root / tmp / org.txm.core / src / java / org / txm / scripts / importer / XMLTXM2WTC.groovy @ 2473
History | View | Annotate | Download (13 kB)
| 1 |
// Copyright © 2010-2013 ENS de Lyon.
|
|---|---|
| 2 |
// Copyright © 2007-2010 ENS de Lyon, CNRS, INRP, University of
|
| 3 |
// Lyon 2, University of Franche-Comté, University of Nice
|
| 4 |
// Sophia Antipolis, University of Paris 3.
|
| 5 |
//
|
| 6 |
// The TXM platform is free software: you can redistribute it
|
| 7 |
// and/or modify it under the terms of the GNU General Public
|
| 8 |
// License as published by the Free Software Foundation,
|
| 9 |
// either version 2 of the License, or (at your option) any
|
| 10 |
// later version.
|
| 11 |
//
|
| 12 |
// The TXM platform is distributed in the hope that it will be
|
| 13 |
// useful, but WITHOUT ANY WARRANTY; without even the implied
|
| 14 |
// warranty of MERCHANTABILITY or FITNESS FOR A PARTICULAR
|
| 15 |
// PURPOSE. See the GNU General Public License for more
|
| 16 |
// details.
|
| 17 |
//
|
| 18 |
// You should have received a copy of the GNU General
|
| 19 |
// Public License along with the TXM platform. If not, see
|
| 20 |
// http://www.gnu.org/licenses.
|
| 21 |
//
|
| 22 |
//
|
| 23 |
//
|
| 24 |
// $LastChangedDate: 2017-04-11 15:30:35 +0200 (mar. 11 avril 2017) $
|
| 25 |
// $LastChangedRevision: 3426 $
|
| 26 |
// $LastChangedBy: mdecorde $
|
| 27 |
//
|
| 28 |
package org.txm.scripts.importer
|
| 29 |
|
| 30 |
import java.text.DateFormat; |
| 31 |
import java.util.Date; |
| 32 |
import java.util.ArrayList; |
| 33 |
import java.util.HashMap; |
| 34 |
import java.util.LinkedHashMap; |
| 35 |
import javax.xml.stream.*; |
| 36 |
import java.net.URL; |
| 37 |
import org.txm.importer.filters.*; |
| 38 |
// TODO: Auto-generated Javadoc
|
| 39 |
|
| 40 |
/**
|
| 41 |
* The Class XMLTXM2CQP.
|
| 42 |
*
|
| 43 |
* @author mdecorde
|
| 44 |
* simple transofmration of a xml-tei-txm file into cqp file
|
| 45 |
*/
|
| 46 |
|
| 47 |
class XMLTXM2CQP |
| 48 |
{
|
| 49 |
|
| 50 |
/** The url. */
|
| 51 |
private def url; |
| 52 |
|
| 53 |
/** The input data. */
|
| 54 |
private def inputData; |
| 55 |
|
| 56 |
/** The factory. */
|
| 57 |
private def factory; |
| 58 |
|
| 59 |
/** The parser. */
|
| 60 |
private XMLStreamReader parser;
|
| 61 |
|
| 62 |
/** The output. */
|
| 63 |
private def output; |
| 64 |
|
| 65 |
/** The hashmap of txm:form and txm:ana values and the attributes hash*/
|
| 66 |
LinkedHashMap<String, String> anahash = new LinkedHashMap<String, String>(); |
| 67 |
LinkedHashMap<String, String> formhash = new LinkedHashMap<String, String>(); |
| 68 |
LinkedHashMap<String, String> wordattributes = new LinkedHashMap<String, String>(); |
| 69 |
|
| 70 |
/** The balisesfound. */
|
| 71 |
HashMap<String, List<String>> balisesfound;// = new HashMap<String, List<String>>(); |
| 72 |
|
| 73 |
/** The balises to keep. */
|
| 74 |
List<String> balisesToKeep; |
| 75 |
|
| 76 |
/** The send to p attributes. */
|
| 77 |
HashMap <String, List<String>> sendToPAttributes;// = new HashMap<String, List<String>>(); |
| 78 |
|
| 79 |
/** The injected p attributes. */
|
| 80 |
List<String> injectedPAttributes = new ArrayList<String>(); |
| 81 |
|
| 82 |
/** The default reference : a pattern + the properties to use */
|
| 83 |
List<String> defaultReferences = new ArrayList<String>(); |
| 84 |
String defaultReferencePattern;
|
| 85 |
|
| 86 |
/** The injected p attributes values. */
|
| 87 |
HashMap <String, String> injectedPAttributesValues;// = new ArrayList<String>(); |
| 88 |
|
| 89 |
/** The addinfos. */
|
| 90 |
boolean addinfos = false; |
| 91 |
|
| 92 |
/** The txtname. */
|
| 93 |
String txtname;
|
| 94 |
|
| 95 |
/** The base. */
|
| 96 |
String base;
|
| 97 |
|
| 98 |
/** The project. */
|
| 99 |
String project;
|
| 100 |
|
| 101 |
/** The lang. */
|
| 102 |
public String lang= "fr"; |
| 103 |
public String currentForm; |
| 104 |
public String currentAna; |
| 105 |
|
| 106 |
/**
|
| 107 |
* Sets the lang.
