Direct access to attributes

[romanchyla]

In [9]: r.bibcode = 'foo'

In [10]: r
Out[10]: <adsmsg.bibrecord.BibRecord at 0x7f9f5ed9acd0>

In [11]: str(r)
Out[11]: '<adsmsg.bibrecord.BibRecord object at 0x7f9f5ed9acd0>'

In [12]: r.__str__()
Out[12]: '<adsmsg.bibrecord.BibRecord object at 0x7f9f5ed9acd0>'

In [13]: r.bibcode
Out[13]: 'foo'

In [14]: r.serialize()
---------------------------------------------------------------------------
EncodeError                               Traceback (most recent call last)
<ipython-input-14-6d620e231470> in <module>()
----> 1 r.serialize()

/dvt/workspace2/ADSPipelineMsg/adsmsg/msg.pyc in serialize(self)
     26         Returns a serialized protocol buffer message
     27         """
---> 28         return self._data.SerializeToString()
     29 
     30     def is_valid(self):

EncodeError: Message adsmsg.BibRecord is missing required fields: bibcode,JSON_fingerprint,metadata,text

so I was expecting the following to work:

r = BibRecord()
r.bibcode = 'foo'
r.serialize()

i doesn't because Msg is storing things in r._data

Also, it would be really really useful to be able to do:

r = BibRecord(bibcode='foo') # and by extension r = BibRecord(**{'bibcode': 'foo', ....})

[marblestation]

Protoc compiler does not generate a class but a metaclass, we cannot just inherit from it. If we want to have our own class, we need to use composition instead.

Direct access to attributes don't seem possible and initialization via kwargs is complex, because you can have an atribute that is a list of attributes that contain other types and lists. It might be doable but I'm not sure that the time investment is proportional to the benefits.

[romanchyla]

I will try to cook something, if devil is on the details I might fail but we'll give it a try. please keep the issue open for now

…

On Jun 12, 2017 3:26 PM, "marblestation" ***@***.***> wrote: Protoc compiler does not generate a class but a metaclass, we cannot just inherit from it. If we want to have our own class, we need to use composition instead. Direct access to attributes don't seem possible and initialization via kwargs is complex, because you can have an atribute that is a list of attributes that contain other types and lists. It might be doable but I'm not sure that the time investment is proportional to the benefits. — You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub <#5 (comment)>, or mute the thread <https://github.com/notifications/unsubscribe-auth/AAZIkmO9M9pWpgFrrk6fsI7n0shHApnWks5sDY-ugaJpZM4N1x9r> .