org.apache.batik.gvt.text.ArabicTextHandler Maven / Gradle / Ivy
Go to download
Show more of this group Show more artifacts with this name
Show all versions of batik-gvt Show documentation
Show all versions of batik-gvt Show documentation
Batik Graphics Vector Tree (GVT)
The newest version!
/*
Licensed to the Apache Software Foundation (ASF) under one or more
contributor license agreements. See the NOTICE file distributed with
this work for additional information regarding copyright ownership.
The ASF licenses this file to You under the Apache License, Version 2.0
(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.apache.org/licenses/LICENSE-2.0
Unless required by applicable law or agreed to in writing, software
distributed under the License is distributed on an "AS IS" BASIS,
WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
See the License for the specific language governing permissions and
limitations under the License.
*/
package org.apache.batik.gvt.text;
import java.text.AttributedCharacterIterator;
import java.text.AttributedString;
import java.util.Map;
/**
* Handles the processing of arabic text. In particular it determines the
* form each arabic char should take. It also contains methods for substituting
* plain arabic glyphs with their shaped forms. This is needed when the arabic
* text is rendered using an AWT font.
*
* @author Bella Robinson
* @version $Id$
*/
public final class ArabicTextHandler {
private static final int arabicStart = 0x0600;
private static final int arabicEnd = 0x06FF;
private static final AttributedCharacterIterator.Attribute ARABIC_FORM =
GVTAttributedCharacterIterator.TextAttribute.ARABIC_FORM;
private static final Integer ARABIC_NONE =
GVTAttributedCharacterIterator.TextAttribute.ARABIC_NONE;
private static final Integer ARABIC_ISOLATED =
GVTAttributedCharacterIterator.TextAttribute.ARABIC_ISOLATED;
private static final Integer ARABIC_TERMINAL =
GVTAttributedCharacterIterator.TextAttribute.ARABIC_TERMINAL;
private static final Integer ARABIC_INITIAL =
GVTAttributedCharacterIterator.TextAttribute.ARABIC_INITIAL;
private static final Integer ARABIC_MEDIAL =
GVTAttributedCharacterIterator.TextAttribute.ARABIC_MEDIAL;
/**
* private ctor prevents unnecessary instantiation of this class.
*/
private ArabicTextHandler() {
}
/**
* If the AttributedString contains any arabic chars, assigns an
* arabic form attribute, i.e. initial|medial|terminal|isolated,
* to each arabic char.
*
* @param as The string to attach the arabic form attributes to.
* @return An attributed string with arabic form attributes.
*/
public static AttributedString assignArabicForms(AttributedString as) {
// first check to see if the string contains any arabic chars
// if not, then don't need to do anything
if (!containsArabic(as)) {
return as;
}
// if the string contains any ligatures with transparent chars
// eg. AtB where AB form a ligature and t is transparent, then
// reorder that part of the string so that it becomes tAB
// construct the reordered ACI
AttributedCharacterIterator aci = as.getIterator();
int numChars = aci.getEndIndex() - aci.getBeginIndex();
int[] charOrder = null;
if (numChars >= 3) {
char prevChar = aci.first();
char c = aci.next();
int i = 1;
for (char nextChar = aci.next();
nextChar != AttributedCharacterIterator.DONE;
prevChar = c, c = nextChar, nextChar = aci.next(), i++) {
if (arabicCharTransparent(c)) {
if (hasSubstitute(prevChar, nextChar)) {
// found a ligature, separated by a transparent char
if (charOrder == null) {
charOrder = new int[numChars];
for (int j = 0; j < numChars; j++) {
charOrder[j] = j + aci.getBeginIndex();
}
}
int temp = charOrder[i];
charOrder[i] = charOrder[i-1];
charOrder[i-1] = temp;
}
}
}
}
if (charOrder != null) {
// need to reconstruct the reordered attributed string
StringBuffer reorderedString = new StringBuffer(numChars);
char c;
for (int i = 0; i < numChars; i++) {
c = aci.setIndex(charOrder[i]);
reorderedString.append( c );
}
AttributedString reorderedAS;
reorderedAS = new AttributedString(reorderedString.toString());
for (int i = 0; i < numChars; i++) {
aci.setIndex(charOrder[i]);