|
| 108 |
*
|
| 109 |
* @param lang the lang
|
| 110 |
* @return the java.lang. object
|
| 111 |
*/
|
| 112 |
public setLang(String lang) |
| 113 |
{
|
| 114 |
this.lang = lang;
|
| 115 |
} |
| 116 |
|
| 117 |
/**
|
| 118 |
* Instantiates a new xMLTX m2 cqp.
|
| 119 |
*
|
| 120 |
* @param url the url
|
| 121 |
*/
|
| 122 |
public XMLTXM2CQP(URL url){ |
| 123 |
try {
|
| 124 |
this.url = url;
|
| 125 |
inputData = url.openStream(); |
| 126 |
factory = XMLInputFactory.newInstance(); |
| 127 |
|
| 128 |
parser = factory.createXMLStreamReader(inputData); |
| 129 |
|
| 130 |
|
| 131 |
} catch (XMLStreamException ex) {
|
| 132 |
System.out.println(ex);
|
| 133 |
}catch (IOException ex) { |
| 134 |
System.out.println("IOException while parsing "); |
| 135 |
} |
| 136 |
} |
| 137 |
|
| 138 |
/**
|
| 139 |
* Sets the text info.
|
| 140 |
*
|
| 141 |
* @param name the name
|
| 142 |
* @param base the base
|
| 143 |
* @param project the project
|
| 144 |
*/
|
| 145 |
public void setTextInfo(String name, String base, String project) |
| 146 |
{
|
| 147 |
this.addinfos = true; |
| 148 |
this.txtname= name;
|
| 149 |
this.base = base;
|
| 150 |
this.project = project;
|
| 151 |
} |
| 152 |
|
| 153 |
/**
|
| 154 |
* Creates the output.
|
| 155 |
*
|
| 156 |
* @param outfile the outfile
|
| 157 |
* @return true, if successful
|
| 158 |
*/
|
| 159 |
private boolean createOutput(File outfile) |
| 160 |
{
|
| 161 |
try {
|
| 162 |
output = new BufferedWriter(new OutputStreamWriter(new FileOutputStream(outfile, outfile.exists()) , "UTF-8")); |
| 163 |
return true; |
| 164 |
} catch (Exception e) { |
| 165 |
System.err.println(e);
|
| 166 |
return false; |
| 167 |
} |
| 168 |
} |
| 169 |
|
| 170 |
/** The haspb. */
|
| 171 |
boolean haspb = false; |
| 172 |
|
| 173 |
/** The haslb. */
|
| 174 |
boolean haslb = false; |
| 175 |
|
| 176 |
/**
|
| 177 |
* Transform file.
