Laboratoire ICAR » Plateforme TXM

class XMLTXM2CQP

{

	/** The url. */

	private def url;

	/** The input data. */

	private def inputData;

	/** The factory. */

	private def factory;

	/** The parser. */

	private XMLStreamReader parser;

	/** The output. */

	private def output;

	/** The hashmap of txm:form and txm:ana values and the attributes hash*/

	LinkedHashMap<String, String> anahash = new LinkedHashMap<String, String>();

	LinkedHashMap<String, String> formhash = new LinkedHashMap<String, String>();

	LinkedHashMap<String, String> wordattributes = new LinkedHashMap<String, String>();

	/** The balisesfound. */

	HashMap<String, List<String>> balisesfound;// = new HashMap<String, List<String>>();

	/** The balises to keep. */

	List<String> balisesToKeep;

	/** The send to p attributes. */

	HashMap <String, List<String>> sendToPAttributes;// = new HashMap<String, List<String>>();

	/** The injected p attributes. */

	List<String> injectedPAttributes = new ArrayList<String>();

	/** The default reference : a pattern + the properties to use */

	List<String> defaultReferences = new ArrayList<String>();

	String defaultReferencePattern;

	/** The injected p attributes values. */

	HashMap <String, String> injectedPAttributesValues;// = new ArrayList<String>();

	/** The addinfos. */

	boolean addinfos = false;

	/** The txtname. */

	String txtname;

	/** The base. */

	String base;

	/** The project. */

	String project;

	/** The lang. */

	public String lang= "fr";

	public String currentForm;

	public String currentAna;

	/**

	 * Sets the lang.

	 *

	{

		this.lang = lang;

	}

	/**

	 * Instantiates a new xMLTX m2 cqp.

	 *

			this.url = url;

			inputData = url.openStream();

			factory = XMLInputFactory.newInstance();

			parser = factory.createXMLStreamReader(inputData);

		} catch (XMLStreamException ex) {

			System.out.println(ex);

		}catch (IOException ex) {

			System.out.println("IOException while parsing ");

		}

	}

	/**

	 * Sets the text info.

	 *

		this.base = base;

		this.project = project;

	}

	/**

	 * Creates the output.

	 *

			return false;

		}

	}

	/** The haspb. */

	boolean haspb = false;

	/** The haslb. */

	boolean haslb = false;

	/**

	 * Transform file.

	 *

			println "no element has been defined to be keeped"

			return false;

		}

		haspb = false;

		haslb = false;

		boolean flagAna;

		boolean flagForm;

		boolean flagWord;

		String vWord = "";

		String vForm = "";

		String vAna = "";

		String lb_id = "";

		String pb_id = "";

		wordattributes = [:];

		balisesfound = new HashMap<String, List<String>>();

		if(!createOutput(outfile))

			return false;

		if(sendToPAttributes != null)

		{

			for(String tag: sendToPAttributes.keySet())

					injectedPAttributes.add(tag+attr);

			injectedPAttributesValues = [:];

		}

		//output.write("<txmcorpus lang=\""+lang+"\">\n");

		balisesfound.put("txmcorpus",["lang"]);

		try {

				switch (event) {

					case XMLStreamConstants.START_ELEMENT:

						String localname = parser.getLocalName().toLowerCase();

					// we will only declare found tags in cwb registry

						if(balisesToKeep.contains(localname)) {

							if(!balisesfound.containsKey(localname)) {

								balisesfound.put(localname, []);

							}

							List<String> attrlist = balisesfound.get(localname);

							for (int i= 0 ; i < parser.getAttributeCount() ;i++ )

								if(!attrlist.contains(parser.getAttributeLocalName(i)))

									attrlist.add(parser.getAttributeLocalName(i));

						}

						switch (localname) {

							case "w": // get word id !!

								wordattributes.put("id", parser.getAttributeValue(null, "id"));

								break;

							case "form":

								flagForm = true;

								currentForm = parser.getAttributeValue(null, "type");

									currentForm = "default";

								vForm = "";

								break;

							case "ana":

								flagAna = true;

								vAna ="";

								currentAna = (parser.getAttributeValue(null,"type"));

								if(currentAna != null)

									currentAna = currentAna.substring(1)// remove the #

								else

									flagAna = false;

								break;

							default:

								if (sendToPAttributes != null) {

									//println "should store $localname ? with "+sendToPAttributes.keySet()

									if (sendToPAttributes.keySet().contains(localname)) {

										}

									}

								}

								if (balisesToKeep.contains(localname)) {

									output.write("<"+localname);

