bogus "found character '\t' that cannot start any token" error

[ensc]

I have a gitlab CI yaml file with

variables:
  OS:			fedora:43

yamllint v1.38.0 complains about this with

  15:6      error    syntax error: found character '\t' that cannot start any token (syntax)

which seems to be triggered by the tabulators after "OS:"

YAML 1.2 allows tabulators at this place (https://yaml.org/spec/1.2.2/#62-separation-spaces)

[andrewimeson]

Your document is valid YAML 1.2, but since the tab does is a part of the style and not the data¹, I think disallowing it on the basis of style (which is part of what a linter does, this isn't just a validator) could make sense. YAML 1.2 still forbids using tabs for indentation, so being more strict than the spec and disallowing it here (where it's purely stylistic, not a necessary part of the data) is reasonable.

That's my opinion as a matter of style. If someone disagrees, it would be hard to implement a way to allow this since PyYAML (which yamllint uses) does not support 1.2.

Footnotes

1. YQ (Golang, supports YAML 1.2) at least doesn't include it in the data.

   $ printf '%%YAML 1.2\n---\nvariables:\n  OS:\tfedora:43\n' | yq --output-format json
   {
     "variables": {
       "OS": "fedora:43"
     }
   }
   

   ↩

[andrewimeson]

It looks like even ruyaml, which advertises YAML 1.2 support, rejects using tabs here.

>>> from ruamel.yaml import YAML
>>> yaml = YAML(); yaml.version = (1, 2)
>>> yaml.load(StringIO('key:\tvalue'))
ScannerError: found character '\t' that cannot start any token

I think it's reasonable for yamllint to disallow this syntax since at least one common YAML 1.2 parser doesn't even support it.

[ensc]

I think, there is a difference between "style" issue and the reported "error" severity (clearly wrongly labeled as "syntax error").

For ruamel: I opened https://sourceforge.net/p/ruamel-yaml/tickets/560/