Re: [PATCH v2 06/11] qapi/introspect.py: add _gen_features helper

[Markus Armbruster <armbru@redhat.com>]

John Snow <jsnow@redhat.com> writes:

> On 11/16/20 3:47 AM, Markus Armbruster wrote:
>> John Snow <jsnow@redhat.com> writes:
>> 
>>> _make_tree might receive a dict or some other type.
>> 
>> Are you talking about @obj?
>> 
>
> Yes.

Recommend to be explict: _make_tree()'s first argument can be ...

>      It *usually* takes a dict. sometimes it doesn't.

Yes.  It takes an abstract syntax tree: dict for JSON object, list for
JSON array, str for JSON string, bool for JSON true and false, NoneType
for JSON none.  JSON int isn't implemented, because it doesn't occur in
SchemaInfo.

>>>                                                      Adding features
>>> information should arguably be performed by the caller at such a time
>>> when we know the type of the object and don't have to re-interrogate it.
>> 
>> Fair enough.  There are just two such callers anyway.
>> 
>>> Signed-off-by: John Snow <jsnow@redhat.com>
>>> ---
>>>   scripts/qapi/introspect.py | 19 ++++++++++++-------
>>>   1 file changed, 12 insertions(+), 7 deletions(-)
>>>
>>> diff --git a/scripts/qapi/introspect.py b/scripts/qapi/introspect.py
>>> index 803288a64e7..16282f2634b 100644
>>> --- a/scripts/qapi/introspect.py
>>> +++ b/scripts/qapi/introspect.py
>>> @@ -76,16 +76,12 @@
>>>   
>>>   
>>>   def _make_tree(obj: Union[_DObject, str], ifcond: List[str],
>>> -               features: List[QAPISchemaFeature],
>>>                  extra: Optional[Annotations] = None
>>>                  ) -> TreeValue:
>>>       if extra is None:
>>>           extra = {}
>>>       if ifcond:
>>>           extra['if'] = ifcond
>>> -    if features:
>>> -        assert isinstance(obj, dict)
>>> -        obj['features'] = [(f.name, {'if': f.ifcond}) for f in features]
>>>       if extra:
>>>           return (obj, extra)
>>>       return obj
>>> @@ -221,6 +217,11 @@ def _use_type(self, typ: QAPISchemaType) -> str:
>>>               return '[' + self._use_type(typ.element_type) + ']'
>>>           return self._name(typ.name)
>>>   
>>> +    @classmethod
>>> +    def _gen_features(cls,
>>> +                      features: List[QAPISchemaFeature]) -> List[TreeValue]:
>>> +        return [_make_tree(f.name, f.ifcond) for f in features]
>>> +
>> 
>> Ignorant question: when to use @classmethod, and when to use
>> @staticmethod?
>
> Matter of taste. My preference is to just always use @classmethod, 
> because they can be extended or referenced by subclasses.

Non-issue here, sub-classes are vanishingly unlikely.

> @staticmethod does not take a class argument, @classmethod does. Static 
> methods therefore cannot address any other classmethods, but a 
> classmethod can.
>
> I just always reach for classmethod by default.

Unused cls parameters are slightly annoying, though.

I've been using @staticmethod whenever it suffices.  Makes "this is a
function, i.e. it can't mess with the class or instances" immediately
obvious.

[...]