I have a structure like this:
<a>
<b>
<c>foo</c>
<d>bar</d>
</b>
<b>
<c>baz</c>
<d>qux</d>
</b>
</a>
I set /a/b as context and loop through all occurrences to get the texts of c and d:
VTDGen vg = new VTDGen();
vg.parseFile("test.xml", true);
VTDNav vn = vg.getNav();
AutoPilot ap = new AutoPilot(vn);
ap.selectXPath("/a/b");
while(ap.evalXPath() != -1) {
String c = extractText(vn, "c");
String d = extractText(vn, "d");
}
Currently, I use the extractText() method I've written to extract the texts of the nodes inside the loop:
public String extractText(VTDNav v, String path) throws Exception {
VTDNav vn = v.cloneNav();
AutoPilot ap = new AutoPilot(vn);
ap.selectXPath(path);
int elementIndex = ap.evalXPath();
if(elementIndex == -1) {
return null;
}
int tokenIndex = (path.contains("@"))
? vn.getAttrVal(vn.toString(elementIndex))
: vn.getText();
return (tokenIndex == -1) ? null : vn.toString(tokenIndex);
}
In the method, I clone the VTDNav and create a new AutoPilot instance so I don't displace the cursor of those objects of my loop when selecting the new XPath.
Furthermore, I have to check the path whether it contains an @, to see if it's an text node or an attribute (i.e. to use getText() or getAttrVal()).
My profiler shows that my extractText() method takes the most time in my huge application, so there must be a better solution here.