: 1230.15
negative infinity: -.inf
not a number: .nan

As valid representations of numbers, which upon parsing will not be strings. And thus need to be quoted to be strings.

For the purpose of melange, we can encounter something like "1.23015e+3" as a string value represetting package version or text git tag. But most commonly it is software with single major or single minor version only that can get miss-parsed.

[luhring]

I'm not sure I'm following what this issue is asking for yet... Are you saying that yam is mishandling YAML tags then they're specified? Are you saying that yam should define its own default behavior for when no tag is present in the input?

An example input and expected output would go a long way here! Thanks!

[xnox]

@luhring

When doing yam on glibc.yaml I would like it to convert

package:
  name: glibc
  version: 2.40
  epoch: 3
  description: "the GNU C library"
  copyright:
    - license: LGPL-2.1-or-later
  resources:
    cpu: 16
    memory: 16Gi

to

package:
  name: glibc
  version: "2.40"
  epoch: 3
  description: "the GNU C library"
  copyright:
    - license: LGPL-2.1-or-later
  resources:
    cpu: 16
    memory: 16Gi

As version is a string, epoch is an integer, and cpu is an integer, and memory is unambiguous string.

[luhring]

Sorry for the delay! 😳

Thanks for this example, it's super helpful, and I think I'm understanding what you're looking for now.

For context, yam is operating on YAML bytes. It doesn't have knowledge of application-specific types. This is by design: it's a general purpose YAML formatter.

Given that, I could see implementing this feature in one of two ways:

1. yam's config has a couple properties that take in "path expressions" — gap and sort. This lets the user indicate preferences for how to handle type-specific aspects of the YAML structure in a way that's loosely coupled to the application domain. This has been working well for Melange files, for example. So we could add another config property that says "quote this string value if it could be interpreted as some other non-string type otherwise" for a given set of path expressions (e.g. ".package.version").

2. If what you're wanting to do is specific to Melange, we could solve this in Melange directly instead of in yam, by adjusting the type-specific YAML marshaling to handle all string-type fields in the manner you're suggesting.

I'd also be curious to know what problems you're seeing in some specific use cases with today's behavior, to help with the discussion about how best to solve it.

[xnox]

Yes adding something to .yam.yaml would be great. Something like "quote:" key to always quote a given subkey.