xml: add preserve_namespaces option to retain XML namespace prefixes

internal/impl/xml/bloblang.go

@@ -42,8 +42,15 @@ func init() {
 				`{"doc":"<root><title>This is a title</title><number id=\"99\">123</number><bool>True</bool></root>"}`,
 				`{"doc":{"root":{"bool":true,"number":{"#text":123,"-id":99},"title":"This is a title"}}}`,
 			}).
+			Example("Parse XML preserving namespace prefixes so the original document is reconstructable", `root.doc = this.doc.parse_xml(preserve_namespaces: true)`, [2]string{
+				`{"doc":"<root xmlns:dc=\"http://my.namespace/dc\"><dc:title>Hello</dc:title></root>"}`,
+				`{"doc":{"root":{"-xmlns:dc":"http://my.namespace/dc","dc:title":"Hello"}}}`,
+			}).
 			Param(bloblang.NewBoolParam("cast").
 				Description("Whether to automatically cast numeric and boolean string values to their proper types. When false, all values remain as strings.").
+				Optional().Default(false)).
+			Param(bloblang.NewBoolParam("preserve_namespaces").
+				Description("Whether to preserve XML namespace prefixes on element and attribute keys, and retain xmlns declarations as attributes. When disabled, namespace prefixes are stripped.").
 				Optional().Default(false)),
 		func(args *bloblang.ParsedParams) (bloblang.Method, error) {
 			castOpt, err := args.GetOptionalBool("cast")
@@ -54,8 +61,24 @@ func init() {
 			if castOpt != nil {
 				cast = *castOpt
 			}
+			preserveOpt, err := args.GetOptionalBool("preserve_namespaces")
+			if err != nil {
+				return nil, err
+			}
+			preserveNS := false
+			if preserveOpt != nil {
+				preserveNS = *preserveOpt
+			}
 			return bloblang.BytesMethod(func(xmlBytes []byte) (any, error) {
-				xmlObj, err := ToMap(xmlBytes, cast)
+				var (
+					xmlObj map[string]any
+					err    error
+				)
+				if preserveNS {
+					xmlObj, err = ToMapPreserveNS(xmlBytes, cast)
+				} else {
+					xmlObj, err = ToMap(xmlBytes, cast)
+				}
 				if err != nil {
 					return nil, fmt.Errorf("parsing value as XML: %w", err)
 				}

internal/impl/xml/bloblang_test.go

@@ -48,6 +48,23 @@ func TestParseXML(t *testing.T) {
 			args:   `true`,
 			exp:    map[string]any{"root": map[string]any{"bool": true, "number": map[string]any{"#text": float64(123), "-id": float64(99)}, "title": "This is a title"}},
 		},
+		{
+			name:   "preserve namespaces for issue 3928",
+			target: `<root xmlns:dc="http://my.namespace/dc"><dc:title>Hello</dc:title></root>`,
+			args:   `preserve_namespaces: true`,
+			exp: map[string]any{"root": map[string]any{
+				"-xmlns:dc": "http://my.namespace/dc",
+				"dc:title":  "Hello",
+			}},
+		},
+		{
+			name:   "preserve namespaces is opt-in",
+			target: `<root xmlns:dc="http://my.namespace/dc"><dc:title>Hello</dc:title></root>`,
+			exp: map[string]any{"root": map[string]any{
+				"-dc":   "http://my.namespace/dc",
+				"title": "Hello",
+			}},
+		},
 	}
 
 	for _, test := range testCases {

internal/impl/xml/package.go

@@ -19,7 +19,12 @@
 package xml
 
 import (
+	"bytes"
 	"encoding/xml"
+	"fmt"
+	"io"
+	"strconv"
+	"strings"
 
