LightXmlWriter 1.0.0

LightXmlWriter

This is a replacement of XmlWriter which can be useful in high-performance scenarios. Therefore there are several goals to reach:

• high performance
  • zero allocation
  • be faster than XmlWriter
  • no runtime validation
  • keep it simple (no configuration)
  • allow writing concrete types (Span<T>, int, double, DateTime, etc.)
  • allow disable escaping values in similar way to Newtonsoft.Json
  • produce as small as possible output XML
• easy migration from XmlWriter

LightXmlWriter vs XmlWriter

Benchmarks

EnterpriseLightXmlWriterBenchmarks.LightXmlWriter_Write_Xml

|                   Method |      Mean |     Error |    StdDev |    Median | Gen 0 | Gen 1 | Gen 2 | Allocated |
|------------------------- |----------:|----------:|----------:|----------:|------:|------:|------:|----------:|
| LightXmlWriter_Write_Xml |  8.585 us | 0.1733 us | 0.2429 us |  8.600 us |     - |     - |     - |         - |
|      XmlWriter_Write_Xml | 17.923 us | 0.4765 us | 1.3671 us | 18.300 us |     - |     - |     - |    1288 B |

Similarities

• Method names are kept the same, and they should work the same way

For example for both writers:

writer.WriteStartElement("soapenv", "Envelope", "http://schemas.xmlsoap.org/soap/envelope/");

produces:

<soapenv:Envelope xmlns:soapenv="http://schemas.xmlsoap.org/soap/envelope/">

• both are not thread-safe

Both writers keep state, so they must not be processed in parallel. LightXmlWriter does not do any validation, so it could produce an unexpeced output.

• both implement these methods:

  • WriteStartElement
  • WriteEndElement
  • WriteElementString
  • WriteStartAttribute
  • WriteEndAttribute
  • WriteAttributeString
  • WriteRaw
  • WriteValue

• both implement IDisposable interface

Differences

• LightXmlWriter contains more overloads of methods

The overload accepts int, double, bool, ReadOnlySpan<char> (only in .NET Core), char[] etc. You can even use overload that uses Action on TextWriter, so you can have all control!

• WriteEndElement must have name of the element

XmlWriter:

writer.WriteStartElement("Person");
writer.WriteEndElement();

LightXmlWriter:

writer.WriteStartElement("Person");
writer.WriteEndElement("Person");

• LightXmlWriter does not use XmlWriterSettings

It only uses its default settings, that's why it is light 😃 What are the default settings? It does not do any validation at run-time, no pretty-print of output, no XML declaration. So it is just simple writer.

• LightXmlWriter has better performance than XmlWriter

LightXmlWriter is about 2x faster and generates less allocations than XmlWriter because of methods overloads that don't need conversion to string before.

• LightXmlWriter produces more compressed self-closed xml tags

The compressed version does not contain space before closing singn. It is still valid XML.

XmlWriter:

writer.WriteStartElement("Person");
writer.WriteEndElement();
//produces <Person />

LightXmlWriter:

writer.WriteStartElement("Person");
writer.WriteEndElement("Person");
//produces <Person/>

• LightXmlWriter does not validate & escape tag and attribute names

XmlReader throws an exception in case of unescaped name

• LightXmlWriter allows write values without escaping

This is similar to Newtonsoft JsonWriter where you can set flag escape: false. It brings better performance than with enabled escape. Use it where you are sure that value does not need escaping.

Example:

writer.WriteStartElement("Code");
writer.WriteValue("ABCD", escape: false);
writer.WriteEndElement();
//produces <Code>ABCD</Code>

writer.WriteElementString("Code", "ABCD", escapeValue: false);
//produces <Code>ABCD</Code>

Migration to LightXmlWriter

1. To each WriteEndElement(); add argument with name of closing element, e.g. WriteEndElement("Person");
2. Optionally to WriteEndAttribute(); add argument with name of closing attribute, e.g. WriteEndAttribute("Age");
3. Optionally use more adequate method overload to reduce string allocations, e.g. WriteAttributeString("Age", "25"); to WriteAttributeString("Age", 25);
4. Optionally disable value escaping where it is not needed, e.g. WriteAttributeString("Code", "ABC123"); to WriteAttributeString("Code", "ABC123", escapeValue: false);
5. If you see any difference in output, report it here

Contributing

You are welcome to help with this package. There are also a lot to do: write more benchmark tests, more tests, more overloads, finding bugs, making optimisations of existing code.

Acknowledgments

Please use it carefully. LightXmlWriter can produce an invalid XML, so write tests for each output.

Requirements

Framework compatible with .NET Standard 1.3 (.NET Core 1.0, .NET Framework 4.6, Mono 4.6) or higher.
Currently ReadOnlySpan<char> overloads are only available in .NET Core 2.1 and .NET 5.0 builds.