Fields — Proto Plus for Python documentation (original) (raw)

Fields are assigned using the Field class, instantiated within aMessage declaration.

Fields always have a type (either a primitive, a message, or an enum) and anumber.

import proto

class Composer(proto.Message): given_name = proto.Field(proto.STRING, number=1) family_name = proto.Field(proto.STRING, number=2)

class Song(proto.Message): composer = proto.Field(Composer, number=1) title = proto.Field(proto.STRING, number=2) lyrics = proto.Field(proto.STRING, number=3) year = proto.Field(proto.INT32, number=4)

For messages and enums, assign the message or enum class directly (as shown in the example above).

Note

For messages declared in the same module, it is also possible to use a string with the message class’ name if the class is not yet declared, which allows for declaring messages out of order or with circular references.

Repeated fields¶

Some fields are actually repeated fields. In protocol buffers, repeated fields are generally equivalent to typed lists. In protocol buffers, these are declared using the repeated keyword:

message Album { repeated Song songs = 1; string publisher = 2; }

Declare them in Python using the RepeatedField class:

class Album(proto.Message): songs = proto.RepeatedField(Song, number=1) publisher = proto.Field(proto.STRING, number=2)

Note

Elements must be appended individually for repeated fields of struct.Value.

class Row(proto.Message): values = proto.RepeatedField(proto.MESSAGE, number=1, message=struct.Value,)

row = Row() values = [struct_pb2.Value(string_value="hello")] for v in values: row.values.append(v)

Direct assignment will result in an error.

class Row(proto.Message): values = proto.RepeatedField(proto.MESSAGE, number=1, message=struct.Value,)

row = Row() row.values = [struct_pb2.Value(string_value="hello")] Traceback (most recent call last): File "", line 1, in File "/usr/local/google/home/busunkim/github/python-automl/.nox/unit-3-8/lib/python3.8/site-packages/proto/message.py", line 543, in setattr self._pb.MergeFrom(self._meta.pb(**{key: pb_value})) TypeError: Value must be iterable

Map fields¶

Similarly, some fields are map fields. In protocol buffers, map fields are equivalent to typed dictionaries, where the keys are either strings or integers, and the values can be any type. In protocol buffers, these use a special map syntax:

message Album { map<uint32, Song> track_list = 1; string publisher = 2; }

Declare them in Python using the MapField class:

class Album(proto.Message): track_list = proto.MapField(proto.UINT32, Song, number=1) publisher = proto.Field(proto.STRING, number=2)

Oneofs (mutually-exclusive fields)¶

Protocol buffers allows certain fields to be declared as mutually exclusive. This is done by wrapping fields in a oneof syntax:

import "google/type/postal_address.proto";

message AlbumPurchase { Album album = 1; oneof delivery { google.type.PostalAddress postal_address = 2; string download_uri = 3; } }

When using this syntax, protocol buffers will enforce that only one of the given fields is set on the message, and setting a field within the oneof will clear any others.

Declare this in Python using the oneof keyword-argument, which takes a string (which should match for all fields within the oneof):

from google.type.postal_address import PostalAddress

class AlbumPurchase(proto.Message): album = proto.Field(Album, number=1) postal_address = proto.Field(PostalAddress, number=2, oneof='delivery') download_uri = proto.Field(proto.STRING, number=3, oneof='delivery')

Warning

oneof fields must be declared consecutively, otherwise the C implementation of protocol buffers will reject the message. They need not have consecutive field numbers, but they must be declared in consecutive order.

Warning

If a message is constructed with multiple variants of a single oneof passed to its constructor, the last keyword/value pair passed will be the final value set.

This is consistent with PEP-468, which specifies the order that keyword args are seen by called functions, and with the regular protocol buffers runtime, which exhibits the same behavior.

Example:

import proto

class Song(proto.Message): name = proto.Field(proto.STRING, number=1, oneof="identifier") database_id = proto.Field(proto.STRING, number=2, oneof="identifier")

s = Song(name="Canon in D minor", database_id="b5a37aad3") assert "database_id" in s and "name" not in s

s = Song(database_id="e6aa708c7e", name="Little Fugue") assert "name" in s and "database_id" not in s

Optional fields¶

All fields in protocol buffers are optional, but it is often necessary to check for field presence. Sometimes legitimate values for fields can be falsy, so checking for truthiness is not sufficient. Proto3 v3.12.0 added theoptional keyword to field descriptions, which enables a mechanism for checking field presence.

In proto plus, fields can be marked as optional by passing optional=Truein the constructor. The message class then gains a field of the same name that can be used to detect whether the field is present in message instances.

s = Song( ... composer={'given_name': 'Johann', 'family_name': 'Pachelbel'}, ... title='Canon in D', ... year=1680, ... genre=Genre.CLASSICAL, ... ) Song.performer in s False s.performer = 'Brahms' Song.performer in s True del s.performer Song.performer in s False s.performer = "" # The mysterious, unnamed composer Song.performer in s True

Under the hood, fields marked as optional are implemented via a synthetic one-variant oneof. See the protocolbuffers documentation for more information.