encoding/xml: add InputOffset method to Decoder

Among other things, this allows users to match the decoded
pieces with the original XML, which can be necessary for
implementing standards like XML signatures.

Fixes #8484.

LGTM=bradfitz
R=bradfitz
CC=golang-codereviews
https://golang.org/cl/122960043

2 files changed, 38 insertions, 4 deletions

diff --git a/src/pkg/encoding/xml/xml.go b/src/pkg/encoding/xml/xml.go
index b473cb8458..a4cd4e29e0 100644
--- a/src/pkg/encoding/xml/xml.go
+++ b/src/pkg/encoding/xml/xml.go
@@ -29,6 +29,7 @@ import (
 type SyntaxError struct {
 	Msg  string
 	Line int
+	Byte int64 // byte offset from start of stream
 }
 
 func (e *SyntaxError) Error() string {
@@ -196,6 +197,7 @@ type Decoder struct {
 	ns             map[string]string
 	err            error
 	line           int
+	offset         int64
 	unmarshalDepth int
 }
 
@@ -859,9 +861,17 @@ func (d *Decoder) getc() (b byte, ok bool) {
 	if b == '\n' {
 		d.line++
 	}
+	d.offset++
 	return b, true
 }
 
+// InputOffset returns the input stream byte offset of the current decoder position.
+// The offset gives the location of the end of the most recently returned token
+// and the beginning of the next token.
+func (d *Decoder) InputOffset() int64 {
+	return d.offset
+}
+
 // Return saved offset.
 // If we did ungetc (nextByte >= 0), have to back up one.
 func (d *Decoder) savedOffset() int {
@@ -891,6 +901,7 @@ func (d *Decoder) ungetc(b byte) {
 		d.line--
 	}
 	d.nextByte = int(b)
+	d.offset--
 }
 
 var entity = map[string]int{
diff --git a/src/pkg/encoding/xml/xml_test.go b/src/pkg/encoding/xml/xml_test.go
index 7723ab1c9f..be995c0d52 100644
--- a/src/pkg/encoding/xml/xml_test.go
+++ b/src/pkg/encoding/xml/xml_test.go
@@ -170,7 +170,7 @@ var xmlInput = []string{
 func TestRawToken(t *testing.T) {
 	d := NewDecoder(strings.NewReader(testInput))
 	d.Entity = testEntity
-	testRawToken(t, d, rawTokens)
+	testRawToken(t, d, testInput, rawTokens)
 }
 
 const nonStrictInput = `
@@ -225,7 +225,7 @@ var nonStrictTokens = []Token{
 func TestNonStrictRawToken(t *testing.T) {
 	d := NewDecoder(strings.NewReader(nonStrictInput))
 	d.Strict = false
-	testRawToken(t, d, nonStrictTokens)
+	testRawToken(t, d, nonStrictInput, nonStrictTokens)
 }
 
 type downCaser struct {
@@ -254,7 +254,7 @@ func TestRawTokenAltEncoding(t *testing.T) {
 		}
 		return &downCaser{t, input.(io.ByteReader)}, nil
 	}
-	testRawToken(t, d, rawTokensAltEncoding)
+	testRawToken(t, d, testInputAltEncoding, rawTokensAltEncoding)
 }
 
 func TestRawTokenAltEncodingNoConverter(t *testing.T) {
@@ -280,9 +280,12 @@ func TestRawTokenAltEncodingNoConverter(t *testing.T) {
 	}
 }
 
-func testRawToken(t *testing.T, d *Decoder, rawTokens []Token) {
+func testRawToken(t *testing.T, d *Decoder, raw string, rawTokens []Token) {
+	lastEnd := int64(0)
 	for i, want := range rawTokens {
+		start := d.InputOffset()
 		have, err := d.RawToken()
+		end := d.InputOffset()
 		if err != nil {
 			t.Fatalf("token %d: unexpected error: %s", i, err)
 		}
@@ -300,6 +303,26 @@ func testRawToken(t *testing.T, d *Decoder, rawTokens []Token) {
 			}
 			t.Errorf("token %d = %s, want %s", i, shave, swant)
 		}
+
+		// Check that InputOffset returned actual token.
+		switch {
+		case start < lastEnd:
+			t.Errorf("token %d: position [%d,%d) for %T is before previous token", i, start, end, have)
+		case start >= end:
+			// Special case: EndElement can be synthesized.
+			if start == end && end == lastEnd {
+				break
+			}
+			t.Errorf("token %d: position [%d,%d) for %T is empty", i, start, end, have)
+		case end > int64(len(raw)):
+			t.Errorf("token %d: position [%d,%d) for %T extends beyond input", i, start, end, have)
+		default:
+			text := raw[start:end]
+			if strings.ContainsAny(text, "<>") && (!strings.HasPrefix(text, "<") || !strings.HasSuffix(text, ">")) {
+				t.Errorf("token %d: misaligned raw token %#q for %T", i, text, have)
+			}
+		}
+		lastEnd = end
 	}
 }