Re: [Qemu-devel] [RFC PATCH] qapi: Use callback to determine visit filtering

[Eric Blake <eblake@redhat.com>]

[-- Attachment #1: Type: text/plain, Size: 1493 bytes --]

On 10/01/2015 12:12 AM, Markus Armbruster wrote:

>>
>>> We can always add an indirection: instead of parameterizing a fixed
>>> predicate (ignore and isinstance(entity, ignore)) with a type ignore, we
>>> could pass a predicate, i.e. a function mapping entity to bool.
>>
>> Like this?  Generates the same code before and after the
>> patch.  I'm open to suggestions on any way to make it more
>> idiomatic python, althouth it at least passed pep8.  In
>> particular, I'm wondering if the predicate should have its
>> sense reversed (pass False to skip, True to visit).
>>
>> I'll probably drop the 'assert info' lines in the two
>> visit_object_type() calls (I put it there to make sure the
>> predicate was working).
> 
> Yes, please.

Returning True to visit and False to skip was easier to rationalize
about, so I've made that change in my local tree.


> 
> I think this turned out rather nicely.
> 
> Can we go one step further?  Unconditionally call visitor.visit_filter()
> here, define the pass-everything filter in QAPISchemaVisitor, override
> it as needed.
> 
> Name it visit_filter_out() to make the sense of the return value
> obvious?

Oh, nice idea. Then we don't even have to return it during visit_begin()
- we just blindly call it.  Will work that into my local tree, and it
will be ready when I post (the next subset) of v6.

-- 
Eric Blake   eblake redhat com    +1-919-301-3266
Libvirt virtualization library http://libvirt.org


[-- Attachment #2: OpenPGP digital signature --]
[-- Type: application/pgp-signature, Size: 604 bytes --]