Laboratoire ICAR » Plateforme TXM

package org.txm.scripts.importer

import javax.xml.stream.XMLStreamException

import org.apache.tools.ant.types.resources.selectors.InstanceOf;

import groovy.xml.*

def root = new File("/home/mdecorde/xml/temoignagesnice/corpus Matrice - fichiers xmlisés/")

File srcdir = new File(root, "orig");

File outdir = new File(root, "tmp");

File okdir = new File(root, "ok");

File ok2dir = new File(root, "ok2");

File temoignagedir = new File(root, "temoignages");

ok2dir.deleteDir()

ok2dir.mkdir()

//rename title -> head

for (def file : outdir.listFiles()) {

        if (!file.getName().endsWith(".xml")) continue;

        def doc = new XmlParser().parse(file);

        for (def note : doc.body.chapter.title) {

                println note

                note.name = "head"

        }

        new File(ok2dir, file.getName()).withWriter("UTF-8") { writer ->

                new XmlNodePrinter(new PrintWriter(writer)).print(doc)

        }

}

/*

// ADD chapter@title

for (def file : outdir.listFiles()) {

        if (!file.getName().endsWith(".xml")) continue;

        def doc = new XmlParser().parse(file);

        for (def chapter : doc.body.chapter) {

                for (def title : chapter.title) {

                        chapter.@title = title.text()

                        break;

                }

        }

        new File(ok2dir, file.getName()).withWriter("UTF-8") { writer ->

                new XmlNodePrinter(new PrintWriter(writer)).print(doc)

        }

}

*/

//FIX figure and caption inclusions

/*

for (def file : outdir.listFiles()) {

        if (!file.getName().endsWith(".xml")) continue;

        def doc = new XmlParser().parse(file);

        for (def note : doc.body."**".figure) {

                note.name = "note"

                //println "fig : $note"

                //if ("Image :" == note.text()) {

                        def children = note.parent().children()

                        int i = children.indexOf(note)

                        //println i + " < "+children.size()

                        def nextChild = children[i+1]

                        if (nextChild != null && nextChild.name().toString() == "caption") {

                                println nextChild

                                note.value = "Images : "+note.text()// + " "+nextChild.text()

                                //println note

                                children.remove(i+1)

                                note.append(nextChild)

                        }

                //}

        }

        new File(ok2dir, file.getName()).withWriter("UTF-8") { writer ->

                new XmlNodePrinter(new PrintWriter(writer)).print(doc)

        }

}

*/

/*

// DOCBOOK -> DOCBOOK TEXT ONLY

for (def file : outdir.listFiles()) {

        def doc = new XmlParser().parse(file);

        def body = null

        def bookinfo = null

        def preface = null

        for (def e : doc.body) body = e

        for (def e : doc.bookinfo) {

                doc.remove(e)

        }

        for (def e : doc.preface) {

                doc.remove(e)

        }

        for (def e : doc.appendix) {

                doc.remove(e)

        }

        for (def e : doc.chapter) {

                doc.remove(e)

        }

        if (body == null) {

                println "error text: "+file

                continue

        }

        new File(ok2dir, file.getName()).withWriter("UTF-8") { writer ->

                new XmlNodePrinter(new PrintWriter(writer)).print(doc)

        }

        //                writer.print XmlUtil.serialize(new StreamingMarkupBuilder().bind {

        //                        mkp.yield body

        //                  })

}

*/

//DOCBOOK to TEI

/*

 for (def file : outdir.listFiles()) {

 def doc = new XmlParser().parse(file);

 def body = null

 //def bookinfo = null

 for (def e : doc.body) body = e

 //for (def e : doc.bookinfo) bookinfo = e

 //println body.getClass()

 if (body == null) {

 println "error text: "+file

 continue

 }

 //        bookinfo.name = "teiHeader"

 body.name = "text"

 def teins = new groovy.xml.Namespace("http://www.tei-c.org/ns/1.0",'tei')

 //        for (def node : body."**") {

 //                if (node instanceof String) continue

 //                def name = node.name()

 //                if (name instanceof String)

 //                        node.name = teins.get(name)

 //                else 

 //                        node.name = teins.get(name.getLocalPart())

 //        }

 for (def figure : body."**".figure) {

 figure.name = "note"

 figure.value = "Image : " + figure.caption.text()

 }

 for (def chapter : body."**".chapter) {

 chapter.name = "div"

 chapter.@type = "chapter"

 }

 for (def caption : body."**".title) {

 caption.name = "head"

 }

 for (def para : body."**".para) {

 para.name = "p"

 }

 def newdoc = new Node(null, "TEI");

 newdoc.@xmlns="http://www.tei-c.org/ns/1.0";

 newdoc.append(new Node(null, "teiHeader"))

 newdoc.append(body)

 new File(okdir, file.getName()).withWriter("UTF-8") { writer ->

 new XmlNodePrinter(new PrintWriter(writer)).print(newdoc)

 }

 //                writer.print XmlUtil.serialize(new StreamingMarkupBuilder().bind {

 //                        mkp.yield body

 //                  })

 }

 */

// remove TEI

/*outdir.deleteDir()

 outdir.mkdir()

 def errors = []

 for (def file : srcdir.listFiles()) {

 if (file.isDirectory()) continue;

 //new EncodingConverter(file, "Windows-1252", "UTF-8")

 File outfile = new File(outdir, file.getName());

 outfile.withWriter("UTF-8") { writer ->

 file.eachLine("UTF-8") { line ->

 if (line.trim() == "<TEI>") {

 } else if (line.trim() == "</TEI>") {

 writer.println("</book>")

 } else if (line.trim() == "<book lang=\"fr\"/>") {

 writer.println("<book lang=\"fr\">")

 } else {

 writer.println(line)

 }

 }                

 }

 try {

 ValidateXml.testAndThrow(outfile);

 } catch (XMLStreamException e) {

 println file.getName() + " : "+ e.getMessage()

 errors << file

 if (e.getMessage().contains('Message: The element type "TEI" must be terminated by the matching end-tag "</TEI>"')) {

 println "Delete line : "+e.location.lineNumber

 }

 println ""

 }

 }

 */

println "done"

//if (errors.size() > 0)

//        println ""+errors.size()+" errors : $errors"

//String content = file.getText("Windows-1252")

//println content