 	"github.com/clbanning/mxj/v2"
 	"golang.org/x/net/html/charset"
@@ -41,3 +46,150 @@ func ToMap(xmlBytes []byte, cast bool) (map[string]any, error) {
 	}
 	return map[string]any(root), nil
 }
+
+// ToMapPreserveNS parses a byte slice as XML with namespace prefixes preserved
+// on element and attribute keys (e.g. "<dc:title>" becomes the key "dc:title")
+// and with xmlns declarations retained as attributes, so the original XML is
+// reconstructable from the resulting JSON. The output shape otherwise matches
+// [ToMap]: attributes are prefixed with "-", mixed text content uses the
+// "#text" key, and repeated elements are collected into arrays.
+func ToMapPreserveNS(xmlBytes []byte, cast bool) (map[string]any, error) {
+	dec := xml.NewDecoder(bytes.NewReader(xmlBytes))
+	dec.Strict = false
+	dec.CharsetReader = charset.NewReaderLabel
+
+	for {
+		tok, err := dec.Token()
+		if err == io.EOF {
+			return nil, fmt.Errorf("xml: no root element found")
+		}
+		if err != nil {
+			return nil, err
+		}
+		if se, ok := tok.(xml.StartElement); ok {
+			key, val, err := parseElementNS(dec, se, map[string]string{}, cast)
+			if err != nil {
+				return nil, err
+			}
+			return map[string]any{key: val}, nil
+		}
+	}
+}
+
+// collectPrefixDecls records URI→prefix mappings from xmlns:* attributes on an
+// element into the provided scope map.
+func collectPrefixDecls(se xml.StartElement, scope map[string]string) {
+	for _, a := range se.Attr {
+		if a.Name.Space == "xmlns" {
+			scope[a.Value] = a.Name.Local
+		}
+	}
+}
+
+// qnameWithPrefix returns "prefix:local" when a prefix is known for the
+// namespace URI; when the namespace was never bound via xmlns, Go's decoder
+// leaves Name.Space as the raw prefix string which is used directly.
+func qnameWithPrefix(n xml.Name, scope map[string]string) string {
+	if n.Space == "" {
+		return n.Local
+	}
+	if p, ok := scope[n.Space]; ok {
+		return p + ":" + n.Local
+	}
+	return n.Space + ":" + n.Local
+}
+
+func parseElementNS(dec *xml.Decoder, se xml.StartElement, parent map[string]string, cast bool) (string, any, error) {
+	scope := make(map[string]string, len(parent)+len(se.Attr))
+	for k, v := range parent {
+		scope[k] = v
+	}
+	collectPrefixDecls(se, scope)
+
+	out := map[string]any{}
+	for _, a := range se.Attr {
+		var key string
+		isNSDecl := false
+		switch {
+		case a.Name.Space == "xmlns":
+			key = "-xmlns:" + a.Name.Local
+			isNSDecl = true
+		case a.Name.Space == "" && a.Name.Local == "xmlns":
+			key = "-xmlns"
+			isNSDecl = true
+		default:
+			key = "-" + qnameWithPrefix(a.Name, scope)
+		}
+		if isNSDecl {
+			out[key] = a.Value
+		} else {
+			out[key] = castString(a.Value, cast)
+		}
+	}
+
+	var text strings.Builder
+	for {
+		tok, err := dec.Token()
+		if err != nil {
+			return "", nil, err
+		}
+		switch t := tok.(type) {
+		case xml.StartElement:
+			k, v, err := parseElementNS(dec, t, scope, cast)
+			if err != nil {
+				return "", nil, err
+			}
+			if existing, ok := out[k]; ok {
+				if arr, isArr := existing.([]any); isArr {
+					out[k] = append(arr, v)
+				} else {
+					out[k] = []any{existing, v}
+				}
+			} else {
+				out[k] = v
+			}
+		case xml.CharData:
+			text.Write(t)
+		case xml.EndElement:
+			s := strings.TrimSpace(text.String())
+			key := qnameWithPrefix(se.Name, scope)
+			if len(out) == 0 {
+				return key, castString(s, cast), nil
+			}
+			if s != "" {
+				out["#text"] = castString(s, cast)
+			}
+			return key, out, nil
+		}
+	}
+}
+
+// castString mirrors clbanning/mxj's default cast order when casting is
+// enabled: int → float → bool, with NaN/Inf left as strings.
+func castString(s string, cast bool) any {
+	if !cast || s == "" {
+		return s
+	}
+	switch strings.ToLower(s) {
+	case "nan", "inf", "-inf":
+		return s
+	}
+	if i, err := strconv.ParseInt(s, 10, 64); err == nil {
+		return i
+	}
+	if u, err := strconv.ParseUint(s, 10, 64); err == nil {
+		return u
+	}
+	if f, err := strconv.ParseFloat(s, 64); err == nil {
+		return f
+	}
+	if len(s) < 6 {
+		switch s[:1] {
+		case "t", "T", "f", "F":
+			if b, err := strconv.ParseBool(s); err == nil {
+				return b
+			}
+		}
+	}
+	return s
+}

