com.lowagie.text.pdf.XfaForm Maven / Gradle / Ivy
Go to download
Show more of this group Show more artifacts with this name
Show all versions of itext2 Show documentation
Show all versions of itext2 Show documentation
Itext is a java library to create and manipulate PDFs.
This is a fork of version 2.1.7 the last MPL/LGPL version.
It's focused basically on mantain compatibility with newer bouncycastle releases and small bugfixes.
/*
* $Id: XfaForm.java 3961 2009-06-09 10:06:10Z blowagie $
*
* Copyright 2006 Paulo Soares
*
* The contents of this file are subject to the Mozilla Public License Version 1.1
* (the "License"); you may not use this file except in compliance with the License.
* You may obtain a copy of the License at http://www.mozilla.org/MPL/
*
* Software distributed under the License is distributed on an "AS IS" basis,
* WITHOUT WARRANTY OF ANY KIND, either express or implied. See the License
* for the specific language governing rights and limitations under the License.
*
* The Original Code is 'iText, a free JAVA-PDF library'.
*
* The Initial Developer of the Original Code is Bruno Lowagie. Portions created by
* the Initial Developer are Copyright (C) 1999, 2000, 2001, 2002 by Bruno Lowagie.
* All Rights Reserved.
* Co-Developer of the code is Paulo Soares. Portions created by the Co-Developer
* are Copyright (C) 2000, 2001, 2002 by Paulo Soares. All Rights Reserved.
*
* Contributor(s): all the names of the contributors are added in the source code
* where applicable.
*
* Alternatively, the contents of this file may be used under the terms of the
* LGPL license (the "GNU LIBRARY GENERAL PUBLIC LICENSE"), in which case the
* provisions of LGPL are applicable instead of those above. If you wish to
* allow use of your version of this file only under the terms of the LGPL
* License and not to allow others to use your version of this file under
* the MPL, indicate your decision by deleting the provisions above and
* replace them with the notice and other provisions required by the LGPL.
* If you do not delete the provisions above, a recipient may use your version
* of this file under either the MPL or the GNU LIBRARY GENERAL PUBLIC LICENSE.
*
* This library is free software; you can redistribute it and/or modify it
* under the terms of the MPL as stated above or under the terms of the GNU
* Library General Public License as published by the Free Software Foundation;
* either version 2 of the License, or any later version.
*
* This library is distributed in the hope that it will be useful, but WITHOUT
* ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS
* FOR A PARTICULAR PURPOSE. See the GNU Library general Public License for more
* details.
*
* If you didn't download this code from the following link, you should check if
* you aren't using an obsolete version:
* http://www.lowagie.com/iText/
*/
package com.lowagie.text.pdf;
import java.io.ByteArrayInputStream;
import java.io.ByteArrayOutputStream;
import java.io.IOException;
import java.util.ArrayList;
import java.util.Collection;
import java.util.EmptyStackException;
import java.util.HashMap;
import java.util.Iterator;
import javax.xml.parsers.DocumentBuilder;
import javax.xml.parsers.DocumentBuilderFactory;
import javax.xml.parsers.ParserConfigurationException;
import org.w3c.dom.Node;
import org.xml.sax.SAXException;
import com.lowagie.text.xml.XmlDomWriter;
/**
* Processes XFA forms.
* @author Paulo Soares ([email protected])
*/
public class XfaForm {
private Xml2SomTemplate templateSom;
private Node templateNode;
private Xml2SomDatasets datasetsSom;
private Node datasetsNode;
private AcroFieldsSearch acroFieldsSom;
private PdfReader reader;
private boolean xfaPresent;
private org.w3c.dom.Document domDocument;
private boolean changed;
public static final String XFA_DATA_SCHEMA = "http://www.xfa.org/schema/xfa-data/1.0/";
/**
* An empty constructor to build on.
*/
public XfaForm() {
}
/**
* Return the XFA Object, could be an array, could be a Stream.
