protobuf.sh

Decode protocol buffer wire format with POSIX Shell
This script is used for quickly glancing the content of protobuf messages during debugging, and it does not provide message parsing based on .proto files.

Dependency

xxd

Usage

A simple hello world example

$ printf "\x08\xd2\xfe\x06\x12\x0bHello world\x1a\x04\x08\xc2\x96\x75" | ./pbdecode.sh
{"1":114514,"2":"Hello world","3":{"1":1919810}}

More usages

# load pb from disk
cat demo_msg.pb | pbdecode.sh

# easily debug API with curl
curl http://example.com | pbdecode.sh

Note for bytes, string, embedded messages

Bytes, string, embedded messages are encoded to bytes.
Without .proto file, it's impossible to distinguish between bytes, string and embedded messages.
pbdecode.sh will try treating bytes as embedded messages and decoding recursively.
If a bytes field cannot be decoded, pbdecode.sh will treat it as a string.

Note for repeated fields

Varint, fixed64, fixed32

[packed=true] option is default for proto3.
Repeated varints, fixed64s and fixed32s will be encoded to bytes. e.g.
Single int32

message A {
  int32 num = 1;
}

{ "num": 64 } => x08(field:1, type:varint) x40(64)

Repeated int32

message B {
  repeated int32 num = 1;
}

{ "num": [64, 65] } => x0a(field:1, type:bytes) x02(length:2) x40(64) x41(65)

pbdecode.sh will treat packed repeated fields as bytes.

Message

Ordinary (not packed) repeated fields emit one record for every element of the field.

Due to implementation, output JSON will also emit one record for every element of the field.

ECMA-404: 6 Objects
The JSON syntax does not impose any restrictions on the strings used as names, does not require that name strings be unique...

e.g.

message C {
  repeated D msgs = 1;
}

message D {
  int32 num = 1;
}

x0a(field:1, type:bytes) x02(length:2) x08(field:1, type:varint) x40(64)
x0a(field:1, type:bytes) x02(length:2) x08(field:1, type:varint) x41(65)
=> {"1":{"1":64},"1":{"1":65}}