Merge pull request #644 from liggitt/omitzero

Add omitzero support

Author: fxamacker

README.md

@@ -39,7 +39,7 @@ Codec passed multiple confidential security assessments in 2022.  No vulnerabili
 
 __🗜️  Data Size__
 
-Struct tags (`toarray`, `keyasint`, `omitempty`) automatically reduce size of encoded structs. Encoding optionally shrinks float64→32→16 when values fit.
+Struct tags (`toarray`, `keyasint`, `omitempty`, `omitzero`) automatically reduce size of encoded structs. Encoding optionally shrinks float64→32→16 when values fit.
 
 __:jigsaw:  Usability__
 
@@ -147,6 +147,7 @@ We can write less code by using struct tags:
 - `toarray`: encode without field names (decode back to original struct)
 - `keyasint`: encode field names as integers (decode back to original struct)
 - `omitempty`: omit empty fields when encoding
+- `omitzero`: omit zero-value fields when encoding
 
 ![alt text](https://github.com/fxamacker/images/raw/master/cbor/v2.3.0/cbor_struct_tags_api.svg?sanitize=1 "CBOR API and Go Struct Tags")
 
@@ -350,7 +351,7 @@ err = em.MarshalToBuffer(v, &buf) // encode v to provided buf
 
 ### Struct Tags
 
-Struct tags (`toarray`, `keyasint`, `omitempty`) reduce encoded size of structs.
+Struct tags (`toarray`, `keyasint`, `omitempty`, `omitzero`) reduce encoded size of structs.
 
 <details><summary> 🔎&nbsp; Example encoding 3-level nested Go struct to 1 byte CBOR</summary><p/>
 

cache.go

@@ -17,6 +17,7 @@ import (
 type encodeFuncs struct {
 	ef  encodeFunc
 	ief isEmptyFunc
+	izf isZeroFunc
 }
 
 var (
@@ -237,7 +238,7 @@ func getEncodingStructType(t reflect.Type) (*encodingStructType, error) {
 	e := getEncodeBuffer()
 	for i := 0; i < len(flds); i++ {
 		// Get field's encodeFunc
-		flds[i].ef, flds[i].ief = getEncodeFunc(flds[i].typ)
+		flds[i].ef, flds[i].ief, flds[i].izf = getEncodeFunc(flds[i].typ)
 		if flds[i].ef == nil {
 			err = &UnsupportedTypeError{t}
 			break
@@ -321,7 +322,7 @@ func getEncodingStructType(t reflect.Type) (*encodingStructType, error) {
 func getEncodingStructToArrayType(t reflect.Type, flds fields) (*encodingStructType, error) {
 	for i := 0; i < len(flds); i++ {
 		// Get field's encodeFunc
-		flds[i].ef, flds[i].ief = getEncodeFunc(flds[i].typ)
+		flds[i].ef, flds[i].ief, flds[i].izf = getEncodeFunc(flds[i].typ)
 		if flds[i].ef == nil {
 			structType := &encodingStructType{err: &UnsupportedTypeError{t}}
 			encodingStructTypeCache.Store(t, structType)
@@ -337,14 +338,14 @@ func getEncodingStructToArrayType(t reflect.Type, flds fields) (*encodingStructT
 	return structType, structType.err
 }
 
-func getEncodeFunc(t reflect.Type) (encodeFunc, isEmptyFunc) {
+func getEncodeFunc(t reflect.Type) (encodeFunc, isEmptyFunc, isZeroFunc) {
 	if v, _ := encodeFuncCache.Load(t); v != nil {
 		fs := v.(encodeFuncs)
-		return fs.ef, fs.ief
+		return fs.ef, fs.ief, fs.izf
 	}
-	ef, ief := getEncodeFuncInternal(t)
-	encodeFuncCache.Store(t, encodeFuncs{ef, ief})
-	return ef, ief
+	ef, ief, izf := getEncodeFuncInternal(t)
+	encodeFuncCache.Store(t, encodeFuncs{ef, ief, izf})
+	return ef, ief, izf
 }
 
 func getTypeInfo(t reflect.Type) *typeInfo {

doc.go

@@ -111,9 +111,14 @@ Decoding Options: https://github.com/fxamacker/cbor#decoding-options
 Struct tags like `cbor:"name,omitempty"` and `json:"name,omitempty"` work as expected.
 If both struct tags are specified then `cbor` is used.
 
