The real solution seems to be to build the objects from the bottom up. 
What's going on is that nested messages are "immmutable" objects from the 
point of view of their parent object. You can't just set them from the top 
down, but you can build the object from the bottom up. The error messages 
and protobug docs don't use the words 'immutable' or 'mutable' but their 
use would explain a lot (at least to me)

For example:

syntax = "proto3";

message SubTask {
   string secret = 2;

message Task {
   int32 id = 1;
   string msg = 2;
   SubTask mission = 3;

message TaskInfo {
   Task task = 1;

if we want to create a TaskInfo object dynamically, we first need to create 
a SubTask object, then create a Task object passing in the SubTask object 
in its constructor, then create the TaskInfo object passing in the Task 
object in that constructor

args = { "secret": "secret message"}
subtask = SubTask(**args)

args = { "id": 100, "msg", "some message", "mission": subtask }
task = Task(**args)

args= {"task": task}
task_info = TaskInfo(**args)

In this way you can also use variables for the field names in the **args 
dictionary when creating objects dynamically for any random proto, where 
you don't know the field names apriori but discover them. Create the 
sub_messages first, then their parent, then its parent and so on, from the 
bottom up

On Friday, March 3, 2017 at 2:44:54 PM UTC-8, Tom Lichtenberg wrote:
> One other thing I am now trying is to take the task_instance object and 
> call SerializeToString() on it, then use 
> task_info_instance.MergeFromString(serialized_task_instance)
> this is not raising any exception and does seem to be setting the field, 
> but I'm not entirely convinced yet
> On Friday, March 3, 2017 at 1:09:09 PM UTC-8, Tom Lichtenberg wrote:
>> I’m trying to dynamically create objects from discovery of pb2 generated 
>> protobuf files and assign appropriate values to their fields. If the field 
>> types are simple, this is straightforward - e.g. for TYPE_BOOL, 
>> setattr(obj, “field_name”, True) works fine - but if the field is 
>> TYPE_MESSAGE, making an instance of that field and using setattr(obj, 
>> “field_name”, instance) results in an error:
>>       Assignment not allowed to composite field "task" in protocol 
>> message object.
>> from python2.7/site-packages/google/protobuf/internal/", 
>> line 736, in setter
>> sample code: dummy/test_dummy.proto
>> ============================
>> syntax = "proto3";
>> message task {
>>    int32 id = 1;
>>    string msg = 2;
>> }
>> message task_info {
>>    task task = 1;
>> }
>> from this I used protoc to generate a
>> For example, I want to create a task_info object, so by exploring the 
>> imported pb2 module, I find the task_info descriptor and create an instance 
>> of it. I see that its field ‘task’ is of TYPE_MESSAGE so I create an 
>> instance of that as well and call setattr(task_instance, “id”, 1) and 
>> setattr(task_instance, “msg”, “some message”) to populate it, and then go 
>> back and call setattr(task_info_instance, “task”, task_instance). Here I 
>> get the error:  Assignment not allowed to composite field "task" in 
>> protocol message object.

>> programatically I can import this module and assign values well enough, 
>> for example in the python shell:
>> >> import importlib
>> >> oobj = importlib.import_module("dummy.dummy.test_dummy_pb2", 
>> package=None)
>> >> obj = getattr(oobj, "task_info")
>> >> task_info_instance = obj()
>> >> = 1
>> >> task_info_instance.task.msg = "some message"
>> >> print task_info_instance
>> task {
>>   id: 1
>>   msg: "some message"
>> }
>> But dynamically I can’t explicitly assign it like that. I would need 
>> something like a setattr() function that can set an object’s composite 
>> object field. I’m hoping that someone can shed some light on this for me.

You received this message because you are subscribed to the Google Groups 
"Protocol Buffers" group.
To unsubscribe from this group and stop receiving emails from it, send an email 
To post to this group, send email to
Visit this group at
For more options, visit

Reply via email to