|
| 178 |
*
|
| 179 |
* @param outfile the outfile
|
| 180 |
* @return true, if successful
|
| 181 |
*/
|
| 182 |
public boolean transformFile(File outfile) |
| 183 |
{
|
| 184 |
if(balisesToKeep == null) |
| 185 |
{
|
| 186 |
println "no element has been defined to be keeped"
|
| 187 |
return false; |
| 188 |
} |
| 189 |
|
| 190 |
haspb = false;
|
| 191 |
haslb = false;
|
| 192 |
|
| 193 |
boolean flagAna;
|
| 194 |
boolean flagForm;
|
| 195 |
boolean flagWord;
|
| 196 |
String vWord = ""; |
| 197 |
String vForm = ""; |
| 198 |
String vAna = ""; |
| 199 |
|
| 200 |
String lb_id = ""; |
| 201 |
String pb_id = ""; |
| 202 |
|
| 203 |
wordattributes = [:]; |
| 204 |
balisesfound = new HashMap<String, List<String>>(); |
| 205 |
|
| 206 |
|
| 207 |
if(!createOutput(outfile))
|
| 208 |
return false; |
| 209 |
|
| 210 |
if(sendToPAttributes != null) |
| 211 |
{
|
| 212 |
for(String tag: sendToPAttributes.keySet()) |
| 213 |
for(String attr : sendToPAttributes.get(tag)) |
| 214 |
injectedPAttributes.add(tag+attr); |
| 215 |
injectedPAttributesValues = [:]; |
| 216 |
} |
| 217 |
|
| 218 |
//output.write("<txmcorpus lang=\""+lang+"\">\n");
|
| 219 |
balisesfound.put("txmcorpus",["lang"]); |
| 220 |
try {
|
| 221 |
for (int event = parser.next(); event != XMLStreamConstants.END_DOCUMENT; event = parser.next()) { |
| 222 |
switch (event) {
|
| 223 |
case XMLStreamConstants.START_ELEMENT:
|
| 224 |
String localname = parser.getLocalName().toLowerCase();
|
| 225 |
|
| 226 |
// we will only declare found tags in cwb registry
|
| 227 |
if(balisesToKeep.contains(localname)) {
|
| 228 |
if(!balisesfound.containsKey(localname)) {
|
| 229 |
balisesfound.put(localname, []);
|
| 230 |
} |
| 231 |
|
| 232 |
List<String> attrlist = balisesfound.get(localname); |
| 233 |
for (int i= 0 ; i < parser.getAttributeCount() ;i++ ) |
| 234 |
if(!attrlist.contains(parser.getAttributeLocalName(i)))
|
| 235 |
attrlist.add(parser.getAttributeLocalName(i)); |
| 236 |
} |
| 237 |
|
| 238 |
switch (localname) {
|
| 239 |
case "w": // get word id !! |
| 240 |
wordattributes.put("id", parser.getAttributeValue(null, "id")); |
| 241 |
break;
|
| 242 |
|
| 243 |
case "form": |
| 244 |
flagForm = true;
|
| 245 |
currentForm = parser.getAttributeValue(null, "type"); |
| 246 |
if(currentForm == null) |
| 247 |
currentForm = "default";
|
| 248 |
vForm = "";
|
| 249 |
break;
|
| 250 |
|
| 251 |
case "ana": |
| 252 |
flagAna = true;
|
| 253 |
vAna ="";
|
| 254 |
|
| 255 |
currentAna = (parser.getAttributeValue(null,"type")); |
| 256 |
if(currentAna != null) |
| 257 |
currentAna = currentAna.substring(1)// remove the # |
| 258 |
else
|
| 259 |
flagAna = false;
|
| 260 |
break;
|
| 261 |
|
| 262 |
default:
|
| 263 |
|
| 264 |
if (sendToPAttributes != null) { |
| 265 |
//println "should store $localname ? with "+sendToPAttributes.keySet()
|
| 266 |
if (sendToPAttributes.keySet().contains(localname)) {
|
| 267 |
//println "store attr of "+localname
|
| 268 |
List<String> attrs = sendToPAttributes.get(localname); |
| 269 |
for (int i= 0 ; i < parser.getAttributeCount() ;i++ ) { |
| 270 |
if (attrs.contains(parser.getAttributeLocalName(i))) {
|
| 271 |
injectedPAttributesValues.put(localname+parser.getAttributeLocalName(i).toLowerCase(),parser.getAttributeValue(i)) |
| 272 |
} |
| 273 |
} |
| 274 |
} |
| 275 |
} |
| 276 |
|
| 277 |
if (balisesToKeep.contains(localname)) {
|
| 278 |
output.write("<"+localname);
|
| 279 |
//println "write <"+localname+"..."
