Re: [PATCH v2] qapi: provide a friendly string representation of QAPI classes

[Markus Armbruster]

Daniel P. Berrangé <berrange@redhat.com> writes:

> If printing a QAPI schema object for debugging we get the classname and
> a hex value for the instance:
>
>   <qapi.schema.QAPISchemaEnumType object at 0x7f0ab4c2dad0>
>   <qapi.schema.QAPISchemaObjectType object at 0x7f0ab4c2dd90>
>   <qapi.schema.QAPISchemaArrayType object at 0x7f0ab4c2df90>
>
> With this change we instead get the classname and the human friendly
> name of the QAPI type instance:
>
>   <QAPISchemaEnumType:CpuS390State>
>   <QAPISchemaObjectType:CpuInfoS390>
>   <QAPISchemaArrayType:CpuInfoFastList>

This gains the QAPI name (good), but loses the address.  The actual
address is rarely useful (when it is, you're deep in Python innards;
good luck, you'll need it).  Except they let me see which objects are
the same, and which are different.  Could that be preserved without
trouble somehow?

> Signed-off-by: Daniel P. Berrangé <berrange@redhat.com>
> ---
>
> v1 was two & half years ago:
>
>   https://mail.gnu.org/archive/html/qemu-devel/2021-03/msg01645.html

Was it my fault?  If yes, I apologize.

>  scripts/qapi/schema.py | 6 ++++++
>  1 file changed, 6 insertions(+)
>
> diff --git a/scripts/qapi/schema.py b/scripts/qapi/schema.py
> index 231ebf61ba..20ffacbdf0 100644
> --- a/scripts/qapi/schema.py
> +++ b/scripts/qapi/schema.py
> @@ -73,6 +73,12 @@ def __init__(self, name: str, info, doc, ifcond=None, 
> features=None):
>          self.features = features or []
>          self._checked = False
>  
> +    def __repr__(self):
> +        if self.name is not None:
> +            return "<%s:%s>" % (type(self).__name__, self.name)
> +        else:
> +            return "<%s>" % type(self).__name__
> +
>      def c_name(self):
>          return c_name(self.name)