									//println "write <"+localname+"..."

										output.write(" "+attrname+"=\""+parser.getAttributeValue(i).replace("&", "&").replace("\"", """)+"\"" );

									}

									if (localname.equals("text"))

										if (addinfos) {

											List<String> attrlist = balisesfound.get(localname);

											if (!idwritten) {

												output.write(" id=\""+txtname+"\"")

												attrlist.add("id");

												attrlist.add("project");

											}

										}

									// finalize tag

									output.write(">\n");

								}

						}

						break;

					case XMLStreamConstants.END_ELEMENT:

						String localname = parser.getLocalName().toLowerCase();

						switch (localname) {

									formhash.put(currentForm, vForm);

								flagForm = false;

								break;

							case "ana":

								if(flagAna)

									anahash.put(currentAna, vAna);

								flagAna = false;

								break;

							case "w":

								vWord = "";

								vWord = formhash.get("default").replaceAll("&", "&amp;").replaceAll("<", "&lt;"); // get default form

								for (String form : formhash.keySet()) // and the others

									if (form != "default")

										vWord += "\t"+formhash.get(form);

								for (String type : wordattributes.keySet()) // only word id ?

									vWord+="\t"+wordattributes.get(type)

								if (sendToPAttributes != null) // word attributes from structure properties

								{

									//println "injectedPAttributesValues: "+injectedPAttributesValues

									for(String pattr : injectedPAttributes)

										vWord+="\t"+injectedPAttributesValues.get(pattr) ;//les attributs injecter

								}

								for (String type : anahash.keySet()) // word annotations in txm:ana

									vWord+="\t"+anahash.get(type)

								output.write(vWord+"\n");

								vWord= "";

								break;

							default:

								if (sendToPAttributes != null) // reset structure properties

								{

										}

									}

								}

								if (balisesToKeep.contains(localname)) {

									output.write("</"+localname+">\n");

								}

						}

						break;

					case XMLStreamConstants.CHARACTERS:

						if (flagForm) {

							vForm += parser.getText().trim();

			//output.write("</txmcorpus>\n");

			output.close();

			if (parser != null) parser.close();

		} catch (Exception ex) {

			println "Error while parsing $url : "+ex

			ex.printStackTrace();

		}

		return true;

	}

	/**

	 * Gets the p attributs.

	 *

	public List<String> getpAttributs()

	{

		def pAttributs = [];

		for (String wordattr : wordattributes.keySet()) {

			pAttributs.add(wordattr);

		}

		if (sendToPAttributes != null)

			for (String pAttr : this.injectedPAttributes)

				pAttributs.add(pAttr);

		for (String anakey : anahash.keySet()) {

			pAttributs.add(anakey);

		}

		return pAttributs;

	}

	/**

	 * Gets the s attributs.

	 *

	 */

	public List<String> getsAttributs()

	{

		def sAttributs = [];

		for (String balise : this.balisesfound.keySet()) {

			List<String> sAtt = this.balisesfound.get(balise);

			for (String attr : sAtt) {

				attributes+="+"+attr;

			}

			if (sAtt.size() > 0)

				sAttributs.add(balise +":"+attributes);

			else

		}

		return sAttributs;

	}

	/**

	 * Sets the balises to keep.

	 *

		else

			println("Warning: the list of elements to keep is null")

	}

	/**

	 * Sets the defautl reference pattern

	 * TODO: not implemented

			defaultReferencePattern = pattern;

		}

	}

	/**

	 * Sets the send to p attributes.

	 *

		else

			println("Warning: the pAttributes to inject is null")

	}

	/**

	 * The main method.

	 *

	 * @param args the arguments

	 */

	public static void main(String[] args) {

		String rootDir = "/home/mdecorde/TXM/corpora/CORNEILLEMOLIERETER/txm/CORNEILLEMOLIERETER";

		File srcfile = new File(rootDir,"CORNEILLEP_AGESILAS_1666.xml");

		println srcfile.exists()

		File cqpfile = new File(rootDir, "out/CORNEILLEP_AGESILAS_1666.cqp");

		new File(rootDir,"out").deleteDir()

		new File(rootDir,"out").mkdir()

		System.out.println("XMLTXM2CQP : "+srcfile+" >> "+cqpfile);

		def builder = new XMLTXM2CQP(srcfile.toURL());

		def balises = ["text", "s"];

		builder.setBalisesToKeep(balises);

		builder.transformFile(cqpfile);

		println("SATTRIBUTS: "+builder.getsAttributs());

		println("PATTRIBUTS: "+builder.getpAttributs());

		return;