|
| 280 |
//write attributes
|
| 281 |
boolean idwritten = false; |
| 282 |
boolean basewritten = false; |
| 283 |
boolean projectwritten = false; |
| 284 |
for (int i= 0 ; i < parser.getAttributeCount() ;i++ ) { |
| 285 |
String attrname = parser.getAttributeLocalName(i).toLowerCase();
|
| 286 |
if (attrname == "id") |
| 287 |
idwritten = true;
|
| 288 |
if (attrname == "base") |
| 289 |
basewritten = true;
|
| 290 |
if (attrname == "project") |
| 291 |
projectwritten = true;
|
| 292 |
|
| 293 |
output.write(" "+attrname+"=\""+parser.getAttributeValue(i).replace("&", "&").replace("\"", """)+"\"" ); |
| 294 |
} |
| 295 |
|
| 296 |
if (localname.equals("text")) |
| 297 |
if (addinfos) {
|
| 298 |
List<String> attrlist = balisesfound.get(localname); |
| 299 |
|
| 300 |
if (!idwritten) {
|
| 301 |
output.write(" id=\""+txtname+"\"") |
| 302 |
attrlist.add("id");
|
| 303 |
} |
| 304 |
if (!basewritten) {
|
| 305 |
output.write(" base=\""+base+"\""); |
| 306 |
attrlist.add("base");
|
| 307 |
} |
| 308 |
if (!projectwritten) {
|
| 309 |
output.write(" project=\""+project+"\""); |
| 310 |
attrlist.add("project");
|
| 311 |
} |
| 312 |
} |
| 313 |
|
| 314 |
// finalize tag
|
| 315 |
output.write(">\n");
|
| 316 |
} |
| 317 |
} |
| 318 |
break;
|
| 319 |
|
| 320 |
case XMLStreamConstants.END_ELEMENT:
|
| 321 |
String localname = parser.getLocalName().toLowerCase();
|
| 322 |
switch (localname) {
|
| 323 |
case "form": |
| 324 |
if(flagForm)
|
| 325 |
formhash.put(currentForm, vForm); |
| 326 |
flagForm = false;
|
| 327 |
break;
|
| 328 |
|
| 329 |
case "ana": |
| 330 |
if(flagAna)
|
| 331 |
anahash.put(currentAna, vAna); |
| 332 |
flagAna = false;
|
| 333 |
break;
|
| 334 |
|
| 335 |
case "w": |
| 336 |
vWord = "";
|
| 337 |
vWord = formhash.get("default").replaceAll("&", "&").replaceAll("<", "<"); // get default form |
| 338 |
for (String form : formhash.keySet()) // and the others |
| 339 |
if (form != "default") |
| 340 |
vWord += "\t"+formhash.get(form);
|
| 341 |
|
| 342 |
for (String type : wordattributes.keySet()) // only word id ? |
| 343 |
vWord+="\t"+wordattributes.get(type)
|
| 344 |
|
| 345 |
if (sendToPAttributes != null) // word attributes from structure properties |
| 346 |
{
|
| 347 |
//println "injectedPAttributesValues: "+injectedPAttributesValues
|
| 348 |
for(String pattr : injectedPAttributes) |
| 349 |
vWord+="\t"+injectedPAttributesValues.get(pattr) ;//les attributs injecter |
| 350 |
} |
| 351 |
|
| 352 |
for (String type : anahash.keySet()) // word annotations in txm:ana |
| 353 |
vWord+="\t"+anahash.get(type)
|
| 354 |
|
| 355 |
output.write(vWord+"\n");
|
| 356 |
vWord= "";
|
| 357 |
break;
|
| 358 |
|
| 359 |
default:
|
| 360 |
if (sendToPAttributes != null) // reset structure properties |
| 361 |
{
|
| 362 |
if (sendToPAttributes.keySet().contains(localname)) {
|
| 363 |
for (String attr : sendToPAttributes.get(localname)) { |
| 364 |
injectedPAttributesValues.put(attr, "N/A")
|
| 365 |
} |
| 366 |
} |
| 367 |
} |
| 368 |
|
| 369 |
if (balisesToKeep.contains(localname)) {
|
| 370 |
output.write("</"+localname+">\n"); |
| 371 |
} |
| 372 |
} |
| 373 |
break;
|
| 374 |
|
| 375 |
case XMLStreamConstants.CHARACTERS:
|
| 376 |
if (flagForm) {
|
| 377 |
vForm += parser.getText().trim(); |
| 378 |
} |
| 379 |
if (flagAna) {
|
| 380 |
vAna += parser.getText().trim(); |
| 381 |
} |
| 382 |
break;
|
| 383 |
} |
| 384 |
} |
| 385 |
//output.write("</txmcorpus>\n");
|
| 386 |
output.close(); |
| 387 |
if (parser != null) parser.close(); |
| 388 |
if (inputData != null) inputData.close(); |
| 389 |
} catch (Exception ex) { |
| 390 |
println "Error while parsing $url : "+ex
|
| 391 |
ex.printStackTrace(); |
| 392 |
if (parser != null) parser.close(); |
| 393 |
if (inputData != null) inputData.close(); |
| 394 |
return false; |
| 395 |
} |
| 396 |
return true; |
| 397 |
} |
| 398 |
|
| 399 |
/**
|
| 400 |
* Gets the p attributs.