Map attributes = aci.getAttributes();
reorderedAS.addAttributes(attributes, i, i+1);
}
if (charOrder[0] != aci.getBeginIndex()) {
// have swapped the first char. Need to move
// any position attributes
aci.setIndex(charOrder[0]);
Float x = (Float) aci.getAttribute(
GVTAttributedCharacterIterator.TextAttribute.X);
Float y = (Float) aci.getAttribute(
GVTAttributedCharacterIterator.TextAttribute.Y);
if (x != null && !x.isNaN()) {
reorderedAS.addAttribute
(GVTAttributedCharacterIterator.TextAttribute.X,
Float.NaN, charOrder[0], charOrder[0]+1);
reorderedAS.addAttribute
(GVTAttributedCharacterIterator.TextAttribute.X,
x, 0, 1);
}
if (y != null && !y.isNaN()) {
reorderedAS.addAttribute
(GVTAttributedCharacterIterator.TextAttribute.Y,
Float.NaN, charOrder[0], charOrder[0]+1);
reorderedAS.addAttribute
(GVTAttributedCharacterIterator.TextAttribute.Y,
y, 0, 1);
}
}
as = reorderedAS;
}
// first assign none to all arabic letters
aci = as.getIterator();
int runStart = -1;
int idx = aci.getBeginIndex();
for (int c = aci.first();
c != AttributedCharacterIterator.DONE;
c = aci.next(), idx++) {
if ((c >= arabicStart) && (c <= arabicEnd)) {
if (runStart == -1)
runStart = idx;
} else if (runStart != -1) {
as.addAttribute(ARABIC_FORM, ARABIC_NONE, runStart, idx);
runStart = -1;
}
}
if (runStart != -1)
as.addAttribute(ARABIC_FORM, ARABIC_NONE, runStart, idx);
aci = as.getIterator(); // Make sure ACI tracks ARABIC_FORM
int end = aci.getBeginIndex();
Integer currentForm = ARABIC_NONE;
// for each run of arabic chars, assign the appropriate form
while (aci.setIndex(end) != AttributedCharacterIterator.DONE) {
int start = aci.getRunStart(ARABIC_FORM);
end = aci.getRunLimit(ARABIC_FORM);
char currentChar = aci.setIndex(start);
currentForm = (Integer)aci.getAttribute(ARABIC_FORM);
if (currentForm == null) {
// only modify if the chars in the run are arabic
continue;
}
int currentIndex = start;
int prevCharIndex = start-1;
while (currentIndex < end) {
char prevChar = currentChar;
currentChar= aci.setIndex(currentIndex);
while (arabicCharTransparent(currentChar) &&
(currentIndex < end)) {
currentIndex++;
currentChar = aci.setIndex(currentIndex);
}
if (currentIndex >= end) {
break;
}
Integer prevForm = currentForm;
currentForm = ARABIC_NONE;
if (prevCharIndex >= start) { // if not at the start
// if prev char right AND current char left
if (arabicCharShapesRight(prevChar)
&& arabicCharShapesLeft(currentChar)) {
// Increment the form of the previous char
prevForm = prevForm + 1;
as.addAttribute(ARABIC_FORM, prevForm,
prevCharIndex, prevCharIndex+1);
// and set the form of the current char to INITIAL
currentForm = ARABIC_INITIAL;
} else if (arabicCharShaped(currentChar)) {
// set the form of the current char to ISOLATE
currentForm = ARABIC_ISOLATED;
}
// if this is the first arabic char and its
// shaped, set to ISOLATE
} else if (arabicCharShaped(currentChar)) {
// set the form of the current char to ISOLATE
currentForm = ARABIC_ISOLATED;
}
if (currentForm != ARABIC_NONE)
as.addAttribute(ARABIC_FORM, currentForm,
currentIndex, currentIndex+1);
prevCharIndex = currentIndex;
currentIndex++;
}
}
return as;
}
/**
* Returns true if the char is a standard arabic char.
* (ie. within the range U+0600 - U+6FF)
*
* @param c The character to test.
* @return True if the char is arabic, false otherwise.
*/
public static boolean arabicChar(char c) {
if (c >= arabicStart && c <= arabicEnd) {
return true;
}
return false;
}
/**
* Returns true if the string contains any arabic characters.
*
* @param as The string to test.
* @return True if at least one char is arabic, false otherwise.
*/
public static boolean containsArabic(AttributedString as) {
return containsArabic(as.getIterator());
}
/**
* Returns true if the ACI contains any arabic characters.
*
* @param aci The AttributedCharacterIterator to test.
* @return True if at least one char is arabic, false otherwise.
*/
public static boolean containsArabic(AttributedCharacterIterator aci) {
for (char c = aci.first();
c != AttributedCharacterIterator.DONE;
c = aci.next()) {
if (arabicChar(c)) {
return true;
}
}
return false;
}
/**
* Returns true if the char is transparent.
