XML Pipelines: the Process Description
by
·
Join the DZone community and get the full member experience.
Join For FreeThe world of XML is moving fast, in its twelve year old history has raised many technologies. Many ways to validate, transform or query the data carried by the XML. Actually the validations and the transformations are commonly used together in the processes today so commonly called XML Pipelines.
XML Pipeline
By definition “XML Pipeline is formed when XML processes, especially XML transformations and XML validations, are connected together”, that means if we send the output of the validation as input to the transformation (and that is a common use case), we have just created XML Pipeline, more accurately the linear XML Pipeline. Well non-linear pipeline would be the one, where we decide between two process branches based on a condition, where we are looping, running parallel branches or handling errors on the fly.
XProc, W3C Recommended XML pipeline language
XProc is a XML language describing processes, using XSLT for transformations and the XSD, RelaxNG or the Schematron for validation. The greatest thing about XProc is that you describe processing of the XML document by XML language, using the W3C standards on the way.
There is a catch
Well with the XProc, we describe the process by one language, validate by another language and transform by yet another language. By using the three different languages for describing even the simplest use case of the XML Pipelining, a huge space for flaws opens. If we want to make something as trivial as changing the name of one element, we have to make changes in three different languages.
An alternative exist, it’s called XDefinition
If you remember my previous article called “XML Processing and Validation Merging Together” where I wrote briefly about an interesting XML technology called XDefinitions, you will find one language able to transform and validate XML. And because its ability to validate the multiple inputs, transform them, and validate output again, all written in only one highly readable language, the space for the errors is limited to unavoidable minimum.
XML Pipeline by XDefinition
Every XDefinition process can initiate another XDefinition process. This kind of scalability allows us to model linear or nonlinear XML pipelines by one language. If we want to change the structure, we will do it on one place only.
<xd:def xmlns:xd = "http://www.syntea.cz/xdef/2.0" xd:root = "Weather" >
<!-- declaration of variables -->
<xd:declaration>
String $source; /*URL to source data */
float $sum = 0; /*total sum of temperatures */
int $n = 0; /*number of measurements*/
</xd:declaration>
<!-- model of input data -->
<Weather date = "optional datetime('yyyy-M-d')">
<Measurement xd:script = "occurs 1..;"
wind = "required float()"
temperature = "required float(-99, +99);
onTrue {/* process value of temperature. */
$n++; /* increase number of measurements. */
$sum += parseFloat(getText()); /* add value to total sum. */
}"
time = "required datetime('HH:mm')" />
</Weather>
<!-- model of result -->
<html xmlns = "http://www.w3.org/1999/xhtml"
xd:script="create parseXml($source, '*');
/* Parse and validate input data with the Weather xdefinition. */">
<body>
<h1> create "Date: " + from("/Weather/@date") </h1>
<li xd:script="occurs +; create from('//Weather/Measurement')">
create "Time: " + from("@time") +
", wind: " + from("@wind") +
", temperature: " + from("@temperature");
</li>
<h3>
create $n == 0 ? "No data" /* $n is equal to 0 */ :
"Average temperature : " + ($sum/$n); /*print average*/
</h3>
</body>
</html>
</xd:def>
Input document
<?xml version="1.0" encoding="UTF-8"?>
<Weather date = "2005-05-11" >
<Measurement wind = "5.3" temperature ="13.0" time = "05:00" />
<Measurement wind = "7.2" temperature ="15.2" time = "11:00" />
<Measurement wind = "8.7" temperature ="18.1" time = "15:00" />
<Measurement wind = "3.9" temperature ="16.5" time = "20:00" />
</Weather>
Output document
<html xmlns="http://www.w3.org/1999/xhtml">
<body>
<h1>
Date: 2005-05-11
</h1>
<li>
Time: 05:00, wind: 5.3, temperature: 13.0
</li>
<li>
Time: 11:00, wind: 7.2, temperature: 15.2
</li>
<li>
Time: 15:00, wind: 8.7, temperature: 18.1
</li>
<li>
Time: 20:00, wind: 3.9, temperature: 16.5
</li>
<h3>
Average temperature : 15.7
</h3>
</body>
</html>
Try this example here online. As you can see XDefinition contains a model of the input document against which it’s being validated. You can even use variables from the validation in the transformation/composition as is $n and $sum (used in the example) that are global for every model of XDefinition. That is very important, because XDefinition can complete validation with non fatal errors. So your process doesn’t have to end with invalid but well formed input. Invalid document still contains the information value and you can work with it.
Why “composition”? Because the word composition describes the process better which is being done by the XDefinition. Nowadays the overwhelming majority of the data generated by every transformation process has XML output. If you take that for granted the whole XML Pipeline has to be well formed and the transformations are then compositions of smaller nodes. Any composition with the XDefinition is XML data modeling and your pipeline has a complete right to be called a XML Pipeline.
Resources
- http://www.syntea.cz/syntea_web/?page=xdef_guidepost&hl=en_US
- http://en.wikipedia.org/wiki/XML_pipeline
- http://www.w3.org/TR/xml-pipeline
Opinions expressed by DZone contributors are their own.
Comments