Incorrect tag attribution

[reandreev]

When attempting to process a GitHub Workflow, the parsing logic incorrectly (?) interprets the property on of a workflow as a boolean value. For example, taking the following workflow:

name: Workflow

on:
  push:

jobs:
  job:
    runs-on: ubuntu-latest
    steps:
      - run: echo "test"

Doing yaml.safe_load(workflow) will produce the following dict:

{
    'name': 'Workflow',
    True: {
        'push': None
    },
    'jobs': {
        'job': {
            'runs-on': 'ubuntu-latest',
            'steps': [{ 'run': 'echo "test"' }]
        }
    }
}

The key which should've been a string with value on is instead a boolean key with the value True. This happens in the resolve method, more specifically at this step:

pyyaml/lib/yaml/resolver.py

Lines 150 to 152 in 69c141a

	for tag, regexp in resolvers + wildcard_resolvers:
	if regexp.match(value):
	return tag

where the tag tag:yaml.org,2002:bool is returned after the following regexp is matched to the value:

re.compile('^(?:yes|Yes|YES|no|No|NO\n                    |true|True|TRUE|false|False|FALSE\n                    |on|On|ON|off|Off|OFF)$', re.VERBOSE)

Any string on is converted to a boolean type which leads to this type of results. Is this intentional?

[perlpunk]

Is this intentional?

Yes, and it's a frequently asked question, so there is even a pinned issue for it: #486

It's because PyYAML currently only supports YAML 1.1, but GitHub workflows are using (a sub set of) YAML 1.2.

You can use https://pypi.org/project/yamlcore/ on top of PyYAML to get YAML 1.2 behaviour, where on is simply a string and not a boolean.