* Returns null f no XFA Object is present.
* @param reader a PdfReader instance
* @return the XFA object
* @since 2.1.3
*/
public static PdfObject getXfaObject(PdfReader reader) {
PdfDictionary af = (PdfDictionary)PdfReader.getPdfObjectRelease(reader.getCatalog().get(PdfName.ACROFORM));
if (af == null) {
return null;
}
return PdfReader.getPdfObjectRelease(af.get(PdfName.XFA));
}
/**
* A constructor from a PdfReader
. It basically does everything
* from finding the XFA stream to the XML parsing.
* @param reader the reader
* @throws java.io.IOException on error
* @throws javax.xml.parsers.ParserConfigurationException on error
* @throws org.xml.sax.SAXException on error
*/
public XfaForm(PdfReader reader) throws IOException, ParserConfigurationException, SAXException {
this.reader = reader;
PdfObject xfa = getXfaObject(reader);
if (xfa == null) {
xfaPresent = false;
return;
}
xfaPresent = true;
ByteArrayOutputStream bout = new ByteArrayOutputStream();
if (xfa.isArray()) {
PdfArray ar = (PdfArray)xfa;
for (int k = 1; k < ar.size(); k += 2) {
PdfObject ob = ar.getDirectObject(k);
if (ob instanceof PRStream) {
byte[] b = PdfReader.getStreamBytes((PRStream)ob);
bout.write(b);
}
}
}
else if (xfa instanceof PRStream) {
byte[] b = PdfReader.getStreamBytes((PRStream)xfa);
bout.write(b);
}
bout.close();
DocumentBuilderFactory fact = DocumentBuilderFactory.newInstance();
fact.setNamespaceAware(true);
DocumentBuilder db = fact.newDocumentBuilder();
domDocument = db.parse(new ByteArrayInputStream(bout.toByteArray()));
extractNodes();
}
/**
* Extracts the nodes from the domDocument.
* @since 2.1.5
*/
private void extractNodes() {
Node n = domDocument.getFirstChild();
while (n.getChildNodes().getLength() == 0) {
n = n.getNextSibling();
}
n = n.getFirstChild();
while (n != null) {
if (n.getNodeType() == Node.ELEMENT_NODE) {
String s = n.getLocalName();
if (s.equals("template")) {
templateNode = n;
templateSom = new Xml2SomTemplate(n);
}
else if (s.equals("datasets")) {
datasetsNode = n;
datasetsSom = new Xml2SomDatasets(n.getFirstChild());
}
}
n = n.getNextSibling();
}
}
/**
* Sets the XFA key from a byte array. The old XFA is erased.
* @param form the data
* @param reader the reader
* @param writer the writer
* @throws java.io.IOException on error
*/
public static void setXfa(XfaForm form, PdfReader reader, PdfWriter writer) throws IOException {
PdfDictionary af = (PdfDictionary)PdfReader.getPdfObjectRelease(reader.getCatalog().get(PdfName.ACROFORM));
if (af == null) {
return;
}
PdfObject xfa = getXfaObject(reader);
if (xfa.isArray()) {
PdfArray ar = (PdfArray)xfa;
int t = -1;
int d = -1;
for (int k = 0; k < ar.size(); k += 2) {
PdfString s = ar.getAsString(k);
if ("template".equals(s.toString())) {
t = k + 1;
}
if ("datasets".equals(s.toString())) {
d = k + 1;
}
}
if (t > -1 && d > -1) {
reader.killXref(ar.getAsIndirectObject(t));
reader.killXref(ar.getAsIndirectObject(d));
PdfStream tStream = new PdfStream(serializeDoc(form.templateNode));
tStream.flateCompress(writer.getCompressionLevel());
ar.set(t, writer.addToBody(tStream).getIndirectReference());
PdfStream dStream = new PdfStream(serializeDoc(form.datasetsNode));
dStream.flateCompress(writer.getCompressionLevel());
ar.set(d, writer.addToBody(dStream).getIndirectReference());
af.put(PdfName.XFA, new PdfArray(ar));
return;
}
}
reader.killXref(af.get(PdfName.XFA));
PdfStream str = new PdfStream(serializeDoc(form.domDocument));
str.flateCompress(writer.getCompressionLevel());
PdfIndirectReference ref = writer.addToBody(str).getIndirectReference();
af.put(PdfName.XFA, ref);
}
/**
* Sets the XFA key from the instance data. The old XFA is erased.
