nme 0.1.4

Basic usage (data serialization)

If You only need to serialize data, then you could use only JSON hooks

import json

from pydantic import BaseModel
from nme import NMEEncoder, nme_object_hook


class SampleModel(BaseModel):
    field1: int
    field2: str


data = SampleModel(field1=4, field2="abc")

with open("sample.json", "w") as f_p:
    json.dump(data, f_p, cls=NMEEncoder)

with open("sample.json") as f_p:
    data2 = json.load(f_p, object_hook=nme_object_hook)

assert data == data2

Migrations

To register this information there is register_class decorator. It has 4 parameters:

• version - version of data structure

• migration_list - list of tuple (version. migration_function).

• old_paths - list of fully qualified python paths to previous class definitions. This is to allow move class during code refactoring.

• use_parent_migrations - if True, then parent class migrations will be used.

Lets imagine that we have such code

from nme import NMEEncoder, nme_object_hook

class SampleModel(BaseModel):
    field1: int
    field_ca_1: str
    field_ca_2: float

with open("sample.json", "w") as f_p:
    json.dump(data, f_p, cls=NMEEncoder)

But there is decision to move both ca field to sub structure:

class CaModel(BaseModel):
    field_1: str
    field_2: float

class SampleModel(BaseModel):
    field1: int
    field_ca: CaModel

Then with nme code may look:

from nme import nme_object_hook, register_class

class CaModel(BaseModel):
    field_1: str
    field_2: float

def ca_migration_function(dkt):
    dkt["field_ca"] = CaModel(field1=dkt.pop("field_ca_1"),
                              field2=dkt.pop("field_ca_2"))
    return dkt

@register_class("0.0.1", [("0.0.1", ca_migration_function)])
class SampleModel(BaseModel):
    field1: int
    field_ca: CaModel

with open("sample.json") as f_p:
    data = json.load(f_p, object_hook=nme_object_hook)

Assume that there is decision to rename field1 to id. Then code may look:

from nme import nme_object_hook, register_class, rename_key

class CaModel(BaseModel):
    field_1: str
    field_2: float

def ca_migration_function(dkt):
    dkt["field_ca"] = CaModel(field1=dkt.pop("field_ca_1"),
                              field2=dkt.pop("field_ca_2"))
    return dkt

@register_class("0.0.2", [("0.0.1", ca_migration_function), ("0.0.2", rename_key("field1", "id"))])
class SampleModel(BaseModel):
    id: int
    field_ca: CaModel

with open("sample.json") as f_p:
    data = json.load(f_p, object_hook=nme_object_hook)

More examples could be found in examples section of documentation

Additional functions

• rename_key(from_key: str, to_key: str, optional=False) -> Callable[[Dict], Dict] - helper function for rename field migrations.

• update_argument(argument_name:str)(func: Callable) -> Callable - decorator to keep backward compatibility by converting dict argument to some class base on function type annotation

Additional notes

This package is extracted from PartSeg project for simplify reuse it in another projects.