-Struct tags like "keyasint", "toarray", and "omitempty" make it easy to use
+Struct tags like "keyasint", "toarray", "omitempty", and "omitzero" make it easy to use
 very compact formats like COSE and CWT (CBOR Web Tokens) with structs.
 
+The "omitzero" option omits zero values from encoding, matching
+[stdlib encoding/json behavior](https://pkg.go.dev/encoding/json#Marshal).
+When specified in the `cbor` tag, the option is always honored.
+When specified in the `json` tag, the option is honored when building with Go 1.24+.
+
 For example, "toarray" makes struct fields encode to array elements.  And "keyasint"
 makes struct fields encode to elements of CBOR map with int keys.
 

encode.go

@@ -58,17 +58,19 @@ import (
 //
 // Marshal supports format string stored under the "cbor" key in the struct
 // field's tag.  CBOR format string can specify the name of the field,
-// "omitempty" and "keyasint" options, and special case "-" for field omission.
-// If "cbor" key is absent, Marshal uses "json" key.
+// "omitempty", "omitzero" and "keyasint" options, and special case "-" for
+// field omission. If "cbor" key is absent, Marshal uses "json" key.
+// When using the "json" key, the "omitzero" option is honored when building
+// with Go 1.24+ to match stdlib encoding/json behavior.
 //
 // Struct field name is treated as integer if it has "keyasint" option in
 // its format string.  The format string must specify an integer as its
 // field name.
 //
 // Special struct field "_" is used to specify struct level options, such as
 // "toarray". "toarray" option enables Go struct to be encoded as CBOR array.
-// "omitempty" is disabled by "toarray" to ensure that the same number
-// of elements are encoded every time.
+// "omitempty" and "omitzero" are disabled by "toarray" to ensure that the
+// same number of elements are encoded every time.
 //
 // Anonymous struct fields are marshaled as if their exported fields
 // were fields in the outer struct.  Marshal follows the same struct fields
@@ -975,6 +977,7 @@ func putEncodeBuffer(e *bytes.Buffer) {
 
 type encodeFunc func(e *bytes.Buffer, em *encMode, v reflect.Value) error
 type isEmptyFunc func(em *encMode, v reflect.Value) (empty bool, err error)
+type isZeroFunc func(v reflect.Value) (zero bool, err error)
 
 func encode(e *bytes.Buffer, em *encMode, v reflect.Value) error {
 	if !v.IsValid() {
@@ -983,7 +986,7 @@ func encode(e *bytes.Buffer, em *encMode, v reflect.Value) error {
 		return nil
 	}
 	vt := v.Type()
-	f, _ := getEncodeFunc(vt)
+	f, _, _ := getEncodeFunc(vt)
 	if f == nil {
 		return &UnsupportedTypeError{vt}
 	}
@@ -1483,6 +1486,15 @@ func encodeStruct(e *bytes.Buffer, em *encMode, v reflect.Value) (err error) {
 				continue
 			}
 		}
+		if f.omitZero {
+			zero, err := f.izf(fv)
+			if err != nil {
+				return err
+			}
+			if zero {
+				continue
+			}
+		}
 
 		if !f.keyAsInt && em.fieldName == FieldNameToByteString {
 			e.Write(f.cborNameByteString)
@@ -1775,32 +1787,32 @@ var (
 	typeByteString      = reflect.TypeOf(ByteString(""))
 )
 
-func getEncodeFuncInternal(t reflect.Type) (ef encodeFunc, ief isEmptyFunc) {
+func getEncodeFuncInternal(t reflect.Type) (ef encodeFunc, ief isEmptyFunc, izf isZeroFunc) {
 	k := t.Kind()
 	if k == reflect.Pointer {
-		return getEncodeIndirectValueFunc(t), isEmptyPtr
+		return getEncodeIndirectValueFunc(t), isEmptyPtr, getIsZeroFunc(t)
 	}
 	switch t {
 	case typeSimpleValue:
-		return encodeMarshalerType, isEmptyUint
+		return encodeMarshalerType, isEmptyUint, getIsZeroFunc(t)
 
 	case typeTag:
-		return encodeTag, alwaysNotEmpty
+		return encodeTag, alwaysNotEmpty, getIsZeroFunc(t)
 