* @param writer the writer
* @throws java.io.IOException on error
*/
public void setXfa(PdfWriter writer) throws IOException {
setXfa(this, reader, writer);
}
/**
* Serializes a XML document to a byte array.
* @param n the XML document
* @throws java.io.IOException on error
* @return the serialized XML document
*/
public static byte[] serializeDoc(Node n) throws IOException {
XmlDomWriter xw = new XmlDomWriter();
ByteArrayOutputStream fout = new ByteArrayOutputStream();
xw.setOutput(fout, null);
xw.setCanonical(false);
xw.write(n);
fout.close();
return fout.toByteArray();
}
/**
* Returns true
if it is a XFA form.
* @return true
if it is a XFA form
*/
public boolean isXfaPresent() {
return xfaPresent;
}
/**
* Gets the top level DOM document.
* @return the top level DOM document
*/
public org.w3c.dom.Document getDomDocument() {
return domDocument;
}
/**
* Finds the complete field name contained in the "classic" forms from a partial
* name.
* @param name the complete or partial name
* @param af the fields
* @return the complete name or null
if not found
*/
public String findFieldName(String name, AcroFields af) {
HashMap items = af.getFields();
if (items.containsKey(name))
return name;
if (acroFieldsSom == null) {
if (items.isEmpty() && xfaPresent)
acroFieldsSom = new AcroFieldsSearch(datasetsSom.getName2Node().keySet());
else
acroFieldsSom = new AcroFieldsSearch(items.keySet());
}
if (acroFieldsSom.getAcroShort2LongName().containsKey(name))
return (String)acroFieldsSom.getAcroShort2LongName().get(name);
return acroFieldsSom.inverseSearchGlobal(Xml2Som.splitParts(name));
}
/**
* Finds the complete SOM name contained in the datasets section from a
* possibly partial name.
* @param name the complete or partial name
* @return the complete name or null
if not found
*/
public String findDatasetsName(String name) {
if (datasetsSom.getName2Node().containsKey(name))
return name;
return datasetsSom.inverseSearchGlobal(Xml2Som.splitParts(name));
}
/**
* Finds the Node
contained in the datasets section from a
* possibly partial name.
* @param name the complete or partial name
* @return the Node
or null
if not found
*/
public Node findDatasetsNode(String name) {
if (name == null)
return null;
name = findDatasetsName(name);
if (name == null)
return null;
return (Node)datasetsSom.getName2Node().get(name);
}
/**
* Gets all the text contained in the child nodes of this node.
* @param n the Node
* @return the text found or "" if no text was found
*/
public static String getNodeText(Node n) {
if (n == null)
return "";
return getNodeText(n, "");
}
private static String getNodeText(Node n, String name) {
Node n2 = n.getFirstChild();
while (n2 != null) {
if (n2.getNodeType() == Node.ELEMENT_NODE) {
name = getNodeText(n2, name);
}
else if (n2.getNodeType() == Node.TEXT_NODE) {
name += n2.getNodeValue();
}
n2 = n2.getNextSibling();
}
return name;
}
/**
* Sets the text of this node. All the child's node are deleted and a new
* child text node is created.