*
* @param c The character to test.
* @return True if the character is transparent, false otherwise.
*/
public static boolean arabicCharTransparent(char c) {
int charVal = c;
if ((charVal < 0x064B) || (charVal > 0x06ED))
return false;
if ((charVal <= 0x0655) ||
(charVal == 0x0670) ||
(charVal >= 0x06D6 && charVal <= 0x06E4) ||
(charVal >= 0x06E7 && charVal <= 0x06E8) ||
(charVal >= 0x06EA)) {
return true;
}
return false;
}
/**
* Returns true if the character shapes to the right. Note that duel
* shaping characters also shape to the right and so will return true.
*
* @param c The character to test.
* @return True if the character shapes to the right, false otherwise.
*/
private static boolean arabicCharShapesRight(char c) {
int charVal = c;
if ((charVal >= 0x0622 && charVal <= 0x0625)
|| (charVal == 0x0627)
|| (charVal == 0x0629)
|| (charVal >= 0x062F && charVal <= 0x0632)
|| (charVal == 0x0648)
|| (charVal >= 0x0671 && charVal <= 0x0673)
|| (charVal >= 0x0675 && charVal <= 0x0677)
|| (charVal >= 0x0688 && charVal <= 0x0699)
|| (charVal == 0x06C0)
|| (charVal >= 0x06C2 && charVal <= 0x06CB)
|| (charVal == 0x06CD)
|| (charVal == 0x06CF)
|| (charVal >= 0x06D2 && charVal <= 0x06D3)
// check for duel shaping too
|| arabicCharShapesDuel(c)) {
return true;
}
return false;
}
/**
* Returns true if character has duel shaping.
*
* @param c The character to test.
* @return True if the character is duel shaping, false otherwise.
*/
private static boolean arabicCharShapesDuel(char c) {
int charVal = c;
if ((charVal == 0x0626)
|| (charVal == 0x0628)
|| (charVal >= 0x062A && charVal <= 0x062E)
|| (charVal >= 0x0633 && charVal <= 0x063A)
|| (charVal >= 0x0641 && charVal <= 0x0647)
|| (charVal >= 0x0649 && charVal <= 0x064A)
|| (charVal >= 0x0678 && charVal <= 0x0687)
|| (charVal >= 0x069A && charVal <= 0x06BF)
|| (charVal == 0x6C1)
|| (charVal == 0x6CC)
|| (charVal == 0x6CE)
|| (charVal >= 0x06D0 && charVal <= 0x06D1)
|| (charVal >= 0x06FA && charVal <= 0x06FC)) {
return true;
}
return false;
}
/**
* Returns true if character shapes to the left. Note that duel
* shaping characters also shape to the left and so will return true.
*
* @param c The character to test.
* @return True if the character shapes to the left, false otherwise.
*/
private static boolean arabicCharShapesLeft(char c) {
return arabicCharShapesDuel(c);
}
/**
* Returns true if character is shaped.
*
* @param c The character to test.
* @return True if the character is shaped, false otherwise.
*/
private static boolean arabicCharShaped(char c) {
return arabicCharShapesRight(c);
}
public static boolean hasSubstitute(char ch1, char ch2) {
if ((ch1 < doubleCharFirst) || (ch1 > doubleCharLast)) return false;
int [][]remaps = doubleCharRemappings[ch1-doubleCharFirst];
if (remaps == null) return false;
for (int[] remap : remaps) {
if (remap[0] == ch2)
return true;
}
return false;
}
/**
* Will try and find a substitute character of the specified form.
*
* @param ch1 The first character of two to replace.
* @param ch2 The second character of two to replace.
* @param form Indicates the required arabic form.
* (isolated = 1, final = 2, initial = 3, medial = 4)
*
* @return The unicode value of the substutute char, or -1 if no substitute
* exists.
*/
public static int getSubstituteChar(char ch1, char ch2, int form) {
if (form == 0) return -1;
if ((ch1 < doubleCharFirst) || (ch1 > doubleCharLast)) return -1;
int [][]remaps = doubleCharRemappings[ch1-doubleCharFirst];
if (remaps == null) return -1;
for (int[] remap : remaps) {
if (remap[0] == ch2)
return remap[form];
}
return -1;
}
public static int getSubstituteChar(char ch, int form) {
if (form == 0) return -1;
if ((ch < singleCharFirst) || (ch > singleCharLast)) return -1;
int[] chars = singleCharRemappings[ch-singleCharFirst];
if (chars == null) return -1;
return chars[form-1];
}
/**
* Where possible substitues plain arabic glyphs with their shaped
* forms. This is needed when the arabic text is rendered using
* an AWT font. Simple arabic ligatures will also be recognised
* and replaced by a single character so the length of the
* resulting string may be shorter than the number of characters
* in the aci.