internal/impl/xml/processor.go

@@ -22,8 +22,9 @@ import (
 )
 
 const (
-	pFieldOperator = "operator"
-	pFieldCast     = "cast"
+	pFieldOperator          = "operator"
+	pFieldCast              = "cast"
+	pFieldPreserveNamespace = "preserve_namespaces"
 )
 
 func xmlProcSpec() *service.ConfigSpec {
@@ -91,6 +92,43 @@ With cast set to true, the resulting JSON structure would look like this:
     ]
   }
 }
+`+"```"+`
+
+== Preserving XML namespaces
+
+By default namespace prefixes on elements and attributes are dropped during conversion (e.g. `+"`<dc:title>`"+` becomes the key `+"`title`"+`), which makes the original XML impossible to reconstruct from the resulting JSON. Set `+"`preserve_namespaces`"+` to `+"`true`"+` to retain prefixes on element and attribute keys and to keep `+"`xmlns:*`"+` declarations as attributes.
+
+For example, given the following XML:
+
+`+"```xml"+`
+<root xmlns:dc="http://my.namespace/dc" xmlns:ot="http://my.namespace/ot">
+  <dc:title>This is a title</dc:title>
+  <dc:description tone="boring">This is a description</dc:description>
+  <ot:elements id="1">foo1</ot:elements>
+  <ot:elements id="2">foo2</ot:elements>
+  <ot:elements>foo3</ot:elements>
+</root>
+`+"```"+`
+
+With `+"`preserve_namespaces: true`"+` the resulting JSON structure would look like this:
+
+`+"```json"+`
+{
+  "root":{
+    "-xmlns:dc":"http://my.namespace/dc",
+    "-xmlns:ot":"http://my.namespace/ot",
+    "dc:title":"This is a title",
+    "dc:description":{
+      "#text":"This is a description",
+      "-tone":"boring"
+    },
+    "ot:elements":[
+      {"#text":"foo1","-id":"1"},
+      {"#text":"foo2","-id":"2"},
+      "foo3"
+    ]
+  }
+}
 `+"```").
 		Fields(
 			service.NewStringEnumField(pFieldOperator, "to_json").
@@ -99,6 +137,9 @@ With cast set to true, the resulting JSON structure would look like this:
 			service.NewBoolField(pFieldCast).
 				Description("Whether to try to cast values that are numbers and booleans to the right type. Default: all values are strings.").
 				Default(false),
+			service.NewBoolField(pFieldPreserveNamespace).
+				Description("Whether to preserve XML namespace prefixes on element and attribute keys, and retain xmlns declarations as attributes. When disabled, namespace prefixes are stripped.").
+				Default(false),
 		)
 }
 
@@ -111,8 +152,9 @@ func init() {
 }
 
 type xmlProc struct {
-	log  *service.Logger
-	cast bool
+	log              *service.Logger
+	cast             bool
+	preserveNSPrefix bool
 }
 
 func xmlProcFromParsed(pConf *service.ParsedConfig, mgr *service.Resources) (*xmlProc, error) {
@@ -129,9 +171,15 @@ func xmlProcFromParsed(pConf *service.ParsedConfig, mgr *service.Resources) (*xm
 		return nil, err
 	}
 
+	preserveNS, err := pConf.FieldBool(pFieldPreserveNamespace)
+	if err != nil {
+		return nil, err
+	}
+
 	j := &xmlProc{
-		log:  mgr.Logger(),
-		cast: cast,
+		log:              mgr.Logger(),
+		cast:             cast,
+		preserveNSPrefix: preserveNS,
 	}
 	return j, nil
 }
@@ -142,7 +190,12 @@ func (p *xmlProc) Process(_ context.Context, msg *service.Message) (service.Mess
 		return nil, err
 	}
 
-	root, err := ToMap(mBytes, p.cast)
+	var root map[string]any
+	if p.preserveNSPrefix {
+		root, err = ToMapPreserveNS(mBytes, p.cast)
+	} else {
+		root, err = ToMap(mBytes, p.cast)
+	}
 	if err != nil {
 		p.log.Debugf("Failed to parse part as XML: %v", err)
 		return nil, err