* @param n the Node
to add the text to
* @param text the text to add
*/
public void setNodeText(Node n, String text) {
if (n == null)
return;
Node nc = null;
while ((nc = n.getFirstChild()) != null) {
n.removeChild(nc);
}
if (n.getAttributes().getNamedItemNS(XFA_DATA_SCHEMA, "dataNode") != null)
n.getAttributes().removeNamedItemNS(XFA_DATA_SCHEMA, "dataNode");
n.appendChild(domDocument.createTextNode(text));
changed = true;
}
/**
* Sets the XFA form flag signaling that this is a valid XFA form.
* @param xfaPresent the XFA form flag signaling that this is a valid XFA form
*/
public void setXfaPresent(boolean xfaPresent) {
this.xfaPresent = xfaPresent;
}
/**
* Sets the top DOM document.
* @param domDocument the top DOM document
*/
public void setDomDocument(org.w3c.dom.Document domDocument) {
this.domDocument = domDocument;
extractNodes();
}
/**
* Gets the PdfReader
used by this instance.
* @return the PdfReader
used by this instance
*/
public PdfReader getReader() {
return reader;
}
/**
* Sets the PdfReader
to be used by this instance.
* @param reader the PdfReader
to be used by this instance
*/
public void setReader(PdfReader reader) {
this.reader = reader;
}
/**
* Checks if this XFA form was changed.
* @return true
if this XFA form was changed
*/
public boolean isChanged() {
return changed;
}
/**
* Sets the changed status of this XFA instance.
* @param changed the changed status of this XFA instance
*/
public void setChanged(boolean changed) {
this.changed = changed;
}
/**
* A structure to store each part of a SOM name and link it to the next part
* beginning from the lower hierarchy.
*/
public static class InverseStore {
protected ArrayList part = new ArrayList();
protected ArrayList follow = new ArrayList();
/**
* Gets the full name by traversing the hierarchy using only the
* index 0.
* @return the full name
*/
public String getDefaultName() {
InverseStore store = this;
while (true) {
Object obj = store.follow.get(0);
if (obj instanceof String)
return (String)obj;
store = (InverseStore)obj;
}
}
/**
* Search the current node for a similar name. A similar name starts
* with the same name but has a different index. For example, "detail[3]"
* is similar to "detail[9]". The main use is to discard names that
* correspond to out of bounds records.
* @param name the name to search
* @return true
if a similitude was found
*/
public boolean isSimilar(String name) {
int idx = name.indexOf('[');
name = name.substring(0, idx + 1);
for (int k = 0; k < part.size(); ++k) {
if (((String)part.get(k)).startsWith(name))
return true;
}
return false;
}
}
/**
* Another stack implementation. The main use is to facilitate
* the porting to other languages.
*/
public static class Stack2 extends ArrayList {
private static final long serialVersionUID = -7451476576174095212L;
/**
* Looks at the object at the top of this stack without removing it from the stack.
* @return the object at the top of this stack
*/
public Object peek() {
if (size() == 0)
throw new EmptyStackException();
return get(size() - 1);
}
/**
* Removes the object at the top of this stack and returns that object as the value of this function.
* @return the object at the top of this stack
*/
public Object pop() {
if (size() == 0)
throw new EmptyStackException();
Object ret = get(size() - 1);
remove(size() - 1);
return ret;
}
/**
* Pushes an item onto the top of this stack.
* @param item the item to be pushed onto this stack
* @return the item
argument
*/
public Object push(Object item) {
add(item);
return item;
}
/**
* Tests if this stack is empty.
* @return true
if and only if this stack contains no items; false
otherwise
*/
public boolean empty() {
return size() == 0;
}
}
/**
* A class for some basic SOM processing.
*/
public static class Xml2Som {
/**
* The order the names appear in the XML, depth first.
*/
protected ArrayList order;
/**
* The mapping of full names to nodes.
*/
protected HashMap name2Node;
/**
* The data to do a search from the bottom hierarchy.
*/
protected HashMap inverseSearch;
/**
* A stack to be used when parsing.
*/
protected Stack2 stack;
/**
* A temporary store for the repetition count.
*/
protected int anform;
/**
* Escapes a SOM string fragment replacing "." with "\.".