 	case typeTime:
-		return encodeTime, alwaysNotEmpty
+		return encodeTime, alwaysNotEmpty, getIsZeroFunc(t)
 
 	case typeBigInt:
-		return encodeBigInt, alwaysNotEmpty
+		return encodeBigInt, alwaysNotEmpty, getIsZeroFunc(t)
 
 	case typeRawMessage:
-		return encodeMarshalerType, isEmptySlice
+		return encodeMarshalerType, isEmptySlice, getIsZeroFunc(t)
 
 	case typeByteString:
-		return encodeMarshalerType, isEmptyString
+		return encodeMarshalerType, isEmptyString, getIsZeroFunc(t)
 	}
 	if reflect.PointerTo(t).Implements(typeMarshaler) {
-		return encodeMarshalerType, alwaysNotEmpty
+		return encodeMarshalerType, alwaysNotEmpty, getIsZeroFunc(t)
 	}
 	if reflect.PointerTo(t).Implements(typeBinaryMarshaler) {
 		defer func() {
@@ -1815,63 +1827,63 @@ func getEncodeFuncInternal(t reflect.Type) (ef encodeFunc, ief isEmptyFunc) {
 	}
 	switch k {
 	case reflect.Bool:
-		return encodeBool, isEmptyBool
+		return encodeBool, isEmptyBool, getIsZeroFunc(t)
 
 	case reflect.Int, reflect.Int8, reflect.Int16, reflect.Int32, reflect.Int64:
-		return encodeInt, isEmptyInt
+		return encodeInt, isEmptyInt, getIsZeroFunc(t)
 
 	case reflect.Uint, reflect.Uint8, reflect.Uint16, reflect.Uint32, reflect.Uint64:
-		return encodeUint, isEmptyUint
+		return encodeUint, isEmptyUint, getIsZeroFunc(t)
 
 	case reflect.Float32, reflect.Float64:
-		return encodeFloat, isEmptyFloat
+		return encodeFloat, isEmptyFloat, getIsZeroFunc(t)
 
 	case reflect.String:
-		return encodeString, isEmptyString
+		return encodeString, isEmptyString, getIsZeroFunc(t)
 
 	case reflect.Slice:
 		if t.Elem().Kind() == reflect.Uint8 {
-			return encodeByteString, isEmptySlice
+			return encodeByteString, isEmptySlice, getIsZeroFunc(t)
 		}
 		fallthrough
 
 	case reflect.Array:
-		f, _ := getEncodeFunc(t.Elem())
+		f, _, _ := getEncodeFunc(t.Elem())
 		if f == nil {
-			return nil, nil
+			return nil, nil, nil
 		}
-		return arrayEncodeFunc{f: f}.encode, isEmptySlice
+		return arrayEncodeFunc{f: f}.encode, isEmptySlice, getIsZeroFunc(t)
 
 	case reflect.Map:
 		f := getEncodeMapFunc(t)
 		if f == nil {
-			return nil, nil
+			return nil, nil, nil
 		}
-		return f, isEmptyMap
+		return f, isEmptyMap, getIsZeroFunc(t)
 
 	case reflect.Struct:
 		// Get struct's special field "_" tag options
 		if f, ok := t.FieldByName("_"); ok {
 			tag := f.Tag.Get("cbor")
 			if tag != "-" {
 				if hasToArrayOption(tag) {
-					return encodeStructToArray, isEmptyStruct
+					return encodeStructToArray, isEmptyStruct, isZeroFieldStruct
 				}
 			}
 		}
-		return encodeStruct, isEmptyStruct
+		return encodeStruct, isEmptyStruct, getIsZeroFunc(t)
 
 	case reflect.Interface:
-		return encodeIntf, isEmptyIntf
+		return encodeIntf, isEmptyIntf, getIsZeroFunc(t)
 	}
-	return nil, nil
+	return nil, nil, nil
 }
 
