Protocol Buffers Encoding Reference
Protobuf wire 格式化: varint, 长度-delimited, 固定-宽度, 和字段 tag 编码.
Wire Types
Protocol Buffers encode each field as a tag followed by a value, where the tag contains the field number and wire type. The wire type determines how the value is read: varint, 64-bit fixed, length-delimited, 32-bit fixed, or deprecated group. This compact binary format avoids field names on the wire, relying on field numbers for identification. Unknown fields can be skipped using the wire type without a schema.
- varint
- Variable-length integer, wire type 0
- fixed64
- 8-byte little-endian, wire type 1
- length-delimited
- Length prefix then bytes, wire type 2
- fixed32
- 4-byte little-endian, wire type 5
int32, int64, boolfixed64, doublestring, bytes, messagefixed32, floatVarint Encoding
Varints encode integers in groups of 7 bits with the high bit indicating whether more bytes follow. Small values use a single byte, while large 64-bit values use up to 10 bytes. This makes protobuf efficient for typical small integer values. Signed integers use zigzag encoding to avoid the 10-byte penalty for negative numbers, mapping signed to unsigned before varint encoding.
Schema Evolution
Protobuf supports forward and backward compatibility when fields are added or removed. New optional fields are ignored by old parsers as unknown fields, and old fields can be reserved to prevent reuse. Changing a field type or number breaks compatibility. Repeated fields can be packed to save space when the elements are scalar. Following the compatibility rules lets services evolve their schemas without coordinated deployments.