|
| 401 |
*
|
| 402 |
* @return the p attributs
|
| 403 |
*/
|
| 404 |
public List<String> getpAttributs() |
| 405 |
{
|
| 406 |
def pAttributs = []; |
| 407 |
|
| 408 |
for (String wordattr : wordattributes.keySet()) { |
| 409 |
pAttributs.add(wordattr); |
| 410 |
} |
| 411 |
|
| 412 |
if (sendToPAttributes != null) |
| 413 |
for (String pAttr : this.injectedPAttributes) |
| 414 |
pAttributs.add(pAttr); |
| 415 |
|
| 416 |
for (String anakey : anahash.keySet()) { |
| 417 |
pAttributs.add(anakey); |
| 418 |
} |
| 419 |
|
| 420 |
return pAttributs;
|
| 421 |
} |
| 422 |
|
| 423 |
/**
|
| 424 |
* Gets the s attributs.
|
| 425 |
*
|
| 426 |
* @return the s attributs
|
| 427 |
*/
|
| 428 |
public List<String> getsAttributs() |
| 429 |
{
|
| 430 |
println balisesfound |
| 431 |
def sAttributs = []; |
| 432 |
for (String balise : this.balisesfound.keySet()) { |
| 433 |
List<String> sAtt = this.balisesfound.get(balise); |
| 434 |
String attributes = ""; |
| 435 |
for (String attr : sAtt) { |
| 436 |
attributes+="+"+attr;
|
| 437 |
} |
| 438 |
|
| 439 |
if (sAtt.size() > 0) |
| 440 |
sAttributs.add(balise +":"+attributes);
|
| 441 |
else
|
| 442 |
sAttributs.add(balise); |
| 443 |
} |
| 444 |
return sAttributs;
|
| 445 |
} |
| 446 |
|
| 447 |
/**
|
| 448 |
* Sets the balises to keep.
|
| 449 |
*
|
| 450 |
* @param balisesToKeep the new balises to keep
|
| 451 |
*/
|
| 452 |
public void setBalisesToKeep(List<String> balisesToKeep) |
| 453 |
{
|
| 454 |
if (balisesToKeep != null) |
| 455 |
this.balisesToKeep = balisesToKeep;
|
| 456 |
else
|
| 457 |
println("Warning: the list of elements to keep is null")
|
| 458 |
} |
| 459 |
|
| 460 |
/**
|
| 461 |
* Sets the defautl reference pattern
|
| 462 |
* TODO: not implemented
|
| 463 |
*
|
| 464 |
* @param balisesToKeep the new balises to keep
|
| 465 |
*/
|
| 466 |
public void setDefaultReference(String pattern, List<String> strucProperties) |
| 467 |
{
|
| 468 |
if (defaultReferencePattern != null) { |
| 469 |
this.defaultReferences = defaultReferences;
|
| 470 |
defaultReferencePattern = pattern; |
| 471 |
} |
| 472 |
} |
| 473 |
|
| 474 |
|
| 475 |
/**
|
| 476 |
* Sets the send to p attributes.
|
| 477 |
*
|
| 478 |
* @param sendus the sendus
|
| 479 |
*/
|
| 480 |
public void setSendToPAttributes(HashMap<String, List<String>> sendus) |
| 481 |
{
|
| 482 |
if (sendus != null) |
| 483 |
this.sendToPAttributes = sendus;
|
| 484 |
else
|
| 485 |
println("Warning: the pAttributes to inject is null")
|
| 486 |
} |
| 487 |
|
| 488 |
|
| 489 |
/**
|
| 490 |
* The main method.
|
| 491 |
*
|
| 492 |
* @param args the arguments
|
| 493 |
*/
|
| 494 |
public static void main(String[] args) { |
| 495 |
|
| 496 |
String rootDir = "/home/mdecorde/TXM/corpora/CORNEILLEMOLIERETER/txm/CORNEILLEMOLIERETER"; |
| 497 |
|
| 498 |
File srcfile = new File(rootDir,"CORNEILLEP_AGESILAS_1666.xml"); |
| 499 |
println srcfile.exists() |
| 500 |
File cqpfile = new File(rootDir, "out/CORNEILLEP_AGESILAS_1666.cqp"); |
| 501 |
new File(rootDir,"out").deleteDir() |
| 502 |
new File(rootDir,"out").mkdir() |
| 503 |
|
| 504 |
System.out.println("XMLTXM2CQP : "+srcfile+" >> "+cqpfile); |
| 505 |
def builder = new XMLTXM2CQP(srcfile.toURL()); |
| 506 |
def balises = ["text", "s"]; |
| 507 |
builder.setBalisesToKeep(balises); |
| 508 |
builder.transformFile(cqpfile); |
| 509 |
|
| 510 |
println("SATTRIBUTS: "+builder.getsAttributs());
|
| 511 |
println("PATTRIBUTS: "+builder.getpAttributs());
|
| 512 |
return;
|
| 513 |
} |
| 514 |
} |
| 515 |
|