* @param s the unescaped string
* @return the escaped string
*/
public static String escapeSom(String s) {
int idx = s.indexOf('.');
if (idx < 0)
return s;
StringBuffer sb = new StringBuffer();
int last = 0;
while (idx >= 0) {
sb.append(s.substring(last, idx));
sb.append('\\');
last = idx;
idx = s.indexOf('.', idx + 1);
}
sb.append(s.substring(last));
return sb.toString();
}
/**
* Unescapes a SOM string fragment replacing "\." with ".".
* @param s the escaped string
* @return the unescaped string
*/
public static String unescapeSom(String s) {
int idx = s.indexOf('\\');
if (idx < 0)
return s;
StringBuffer sb = new StringBuffer();
int last = 0;
while (idx >= 0) {
sb.append(s.substring(last, idx));
last = idx + 1;
idx = s.indexOf('\\', idx + 1);
}
sb.append(s.substring(last));
return sb.toString();
}
/**
* Outputs the stack as the sequence of elements separated
* by '.'.
* @return the stack as the sequence of elements separated by '.'
*/
protected String printStack() {
if (stack.empty())
return "";
StringBuffer s = new StringBuffer();
for (int k = 0; k < stack.size(); ++k)
s.append('.').append((String)stack.get(k));
return s.substring(1);
}
/**
* Gets the name with the #subform
removed.
* @param s the long name
* @return the short name
*/
public static String getShortName(String s) {
int idx = s.indexOf(".#subform[");
if (idx < 0)
return s;
int last = 0;
StringBuffer sb = new StringBuffer();
while (idx >= 0) {
sb.append(s.substring(last, idx));
idx = s.indexOf("]", idx + 10);
if (idx < 0)
return sb.toString();
last = idx + 1;
idx = s.indexOf(".#subform[", last);
}
sb.append(s.substring(last));
return sb.toString();
}
/**
* Adds a SOM name to the search node chain.
* @param unstack the SOM name
*/
public void inverseSearchAdd(String unstack) {
inverseSearchAdd(inverseSearch, stack, unstack);
}
/**
* Adds a SOM name to the search node chain.
* @param inverseSearch the start point
* @param stack the stack with the separated SOM parts
* @param unstack the full name
*/
public static void inverseSearchAdd(HashMap inverseSearch, Stack2 stack, String unstack) {
String last = (String)stack.peek();
InverseStore store = (InverseStore)inverseSearch.get(last);
if (store == null) {
store = new InverseStore();
inverseSearch.put(last, store);
}
for (int k = stack.size() - 2; k >= 0; --k) {
last = (String)stack.get(k);
InverseStore store2;
int idx = store.part.indexOf(last);
if (idx < 0) {
store.part.add(last);
store2 = new InverseStore();
store.follow.add(store2);
}
else
store2 = (InverseStore)store.follow.get(idx);
store = store2;
}
store.part.add("");
store.follow.add(unstack);
}
/**
* Searches the SOM hierarchy from the bottom.
* @param parts the SOM parts
* @return the full name or null
if not found
*/
public String inverseSearchGlobal(ArrayList parts) {
if (parts.isEmpty())
return null;
InverseStore store = (InverseStore)inverseSearch.get(parts.get(parts.size() - 1));
if (store == null)
return null;
for (int k = parts.size() - 2; k >= 0; --k) {
String part = (String)parts.get(k);
int idx = store.part.indexOf(part);
if (idx < 0) {
if (store.isSimilar(part))
return null;
return store.getDefaultName();
}
store = (InverseStore)store.follow.get(idx);
}
return store.getDefaultName();
}
/**
* Splits a SOM name in the individual parts.
