typegen

cbor-gen

DAOT Labs' fork of whyrusleeping/cbor-gen.

Some basic utilities to generate fast path cbor codecs for your types.

Usage

See testgen/main.go.

New features of this fork

Adds support for types: uint16, uint32, int8, int16, int32

Sort map keys according to RFC7049 & DAG-CBOR strict ordering

This adds proper RFC7049 map key sorting, to both bare maps and structs in map representation. The other DAG-CBOR codec implementations strictly sort by length first, then bytes, as per https://github.com/ipld/specs/blob/master/block-layer/codecs/dag-cbor.md#strictness. The original cbor-gen implementation will render CBOR that won't produce stable CIDs with a round-trip using another DAG-CBOR encoder.

For more details, see: https://github.com/whyrusleeping/cbor-gen/pull/42

Embedded struct flattening

This fork additionally support an option flattenEmbeddedStruct in WriteTupleEncodersToFile and WriteMapEncodersToFile functions to flatten embedded structs in serialization.

For example, we have following structs from testing/flatten_tuple/types.go:

type EmbeddingStructOne struct {
	SimpleTypeOne
	*SimpleTypeTwo
	Foo          string
	Stuff        *SimpleTypeTwo
	Others       []uint64
	SignedOthers []int64
}

type EmbeddingStructTwo struct {
	SimpleTypeOne
	*EmbeddingStructOne
	Foo          string
	Value        uint64
	Stuff        *SimpleTypeTwo
	Others       []uint64
	SignedOthers []int64
	Test         [][]byte
	Dog          string
	Numbers      []NamedNumber
}

type EmbeddingStructThree struct {
	*EmbeddingStructTwo
	Foo          string
	Value        uint64
	Binary       []byte
	Stuff        *SimpleTypeTwo
	Others       []uint64
	SignedOthers []int64
	Test         [][]byte
	Dog          string
	Numbers      []NamedNumber
	Pizza        *uint64
	PointyPizza  *NamedNumber
	Arrrrrghay   [testing.Thingc]SimpleTypeOne
}

type FlatStruct struct {
	Signed  int64
	Foo     string
	Binary  []byte
	NString NamedString
	Value   uint64
}

type EmbeddedStruct struct {
	Value  uint64
	Foo    string
	Binary []byte
}

type StructValueFieldStruct struct {
	Signed  int64
	EmbeddedStruct EmbeddedStruct
	Binary  []byte
	NString NamedString
	Value   uint64
}

type StructPointerFieldStruct struct {
	Signed  int64
	EmbeddedStruct *EmbeddedStruct
	Binary  []byte
	NString NamedString
	Value   uint64
}

type EmbedByValueStruct struct {
	Signed  int64
	EmbeddedStruct
	Binary  []byte
	NString NamedString
	Value   uint64
}

type EmbedByPointerStruct struct {
	Signed  int64
	*EmbeddedStruct
	Binary  []byte
	NString NamedString
	Value   uint64
}

With the original cbor-gen, StructValueFieldStruct, StructPointerFieldStruct, EmbedByValueStruct and EmbedByPointerStruct will all be serialized into the same byte sequence, which is different from that of FlatStruct.

With this fork and flattenEmbeddedStruct option set to true, StructValueFieldStruct and StructPointerFieldStruct will all be serialized into the same byte sequence as that of FlatStruct, which is different from that of StructValueFieldStruct and StructPointerFieldStruct.

The same flattening applies to recursively embedded structs with arbitrary levels of embedding.

Note: For flattenEmbeddedStruct option to work, all structs recursively embedded by pointer (any levels) in the root struct need to generate their InitNilEmbeddedStruct methods with cbor-gen and flattenEmbeddedStruct option also set to true, although their generated MarshalCBOR and UnmarshalCBOR methods won't be used when marshalling and unmarshalling the root struct. Alternatively, you can write the InitNilEmbeddedStruct methods by hand, see Inniter.InitNilEmbeddedStruct in utils.go for description and testing/flatten_tuple/cbor_gen.go for examples.

type Initter interface {
	// InitNilEmbeddedStruct prepare the struct for marshalling & unmarshalling by
	// recursively initialize the structs embedded by pointer to their zero values.
	InitNilEmbeddedStruct()
}

Specifying struct field order

This fork additionally support an option fieldOrder in WriteTupleEncodersToFile to specify the orders of structs' fields. If specified (not nil), fieldOrder is a string slice of struct field names in the order they will be serialized, the remaining field names not found in the slice will be serialized after those specified in the original order of appearance in the code.

For example, we have the following two structs from testing/flatten_tuple/types.go:

type FlatStruct struct {
	Signed  int64
	Foo     string
	Binary  []byte
	NString NamedString
	Value   uint64
}

type ReorderedFlatStruct struct {
	Foo     string
	Value   uint64
	Binary  []byte
	Signed  int64
	NString NamedString
}

They have the same fields with different orders. If we generate the MarshalCBOR and UnmarshalCBOR with the following options, they will be serialized into and deserialized from the same byte sequence:

cbg.WriteTupleEncodersToFile("testing/flatten_tuple/cbor_gen.go",
    "flatten_tuple", true, nil,
    flatten_tuple.FlatStruct{},
)

cbg.WriteTupleEncodersToFile("testing/flatten_tuple/cbor_gen_reordered.go",
    "flatten_tuple", true, []string{"Signed", "Foo", "Binary", "NString", "Value"},
    flatten_tuple.ReorderedFlatStruct{},
)

License

MIT