Accept filters from request arguments and not only metadata

[AMontagu]

In DSG metadata is what used to replace the query parameters systems. It is very flexible as it's not in the proto file.

The main inconvenients are:

• The metadata are not binary serialized so passing a lot of data as filters may result in poor performance
• They not exported in the proto so not documented by default.

To contourn that we want to look about best practices and implement it.

Think to update doc after.

[markdoerr]

@AMontagu,
I fully support this reature request.

Right now, I am using a "decorator" workaround, e.g.:

# filter_decorator.py
import functools
import json

import lara_django_samples_grpc.v1.lara_django_samples_pb2 as lara_django_samp_pb2

def dict_filter(func):
    @functools.wraps(func)
    def wrapper_filter(filter_dict : dict={}, **kwargs):
        metadata = (("filters", (json.dumps(filter_dict))),)
        return func.List(lara_django_samp_pb2.SampleListRequest(), metadata=metadata, **kwargs)
    return wrapper_filter

and then I can do something like:

from filter_decorator import dict_filter

channel = grpc.insecure_channel(f'{grpc_server_host}:{grpc_server_port}')
sample_client = lara_django_samp_pb2_grpc.SampleControllerStub(channel)
list_filter = dict_filter(sample_client)

filter_dict = { "name__contains": "term-to-search"}
res = list_filter(filter_dict=filter_dict)
print(res)

Which is more readable, but still has the problem of using metadata to send the filter.

[AMontagu]

The first part of this issue is completed.
The second part is to precisely detailled the filters and pagination fields names and types instead of using a struct.
This improvement may happen in late 2024.