* @param name the full SOM name
* @return the split name
*/
public static Stack2 splitParts(String name) {
while (name.startsWith("."))
name = name.substring(1);
Stack2 parts = new Stack2();
int last = 0;
int pos = 0;
String part;
while (true) {
pos = last;
while (true) {
pos = name.indexOf('.', pos);
if (pos < 0)
break;
if (name.charAt(pos - 1) == '\\')
++pos;
else
break;
}
if (pos < 0)
break;
part = name.substring(last, pos);
if (!part.endsWith("]"))
part += "[0]";
parts.add(part);
last = pos + 1;
}
part = name.substring(last);
if (!part.endsWith("]"))
part += "[0]";
parts.add(part);
return parts;
}
/**
* Gets the order the names appear in the XML, depth first.
* @return the order the names appear in the XML, depth first
*/
public ArrayList getOrder() {
return order;
}
/**
* Sets the order the names appear in the XML, depth first
* @param order the order the names appear in the XML, depth first
*/
public void setOrder(ArrayList order) {
this.order = order;
}
/**
* Gets the mapping of full names to nodes.
* @return the mapping of full names to nodes
*/
public HashMap getName2Node() {
return name2Node;
}
/**
* Sets the mapping of full names to nodes.
* @param name2Node the mapping of full names to nodes
*/
public void setName2Node(HashMap name2Node) {
this.name2Node = name2Node;
}
/**
* Gets the data to do a search from the bottom hierarchy.
* @return the data to do a search from the bottom hierarchy
*/
public HashMap getInverseSearch() {
return inverseSearch;
}
/**
* Sets the data to do a search from the bottom hierarchy.
* @param inverseSearch the data to do a search from the bottom hierarchy
*/
public void setInverseSearch(HashMap inverseSearch) {
this.inverseSearch = inverseSearch;
}
}
/**
* Processes the datasets section in the XFA form.
*/
public static class Xml2SomDatasets extends Xml2Som {
/**
* Creates a new instance from the datasets node. This expects
* not the datasets but the data node that comes below.
* @param n the datasets node
*/
public Xml2SomDatasets(Node n) {
order = new ArrayList();
name2Node = new HashMap();
stack = new Stack2();
anform = 0;
inverseSearch = new HashMap();
processDatasetsInternal(n);
}
/**
* Inserts a new Node
that will match the short name.
* @param n the datasets top Node
* @param shortName the short name
* @return the new Node
of the inserted name
*/
public Node insertNode(Node n, String shortName) {
Stack2 stack = splitParts(shortName);
org.w3c.dom.Document doc = n.getOwnerDocument();
Node n2 = null;
n = n.getFirstChild();
for (int k = 0; k < stack.size(); ++k) {
String part = (String)stack.get(k);
int idx = part.lastIndexOf('[');
String name = part.substring(0, idx);
idx = Integer.parseInt(part.substring(idx + 1, part.length() - 1));
int found = -1;
for (n2 = n.getFirstChild(); n2 != null; n2 = n2.getNextSibling()) {
if (n2.getNodeType() == Node.ELEMENT_NODE) {
String s = escapeSom(n2.getLocalName());
if (s.equals(name)) {
++found;
if (found == idx)
break;
}
}
}
for (; found < idx; ++found) {
n2 = doc.createElementNS(null, name);
n2 = n.appendChild(n2);
Node attr = doc.createAttributeNS(XFA_DATA_SCHEMA, "dataNode");
attr.setNodeValue("dataGroup");
n2.getAttributes().setNamedItemNS(attr);
}
n = n2;
}
inverseSearchAdd(inverseSearch, stack, shortName);
name2Node.put(shortName, n2);
order.add(shortName);
return n2;
}
private static boolean hasChildren(Node n) {
Node dataNodeN = n.getAttributes().getNamedItemNS(XFA_DATA_SCHEMA, "dataNode");
if (dataNodeN != null) {
String dataNode = dataNodeN.getNodeValue();
if ("dataGroup".equals(dataNode))
return true;
else if ("dataValue".equals(dataNode))
return false;
}
if (!n.hasChildNodes())
return false;
Node n2 = n.getFirstChild();
while (n2 != null) {
if (n2.getNodeType() == Node.ELEMENT_NODE) {
return true;
}
n2 = n2.getNextSibling();
}
return false;
}
private void processDatasetsInternal(Node n) {
HashMap ss = new HashMap();
Node n2 = n.getFirstChild();
while (n2 != null) {
if (n2.getNodeType() == Node.ELEMENT_NODE) {
String s = escapeSom(n2.getLocalName());
Integer i = (Integer)ss.get(s);
if (i == null)
i = Integer.valueOf(0);
else
i = Integer.valueOf(i.intValue() + 1);
ss.put(s, i);
if (hasChildren(n2)) {
stack.push(s + "[" + i.toString() + "]");
processDatasetsInternal(n2);
stack.pop();
}
else {
stack.push(s + "[" + i.toString() + "]");
String unstack = printStack();
order.add(unstack);
inverseSearchAdd(unstack);
name2Node.put(unstack, n2);
stack.pop();
}
}
n2 = n2.getNextSibling();
}
}
}
/**
* A class to process "classic" fields.