 func getEncodeIndirectValueFunc(t reflect.Type) encodeFunc {
 	for t.Kind() == reflect.Pointer {
 		t = t.Elem()
 	}
-	f, _ := getEncodeFunc(t)
+	f, _, _ := getEncodeFunc(t)
 	if f == nil {
 		return nil
 	}
@@ -1987,3 +1999,96 @@ func float32NaNFromReflectValue(v reflect.Value) float32 {
 	f32 := p.Convert(reflect.TypeOf((*float32)(nil))).Elem().Interface().(float32)
 	return f32
 }
+
+type isZeroer interface {
+	IsZero() bool
+}
+
+var isZeroerType = reflect.TypeOf((*isZeroer)(nil)).Elem()
+
+// getIsZeroFunc returns a function for the given type that can be called to determine if a given value is zero.
+// Types that implement `IsZero() bool` are delegated to for non-nil values.
+// Types that do not implement `IsZero() bool` use the reflect.Value#IsZero() implementation.
+// The returned function matches behavior of stdlib encoding/json behavior in Go 1.24+.
+func getIsZeroFunc(t reflect.Type) isZeroFunc {
+	// Provide a function that uses a type's IsZero method if defined.
+	switch {
+	case t == nil:
+		return isZeroDefault
+	case t.Kind() == reflect.Interface && t.Implements(isZeroerType):
+		return isZeroInterfaceCustom
+	case t.Kind() == reflect.Pointer && t.Implements(isZeroerType):
+		return isZeroPointerCustom
+	case t.Implements(isZeroerType):
+		return isZeroCustom
+	case reflect.PointerTo(t).Implements(isZeroerType):
+		return isZeroAddrCustom
+	default:
+		return isZeroDefault
+	}
+}
+
+// isZeroInterfaceCustom returns true for nil or pointer-to-nil values,
+// and delegates to the custom IsZero() implementation otherwise.
+func isZeroInterfaceCustom(v reflect.Value) (bool, error) {
+	kind := v.Kind()
+
+	switch kind {
+	case reflect.Chan, reflect.Func, reflect.Map, reflect.Pointer, reflect.Interface, reflect.Slice:
+		if v.IsNil() {
+			return true, nil
+		}
+	}
+
+	switch kind {
+	case reflect.Interface, reflect.Pointer:
+		if elem := v.Elem(); elem.Kind() == reflect.Pointer && elem.IsNil() {
+			return true, nil
+		}
+	}
+
+	return v.Interface().(isZeroer).IsZero(), nil
+}
+
+// isZeroPointerCustom returns true for nil values,
+// and delegates to the custom IsZero() implementation otherwise.
+func isZeroPointerCustom(v reflect.Value) (bool, error) {
+	if v.IsNil() {
+		return true, nil
+	}
+	return v.Interface().(isZeroer).IsZero(), nil
+}
+
+// isZeroCustom delegates to the custom IsZero() implementation.
+func isZeroCustom(v reflect.Value) (bool, error) {
+	return v.Interface().(isZeroer).IsZero(), nil
+}
+
+// isZeroAddrCustom delegates to the custom IsZero() implementation of the addr of the value.
+func isZeroAddrCustom(v reflect.Value) (bool, error) {
+	if !v.CanAddr() {
+		// Temporarily box v so we can take the address.
+		v2 := reflect.New(v.Type()).Elem()
+		v2.Set(v)
+		v = v2
+	}
+	return v.Addr().Interface().(isZeroer).IsZero(), nil
+}
+
+// isZeroDefault calls reflect.Value#IsZero()
+func isZeroDefault(v reflect.Value) (bool, error) {
+	if !v.IsValid() {
+		// v is zero value
+		return true, nil
+	}
+	return v.IsZero(), nil
+}
+
+// isZeroFieldStruct is used to determine whether to omit toarray structs
+func isZeroFieldStruct(v reflect.Value) (bool, error) {
+	structType, err := getEncodingStructType(v.Type())
+	if err != nil {
+		return false, err
+	}
+	return len(structType.fields) == 0, nil
+}

encode_map.go

@@ -65,8 +65,8 @@ func (me *mapKeyValueEncodeFunc) encodeKeyValues(e *bytes.Buffer, em *encMode, v
 }
 
 func getEncodeMapFunc(t reflect.Type) encodeFunc {
-	kf, _ := getEncodeFunc(t.Key())
-	ef, _ := getEncodeFunc(t.Elem())
+	kf, _, _ := getEncodeFunc(t.Key())
+	ef, _, _ := getEncodeFunc(t.Elem())
 	if kf == nil || ef == nil {
 		return nil
 	}