Remove CDATA preservation on read

This feature, introduced in v1.1.3, was implemented in such a way
that it broke the ability to read XML documents encoded in non-UTF8
character sets.

3 files changed, 8 insertions, 65 deletions

diff --git a/etree.go b/etree.go
index 5d272a3..ac9dc74 100644
--- a/etree.go
+++ b/etree.go
@@ -775,7 +775,6 @@ func (e *Element) readFrom(ri io.Reader, settings ReadSettings) (n int64, err er
 	var stack stack
 	stack.push(e)
 	for {
-		xr.ResetPeek(dec.InputOffset())
 		t, err := dec.RawToken()
 		switch {
 		case err == io.EOF:
@@ -806,9 +805,7 @@ func (e *Element) readFrom(ri io.Reader, settings ReadSettings) (n int64, err er
 		case xml.CharData:
 			data := string(t)
 			var flags charDataFlags
-			if xr.PeekContainsCdata() {
-				flags = cdataFlag
-			} else if isWhitespace(data) {
+			if isWhitespace(data) {
 				flags = whitespaceFlag
 			}
 			newCharData(data, flags, top)
diff --git a/etree_test.go b/etree_test.go
index 180c1a8..07eee08 100644
--- a/etree_test.go
+++ b/etree_test.go
@@ -892,8 +892,8 @@ func TestIndentPreserveWhitespace(t *testing.T) {
 		{"<test>  </test>", "<test>  </test>"},
 		{"<test>\t</test>", "<test>\t</test>"},
 		{"<test>\t\n \t</test>", "<test>\t\n \t</test>"},
-		{"<test><![CDATA[ ]]></test>", "<test><![CDATA[ ]]></test>"},
-		{"<test> <![CDATA[ ]]> </test>", "<test><![CDATA[ ]]></test>"},
+		{"<test><![CDATA[ ]]></test>", "<test> </test>"},
+		{"<test> <![CDATA[ ]]> </test>", "<test/>"},
 		{"<outer> <inner> </inner> </outer>", "<outer>\n  <inner> </inner>\n</outer>"},
 	}
 
@@ -1278,22 +1278,3 @@ func TestWhitespace(t *testing.T) {
 	cd.SetData("")
 	checkBoolEq(t, cd.IsWhitespace(), true)
 }
-
-func TestPreserveCDATA(t *testing.T) {
-	s := `<name><![CDATA[My]] <b>name</b> <![CDATA[is]]></name>`
-
-	doc := NewDocument()
-	err := doc.ReadFromString(s)
-	if err != nil {
-		t.Fatalf("etree: failed to ReadFromString: %v", err)
-	}
-
-	result, err := doc.WriteToString()
-	if err != nil {
-		t.Fatalf("etree: failed to WriteToString: %v", err)
-	}
-
-	if result != s {
-		t.Errorf("etree: wanted %q, got %q", s, result)
-	}
-}
diff --git a/helpers.go b/helpers.go
index 2fddc28..39a31ee 100644
--- a/helpers.go
+++ b/helpers.go
@@ -6,7 +6,6 @@ package etree
 
 import (
 	"bufio"
-	"bytes"
 	"io"
 	"strings"
 	"unicode/utf8"
@@ -88,54 +87,20 @@ func (f *fifo) grow() {
 // bytes read from its encapsulated reader and detects when a CDATA
 // prefix has been parsed.
 type xmlReader struct {
-	r     io.ByteReader
+	r     io.Reader
 	bytes int64
-	peek  []byte
-	last  byte
 }
 
 var cdataPrefix = []byte("<![CDATA[")
 
 func newXmlReader(r io.Reader) *xmlReader {
-	return &xmlReader{
-		r:     bufio.NewReader(r),
-		bytes: 0,
-		peek:  make([]byte, 0, len(cdataPrefix)),
-		last:  0,
-	}
+	return &xmlReader{r, 0}
 }
 
 func (xr *xmlReader) Read(p []byte) (n int, err error) {
-	// Since xmlReader implements the io.ByteReader interface, the XML decoder
-	// bypasses Read in favor of ReadByte.
-	return 0, nil
-}
-
-func (xr *xmlReader) ReadByte() (b byte, err error) {
-	b, err = xr.r.ReadByte()
-	if err == nil {
-		xr.last = b
-		xr.bytes += 1
-		if len(xr.peek) < len(cdataPrefix) {
-			xr.peek = append(xr.peek, b)
-		}
-	}
-	return b, err
-}
-
-func (xr *xmlReader) ResetPeek(decoderOffset int64) {
-	xr.peek = xr.peek[0:0]
-
-	// If the decoder offset doesn't match the number of bytes read so far,
-	// then the decoder performed an "unget" on the last byte read. Return
-	// this byte to the front of the peek buffer.
-	if decoderOffset != xr.bytes {
-		xr.peek = append(xr.peek, xr.last)
-	}
-}
-
-func (xr *xmlReader) PeekContainsCdata() bool {
-	return bytes.Equal(xr.peek, cdataPrefix)
+	n, err = xr.r.Read(p)
+	xr.bytes += int64(n)
+	return n, err
 }
 
 // xmlWriter implements a proxy writer that counts the number of