When I looked it up, there weren't any samples (even in English-speaking countries!), So it was easy.

Trigger

I was writing code in Python that defines the behavior of a program in YAML, and I just wanted to write a process to create an object in YAML.

As far as I can see PyYAMLDocumentation, there is a ** tag **, which seems to do what I want, but there is no usage example anywhere. I don't know if it should be used or how to define it. Well I was in trouble.

How to add a tag

So, suddenly it's the main subject. To add a tag, pass a value to the PyYAML module's global function, ʻadd_multi_constructor () `.

  yaml.add_multi_constructor("tag:yaml.org,2002:var", var_handler)

  # ...

  def var_handler(loader, suffix, node=None):
    #processing

If you write such code, it will react when the YAML file has the following tag and var_handler will be called.

  textvariable: !!var:StringVar
    name: label
    default: ""

Three parameters are passed to the function specified as an argument

loader: The loader object itself processing the YAML file being loaded
suffix: The string that follows the tag prefix. In the case of the above YAML code, : StringVar
node: A node object that indicates where the tag was found. Nodes in the lower layers below this are stored in value. In the case of the above YAML code, MappingNode (value = [(ScalarNode (), ScalarNode ()), (ScalarNode (), ScalarNode ())])

So, if you return some value as the return value of the function, the tag part will be the value specified in the return value. For example, in the above example, returning the character "Yes" is the same as writing textvariable: Yes.

ʻAdd_multi_constructor ()` If you follow the behavior of the function, it seems that if you omit the Loader defined as a keyword argument, the function specified here will be added to the three classes Loader, FullLoader, and UnsafeLoader. So if you use SafeLoader ** no matter what you do here, no function will be added. ** **

If you want to use only tags that you have defined in YAML

When reading YAML, if you want to read only the tags defined by yourself, it is better to specify SafeLoader as an argument and add a function.

I had some other work to do, and I was a little worried about messing with SafeLoader directly, so I defined a new class that inherited SafeLoader.

class GeneratorLoader(yaml.SafeLoader):
  def __init__(self, stream):
    super().__init__(stream)
    yaml.add_multi_constructor("tag:yaml.org,2002:var", GeneratorLoader.var_handler,
      Loader=GeneratorLoader)

  @staticmethod
  def var_handler(loader, suffix, node=None):
    pass

If you use the class created here as a Loader in the yaml.load () method, the tags will be processed.

Global functions in PyYAML do not have a function that receives an instance of Loader

Note that the functions defined in \ _ \ _ init \ _ \ _. Py of PyYAML such as yaml.load () do not have a function that receives an instance of Loader (all receive classes). Therefore, if you want to use the Loader object created above after loading YAML, you need to call Loader # get_single_data () directly.

loader = GeneratorLoader(self.string)
struct = loader.get_single_data()

Other notes

Tags must always start with "tag: yaml.org, 2002". Therefore, the value of the first argument of the ʻadd_multi_constructor ()` function always starts with "tag: yaml.org, 2002". However, this part is not actually written in YAML (If you define the tag "tag: yaml.org, 2002: var", you can only write "var" in YAML)

I don't understand the explanation, so show me the code for the time being

I'm working on it now, but I'm using it in the next module.

TakamiChie/TkSugar: A module that generates a structured Tk window frame from a text file.

As of September 21, 2020, there is such a process in tksugar / generator.py (may move the code later).

Use custom tags with PyYAML