*/
public static class AcroFieldsSearch extends Xml2Som {
private HashMap acroShort2LongName;
/**
* Creates a new instance from a Collection with the full names.
* @param items the Collection
*/
public AcroFieldsSearch(Collection items) {
inverseSearch = new HashMap();
acroShort2LongName = new HashMap();
for (Iterator it = items.iterator(); it.hasNext();) {
String itemName = (String)it.next();
String itemShort = getShortName(itemName);
acroShort2LongName.put(itemShort, itemName);
inverseSearchAdd(inverseSearch, splitParts(itemShort), itemName);
}
}
/**
* Gets the mapping from short names to long names. A long
* name may contain the #subform name part.
* @return the mapping from short names to long names
*/
public HashMap getAcroShort2LongName() {
return acroShort2LongName;
}
/**
* Sets the mapping from short names to long names. A long
* name may contain the #subform name part.
* @param acroShort2LongName the mapping from short names to long names
*/
public void setAcroShort2LongName(HashMap acroShort2LongName) {
this.acroShort2LongName = acroShort2LongName;
}
}
/**
* Processes the template section in the XFA form.
*/
public static class Xml2SomTemplate extends Xml2Som {
private boolean dynamicForm;
private int templateLevel;
/**
* Creates a new instance from the datasets node.
* @param n the template node
*/
public Xml2SomTemplate(Node n) {
order = new ArrayList();
name2Node = new HashMap();
stack = new Stack2();
anform = 0;
templateLevel = 0;
inverseSearch = new HashMap();
processTemplate(n, null);
}
/**
* Gets the field type as described in the template
section of the XFA.
* @param s the exact template name
* @return the field type or null
if not found
*/
public String getFieldType(String s) {
Node n = (Node)name2Node.get(s);
if (n == null)
return null;
if (n.getLocalName().equals("exclGroup"))
return "exclGroup";
Node ui = n.getFirstChild();
while (ui != null) {
if (ui.getNodeType() == Node.ELEMENT_NODE && ui.getLocalName().equals("ui")) {
break;
}
ui = ui.getNextSibling();
}
if (ui == null)
return null;
Node type = ui.getFirstChild();
while (type != null) {
if (type.getNodeType() == Node.ELEMENT_NODE && !(type.getLocalName().equals("extras") && type.getLocalName().equals("picture"))) {
return type.getLocalName();
}
type = type.getNextSibling();
}
return null;
}
private void processTemplate(Node n, HashMap ff) {
if (ff == null)
ff = new HashMap();
HashMap ss = new HashMap();
Node n2 = n.getFirstChild();
while (n2 != null) {
if (n2.getNodeType() == Node.ELEMENT_NODE) {
String s = n2.getLocalName();
if (s.equals("subform")) {
Node name = n2.getAttributes().getNamedItem("name");
String nn = "#subform";
boolean annon = true;
if (name != null) {
nn = escapeSom(name.getNodeValue());
annon = false;
}
Integer i;
if (annon) {
i = Integer.valueOf(anform);
++anform;
}
else {
i = (Integer)ss.get(nn);
if (i == null)
i = Integer.valueOf(0);
else
i = Integer.valueOf(i.intValue() + 1);
ss.put(nn, i);
}
stack.push(nn + "[" + i.toString() + "]");
++templateLevel;
if (annon)
processTemplate(n2, ff);
else
processTemplate(n2, null);
--templateLevel;
