/** Process files. */
protected static int processFiles(String[] args) {
int result = 0;
// Get file name/file wildcard specs.
String[] wildCards = new String[args.length - INITPARAMS];
for (int i = INITPARAMS; i < args.length; i++) {
wildCards[i - INITPARAMS] = args[i];
}
// Expand wildcards to list of
// file names.
String[] fileNames = FileNameUtils.expandFileNameWildcards(wildCards);
docsToProcess = fileNames.length;
System.out.println(
"There are "
+ Formatters.formatIntegerWithCommas(docsToProcess)
+ " documents to process.");
// Process each file.
for (int i = 0; i < fileNames.length; i++) {
processOneFile(fileNames[i]);
}
return fileNames.length;
}
/**
* Process one file.
*
* <p>\u0040param xmlFileName Input file name to check for part of speech/lemma mismatches..
*/
protected static void processOneFile(String xmlFileName) {
try {
// Report document being processed.
System.out.println(
"(" + ++currentDocNumber + "/" + docsToProcess + ") " + "processing " + xmlFileName);
// Create filter to strip <w> and <c>
// elements.
XMLFilter filter = new StripAllWordElementsFilter(XMLReaderFactory.createXMLReader());
// Strip path from input file name.
String strippedFileName = FileNameUtils.stripPathName(xmlFileName);
strippedFileName = FileNameUtils.changeFileExtension(strippedFileName, "");
// Generate output file name.
String xmlOutputFileName =
new File(outputDirectoryName, strippedFileName + ".xml").getAbsolutePath();
// Make sure output directory exists.
FileUtils.createPathForFile(xmlOutputFileName);
// Copy input xml to output xml,
// stripping <w> and <c> elements.
new FilterAdornedFile(xmlFileName, xmlOutputFileName, filter);
// Read it back and fix spacing.
String fixedXML = FileUtils.readTextFile(xmlOutputFileName, "utf-8");
fixedXML = fixedXML.replaceAll("(\\s+)", " ");
fixedXML = fixedXML.replaceAll(" ([\\.?!,;:\\)])", "\u00241");
fixedXML = fixedXML.replaceAll("\\( ", "(");
fixedXML = fixedXML.replaceAll("\u00b6 ", "\u00b6");
fixedXML = fixedXML.replaceAll("__NS1:", "");
fixedXML = fixedXML.replaceAll("__NS1", "");
/*
fixedXML =
fixedXML.replaceAll
(
"</__NS1:" ,
""
);
*/
// Emit unadorned XML.
SAXBuilder builder = new SAXBuilder();
Document document = builder.build(new StringReader(fixedXML));
new AdornedXMLWriter(document, xmlOutputFileName);
} catch (Exception e) {
System.out.println(" Error: " + e.getMessage());
e.printStackTrace();
}
}