XmlReaders and fragments

[Lee Crabtree]

When reading fragments, it seems like XmlReaders try to read too much. I'm
working on a file parser for a new file format, and I've run into a problem.
The format has an XML fragment for a header, then a (frequently) large amount
of binary data beneath. In certain situations, there may be XML fragments
further down in the file. An example of the file might look like this:

<header>
<name>some name</name>
<size>1000</size>
<stuff>more data</stuff>
<otherstuff>19.2</otherstuff>
</header>...some huge block of binary data...

The header isn't a fixed size. In writing a little test to try and parse
out the header, I ran into what seemed like a really weird decision on the
part of the XmlReader. When I try to stream the data through using ReadOuterXml
(to get the header for processing later), it would throw an exception regarding
invalid characters MUCH further down in the file. In other words, the reader
had gotten past the ending element of the fragment, then kept going. ReadOuterXml
is just supposed to get the tags and children of the current node, which
in this case would have been the node labelled "header".

Here's a short program to demonstrate what I mean.

using System;
using System.IO;
using System.Xml;

namespace XmlParseTest
{
class Program
{
static void Main(string[] args)
{
FileStream file;
XmlReader baseReader;
XmlTextReader reader, sReader;
XmlReaderSettings readerSettings;
string xml;

file = new FileStream(@"c:\test.xml", FileMode.Open, FileAccess.Read);
file.Seek(0, SeekOrigin.Begin);

reader = new XmlTextReader(file, XmlNodeType.Element, null);
reader.Normalization = false;

readerSettings = new XmlReaderSettings();
readerSettings.ConformanceLevel = ConformanceLevel.Fragment;
readerSettings.IgnoreWhitespace = false;
readerSettings.IgnoreComments = true;
readerSettings.CheckCharacters = false;

baseReader = XmlReader.Create(reader, readerSettings);
baseReader.MoveToContent();
xml = baseReader.ReadOuterXml();

baseReader.Close();
file.Close();
}
}
}

Am I missing something?

Lee Crabtree

[Martin Honnen]

Lee Crabtree wrote:

When reading fragments, it seems like XmlReaders try to read too much.
I'm working on a file parser for a new file format, and I've run into a
problem. The format has an XML fragment for a header, then a
(frequently) large amount of binary data beneath. In certain
situations, there may be XML fragments further down in the file. An
example of the file might look like this:

<header>
<name>some name</name>
<size>1000</size>
<stuff>more data</stuff>
<otherstuff>19.2</otherstuff>
</header>...some huge block of binary data...

The header isn't a fixed size. In writing a little test to try and
parse out the header, I ran into what seemed like a really weird
decision on the part of the XmlReader. When I try to stream the data
through using ReadOuterXml (to get the header for processing later), it
would throw an exception regarding invalid characters MUCH further down
in the file. In other words, the reader had gotten past the ending
element of the fragment, then kept going. ReadOuterXml is just supposed
to get the tags and children of the current node, which in this case
would have been the node labelled "header".

ReadOuterXml() when positioned on an element reads everything including
the end tag and positions the reader on the next node. If there is
binary data after the end tag then you get an error. I am not sure what
you expect, ConformanceLevel.Fragment does only mean there is no
requirement to have exactly one root element, it does not mean binary
data is allowed.
If you want to consume an element but avoid that the reader is
positioned after the end tag then you might want to try whether using
ReadSubtree() does what you want, it gives you a second XmlReader you
can work with to consume only the element, once you close it the first
main reader is positioned on the end tag, not after it.
--

Martin Honnen --- MVP XML
http://JavaScript.FAQTs.com/

[Lee Crabtree]

Lee Crabtree wrote:

>

>When reading fragments, it seems like XmlReaders try to read too
much. I'm working on a file parser for a new file format, and I've
run into a problem. The format has an XML fragment for a header, then
a (frequently) large amount of binary data beneath. In certain
situations, there may be XML fragments further down in the file. An
example of the file might look like this:

<header>
<name>some name</name>
<size>1000</size>
<stuff>more data</stuff>
<otherstuff>19.2</otherstuff>
</header>...some huge block of binary data...
The header isn't a fixed size. In writing a little test to try and
parse out the header, I ran into what seemed like a really weird
decision on the part of the XmlReader. When I try to stream the data
through using ReadOuterXml (to get the header for processing later),
it would throw an exception regarding invalid characters MUCH further
down in the file. In other words, the reader had gotten past the
ending element of the fragment, then kept going. ReadOuterXml is
just supposed to get the tags and children of the current node, which
in this case would have been the node labelled "header".

ReadOuterXml() when positioned on an element reads everything
including
the end tag and positions the reader on the next node. If there is
binary data after the end tag then you get an error. I am not sure
what
you expect, ConformanceLevel.Fragment does only mean there is no
requirement to have exactly one root element, it does not mean binary
data is allowed.
If you want to consume an element but avoid that the reader is
positioned after the end tag then you might want to try whether using
ReadSubtree() does what you want, it gives you a second XmlReader you
can work with to consume only the element, once you close it the first
main reader is positioned on the end tag, not after it.

Fantastic. That worked. Thanks.

Lee Crabtree