stack.pop();
}
else if (s.equals("field") || s.equals("exclGroup")) {
Node name = n2.getAttributes().getNamedItem("name");
if (name != null) {
String nn = escapeSom(name.getNodeValue());
Integer i = (Integer)ff.get(nn);
if (i == null) {
i = Integer.valueOf(0);
}
else {
i = Integer.valueOf(i.intValue() + 1);
}
ff.put(nn, i);
stack.push(nn + "[" + i.toString() + "]");
String unstack = printStack();
order.add(unstack);
inverseSearchAdd(unstack);
name2Node.put(unstack, n2);
stack.pop();
}
}
else if (!dynamicForm && templateLevel > 0 && s.equals("occur")) {
int initial = 1;
int min = 1;
int max = 1;
Node a = n2.getAttributes().getNamedItem("initial");
if (a != null)
try{initial = Integer.parseInt(a.getNodeValue().trim());}catch(Exception e){}
a = n2.getAttributes().getNamedItem("min");
if (a != null)
try{min = Integer.parseInt(a.getNodeValue().trim());}catch(Exception e){}
a = n2.getAttributes().getNamedItem("max");
if (a != null)
try{max = Integer.parseInt(a.getNodeValue().trim());}catch(Exception e){}
if (initial != min || min != max)
dynamicForm = true;
}
}
n2 = n2.getNextSibling();
}
}
/**
* true
if it's a dynamic form; false
* if it's a static form.
* @return true
if it's a dynamic form; false
* if it's a static form
*/
public boolean isDynamicForm() {
return dynamicForm;
}
/**
* Sets the dynamic form flag. It doesn't change the template.
* @param dynamicForm the dynamic form flag
*/
public void setDynamicForm(boolean dynamicForm) {
this.dynamicForm = dynamicForm;
}
}
/**
* Gets the class that contains the template processing section of the XFA.
* @return the class that contains the template processing section of the XFA
*/
public Xml2SomTemplate getTemplateSom() {
return templateSom;
}
/**
* Sets the class that contains the template processing section of the XFA
* @param templateSom the class that contains the template processing section of the XFA
*/
public void setTemplateSom(Xml2SomTemplate templateSom) {
this.templateSom = templateSom;
}
/**
* Gets the class that contains the datasets processing section of the XFA.
* @return the class that contains the datasets processing section of the XFA
*/
public Xml2SomDatasets getDatasetsSom() {
return datasetsSom;
}
/**
* Sets the class that contains the datasets processing section of the XFA.
* @param datasetsSom the class that contains the datasets processing section of the XFA
*/
public void setDatasetsSom(Xml2SomDatasets datasetsSom) {
this.datasetsSom = datasetsSom;
}
/**
* Gets the class that contains the "classic" fields processing.
* @return the class that contains the "classic" fields processing
*/
public AcroFieldsSearch getAcroFieldsSom() {
return acroFieldsSom;
}
/**
* Sets the class that contains the "classic" fields processing.
* @param acroFieldsSom the class that contains the "classic" fields processing
*/
public void setAcroFieldsSom(AcroFieldsSearch acroFieldsSom) {
this.acroFieldsSom = acroFieldsSom;
}
/**
* Gets the Node
that corresponds to the datasets part.
* @return the Node
that corresponds to the datasets part
*/
public Node getDatasetsNode() {
return datasetsNode;
}
}