Merging schema xml with data xml with Java

[bonjo]

Hello everyone,
I wish, based on two xml:
1. ui.xml: defines a layout, based on variables/conditions in its attributes, such as:

<root>
<item title="if data/gender = 'female': Ms.; else: Mr." />
</root>

2. data.xml: defines actual data came from another data source:

<data>
<gender>male</gender>
</data>

to merge the two into one result.xml, which, in that case would be:

<root>
<item title="Mr." />
</root>

I started from the end, as i think its clearer, I want to define attributes values based on what I have in my data.
Of course, the language I have there is just some pseudo -code, it can be anything, as long as its readable :)
I have read about mustache / handlebars implementations in java, which is more or less what I need (some sort of templating engine) but sadly, those are logic-less approaches.
I need less-logic or full-of-logic kind of approach.
What would you advice based on your knowledge?

[chaarmann]

Read in both XMLs as DOM-tree in Java (see org.w3c.dom.Doc ument).
Then merge them into a resulting DOM-tree.
(You must provide the merge logic by your own java code. Using XPath is very helpful doing it.)
Then save the resulting DOM-tree as an XML file (using TransformerFact ory to get a transformer instance and then using its transform() method).

[bonjo]

So I should, after having a dom-tree, use xpath in java to select all attributes (according to my example, all 'title' attributes), and in those attributes i should write xpath code, and evaluate it on my second (data) dom-tree?
then, replace the attribute's value to the evaluated one?
I can see how I can pull information, but how can I tell in xpath if statements, so that if it found that a certain value is 'X', assign 'Y' to that attribute.
Im not sure it will fit to my needs.
Could you be a bit more specific? Im not quite sure how the evaluation during merge process occurs.

I

[chaarmann]

You got it right. Just get the value of the attribute from the DOM as a String. In your example you would store "if data/gender = 'female': Ms.; else: Mr." in the variable (for example) named scriptString. I don't recognize this language. But let us assume it is javascript or xpath.
If it is javascript, you could use a scripting engine to evaluate this expression:

ScriptEngineManager factory = new ScriptEngineManager();
 ScriptEngine engine = factory.getEngineByName("JavaScript");
 engine.eval(scriptString);

If it is xpath, you could use it directly in your "xpath.evaluate (scriptString)" . It is possible to make if-else-statements with xpath. Example:

if (/data[@gender='female'])
   then 'Ms.' 
   else 'Mr.'

[bonjo]

I prefer using xpath engine ( I guess runtime will be much better comparing to js engine) But its not accepting those if statements.
I try the following:
exp = "if (/row/field[@name='service']) then 'Ms.' else 'Mr.'";
res = xPath.compile(e xp).evaluate(xm lDocument);
but then I get javax.xml.trans form.Transforme rException: Could not find function: if

Is it, for sure, exists in java xpath engine?

[chaarmann]

If-else-syntax exists only in Xpath 2.0 and up.
If you are using the old Xpath 1.0, you can simulate it with:

concat(substring('Ms.', 1 div boolean(/data[@gender='female'])),
       substring('Mr.', 1 div not(/data[@gender='female']))
      )

Why do you use "/row/field[@name='service'])" instead of "/data[@gender='female ']" as given in examples above?

[bonjo]

Yes, That was a mistake.
Using XPath 1.0 seems very problematic.
perhaps I should look at a different direction.
I also need the strength to do if-elseif-elseif-else, and maybe if-if-if statements.
how severe in performance is using the ScriptEngine to evaluate javascript expressions?

[chaarmann]

javascript speed was improved rapidly over the years. I remember the first Netscape browsers where I felt it was thousand times slower than nowadays.
The latest javascript engine for Firefox was also a major performance step . It is much faster than the one in Internet Explorer 8. My javascript game that I programmed a few years ago, and where I used many tricks to speed it up, is now running fast like hell!

I would say, give it a try.
I am not sure, but I bet it is much faster than xpath which did not have major performace improvements over the years.

You can do us a favor and measure the performance for both xpath and javascript and then post it here. I am eager to hear if my assumption is right.

[bonjo]

I'd gladly try to measure that, but first I need to understand how to accomplish it :-)
Its a bit of a mixup, as I cant really use only JS to iterate over my data.xml and evaluate statements over it.
I will still need, this way or another, to use some kind of xpath, just in javascript.
either that, or I will create some multi dimensional array holding all my xml elements and their attributes,
and evaluate the javascript together with my data structure.

Now, after I wrote it down, it sounds like one hell of a mess.

[chaarmann]

You are right, it is more work to do it in JS. And you can't reference variables like "/data/gender", you must use "gender" directly in your expression. How do you get the value for "gender" to set it before executing the JS? Most likely using xpath on your data-holding-XML.
Or use a converter to convert your data-holding-XML to JSON and use that in your JS-snippet.

So avoid premature optimization and start doing it with xpath, and only in case the performance is unacceptable bad at some hotspots, use JS there.
I guess JS would give you a performance boost only for
long and complicated expressions, but not for simple if-else constructs.

[bonjo]

Im still investigating my options.
I might prefer the JS engine after all.
but I wonder what is the best way to reconstruct my xml file to a data structure that JS Engine will work easily with.
What is it that you say about json? can you explain a bit further that point?

[bonjo]

Also, another interesting question, is it possible to debug the ScriptEngine instructions?
Same as you debug a js in browser / attach to process in Script mode with VS?

[chaarmann]

If you can create the GUI.xml yourself, and only the data.xml is given, then I would not use any methods described above.

Just change your GUI.xml slightly to make a GUI.xsl out of it.
Read about XSL and XSLT to know the syntax. XSL is the standard way to do transformation tasks like yours and there are also a lot of online-xsl-converter for easy testing.

Then you can use java to do the transformation via TransformerFact ory.

public void transform(String dataXml, String inputXsl, String outputXml)
	            throws TransformerConfigurationException,
	            TransformerException
	    {
	 
	        TransformerFactory factory = TransformerFactory.newInstance();
	        StreamSource xslStream = new StreamSource(inputXsl);
	        Transformer transformer = factory.newTransformer(xslStream);
	        StreamSource in = new StreamSource(dataXml);
	        StreamResult out = new StreamResult(outputXml);
	        transformer.transform(in, out);
	        System.out.println("The generated XML file is:" + outputXml);
	    }

[bonjo]

Well, I promised to update about time measurements.
so, I tried two approaches:
1. Having an xml with embedded JS.
I construct a dom-tree, locate js scripts and run it.
2. Having an xsl file. using TransformerFact ory to create an output xml of it.

Simulating with 10runs gives the following time (seconds):
JS: 0.03374
XSL: 0.04
Thats already very close.
When going for larger runs, it gets better (10000 runs):
JS: 0.00217
XSL: 0.01

I guess its some static variables and things i tried to manage better.
Perhaps the ScriptEngine is better than the old XPath.

[chaarmann]

Thank you for the statistics.
Can you post the testing code?