Saxon

Correction, just added some introspection, the duplication happens during parsing, not serialization.

Which XML parser are you using? Corruption of attribute values is a common problem with the JDK parser, which is why I always recommend use of Apache Xerces in preference. It's possible it's been fixed in JDK 8, but I'll hold my judgement on that.

Here's what the input file contains in hex, according to net.sf.saxon.functions.UnparsedText:

3c 3f 78 6d 6c 20 76 65 72 73 69 6f 6e 3d 22 31 2e 30 22 20 65 6e 63 6f 64 69 6e

< ? x m l v e r s i o n = " 1 . 0 " e n c o d i n

67 3d 22 55 54 46 2d 38 22 3f 3e 3c 6e 61 6d 65 20 73 6f 72 74 61 62 6c 65 3d 22

g = " U T F - 8 " ? > < n a m e s o r t a b l e = "

f0 9d 9a a4 f0 9d 9a a4 22 2f 3e

ð  š ¤ ð  š ¤ " / >

So yes, the sequence (f0 9d 9a a4) appears twice in the value of the attribute.

When I transform this with:

Saxon-EE 9.7.0.1J from Saxonica

Java version 1.6.0_27

Generating byte code...

Stylesheet compilation time: 355.856ms

Processing file:/Users/mike/bugs/2015/nunes/input.xml

Using parser org.apache.xerces.jaxp.SAXParserImpl$JAXPSAXParser

I get a file whose content is identical:

3c 3f 78 6d 6c 20 76 65 72 73 69 6f 6e 3d 22 31 2e 30 22 20 65 6e 63 6f 64 69 6e

< ? x m l v e r s i o n = " 1 . 0 " e n c o d i n

67 3d 22 55 54 46 2d 38 22 3f 3e 3c 6e 61 6d 65 20 73 6f 72 74 61 62 6c 65 3d 22

g = " U T F - 8 " ? > < n a m e s o r t a b l e = "

f0 9d 9a a4 f0 9d 9a a4 22 2f 3e

ð  š ¤ ð  š ¤ " / >

If I remove the Apache parser from the classpath and use the JDK parser:

Saxon-EE 9.7.0.1J from Saxonica

Java version 1.6.0_27

Stylesheet compilation time: 300.808ms

Processing file:/Users/mike/bugs/2015/nunes/input.xml

Using parser com.sun.org.apache.xerces.internal.jaxp.SAXParserImpl$JAXPSAXParser

I get this output file:

3c 3f 78 6d 6c 20 76 65 72 73 69 6f 6e 3d 22 31 2e 30 22 20 65 6e 63 6f 64 69 6e

< ? x m l v e r s i o n = " 1 . 0 " e n c o d i n

67 3d 22 55 54 46 2d 38 22 3f 3e 3c 6e 61 6d 65 20 73 6f 72 74 61 62 6c 65 3d 22

g = " U T F - 8 " ? > < n a m e s o r t a b l e = "

f0 9d 9a a4 f0 9d 9a a4 f0 9d 9a a4 22 2f 3e

ð  š ¤ ð  š ¤ ð  š ¤ " / >

which contains the character 3 times.

So yes, it's the old JDK parser bug, I'm afraid.

I ran it with Java 8:

Saxon-EE 9.7.0.1J from Saxonica

Java version 1.8.0_25

Stylesheet compilation time: 361.745129ms

Processing file:/Users/mike/bugs/2015/nunes/input.xml

Using parser com.sun.org.apache.xerces.internal.jaxp.SAXParserImpl$JAXPSAXParser

and it seems the bug is still there:

3c 3f 78 6d 6c 20 76 65 72 73 69 6f 6e 3d 22 31 2e 30 22 20 65 6e 63 6f 64 69 6e

< ? x m l v e r s i o n = " 1 . 0 " e n c o d i n

67 3d 22 55 54 46 2d 38 22 3f 3e 3c 6e 61 6d 65 20 73 6f 72 74 61 62 6c 65 3d 22

g = " U T F - 8 " ? > < n a m e s o r t a b l e = "

f0 9d 9a a4 f0 9d 9a a4 f0 9d 9a a4 22 2f 3e

ð  š ¤ ð  š ¤ ð  š ¤ " / >

Just use Apache Xerces!

Update: this bug appears to be fixed in Java 9. See

https://bugs.java.com/bugdatabase/view_bug.do?bug_id=8145969