*
* @param aci Contains the text to process. Arabic form attributes
* should already be assigned to each arabic character.
* @return A String containing the shaped versions of the arabic characters
*/
public static String createSubstituteString(AttributedCharacterIterator aci) {
int start = aci.getBeginIndex();
int end = aci.getEndIndex();
int numChar = end-start;
StringBuffer substString = new StringBuffer(numChar);
for (int i=start; i< end; i++) {
char c = aci.setIndex(i);
if (!arabicChar(c)) {
substString.append(c);
continue;
}
Integer form = (Integer)aci.getAttribute(ARABIC_FORM);
// see if the c is the start of a ligature
if (charStartsLigature(c) && (i+1 < end)) {
char nextChar = aci.setIndex(i+1);
Integer nextForm = (Integer)aci.getAttribute(ARABIC_FORM);
if (form != null && nextForm != null) {
if (form.equals(ARABIC_TERMINAL)
&& nextForm.equals(ARABIC_INITIAL)) {
// look for an isolated ligature
int substChar = ArabicTextHandler.getSubstituteChar
(c, nextChar, ARABIC_ISOLATED);
if (substChar > -1) {
substString.append((char)substChar);
i++;
continue;
}
} else if (form.equals(ARABIC_TERMINAL)) {
// look for a terminal ligature
int substChar = ArabicTextHandler.getSubstituteChar
(c, nextChar, ARABIC_TERMINAL);
if (substChar > -1) {
substString.append((char)substChar);
i++;
continue;
}
} else if (form.equals(ARABIC_MEDIAL)
&& nextForm.equals(ARABIC_MEDIAL)) {
// look for a medial ligature
int substChar = ArabicTextHandler.getSubstituteChar
(c, nextChar, ARABIC_MEDIAL);
if (substChar > -1) {
substString.append((char)substChar);
i++;
continue;
}
}
}
}
// couldn't find a matching ligature so just look for a
// simple substitution
if (form != null && form > 0) {
int substChar = getSubstituteChar(c, form);
if (substChar > -1) {
c = (char)substChar;
}
}
substString.append(c);
}
return substString.toString();
}
/**
* Returns true if a ligature exists that starts with the
* specified character.
*
* @param c The character to test.
* @return True if there is a ligature that starts with c, false otherwise.
*/
public static boolean charStartsLigature(char c) {
int charVal = c;
if (charVal == 0x064B || charVal == 0x064C || charVal == 0x064D
|| charVal == 0x064E || charVal == 0x064F || charVal == 0x0650
|| charVal == 0x0651 || charVal == 0x0652 || charVal == 0x0622
|| charVal == 0x0623 || charVal == 0x0625 || charVal == 0x0627) {
return true;
}
return false;
}
/**
* Returns the number of characters the glyph for the specified
* character represents. If the glyph represents a ligature this
* will be 2, otherwise 1.
*
* @param c The character to test.
* @return The number of characters the glyph for c represents.
*/
public static int getNumChars(char c) {
// if c is a ligature returns 2, else returns 1
if (isLigature(c))
// at the moment only support ligatures with two chars
return 2;
return 1;
}
/**
* Returns true if the glyph for the specified character
* respresents a ligature.
*
* @param c The character to test.
* @return True if c is a ligature, false otherwise.
*/
public static boolean isLigature(char c) {
int charVal = c;
if ((charVal < 0xFE70) || (charVal > 0xFEFC))
return false;
if ((charVal <= 0xFE72) ||
(charVal == 0xFE74) ||
(charVal >= 0xFE76 && charVal <= 0xFE7F) ||
(charVal >= 0xFEF5)) {
return true;
}
return false;
}
// constructs the character map that maps arabic characters and
// ligature to their various forms
// NOTE: the unicode values for ligatures are stored here in
// visual order (not logical order)
// Single char remappings:
static int singleCharFirst=0x0621;
static int singleCharLast =0x064A;
static int [][] singleCharRemappings = {
// isolated, final, initial, medial
{0xFE80, -1, -1, -1}, // 0x0621
{0xFE81, 0xFE82, -1, -1}, // 0x0622
{0xFE83, 0xFE84, -1, -1}, // 0x0623
{0xFE85, 0xFE86, -1, -1}, // 0x0624
{0xFE87, 0xFE88, -1, -1}, // 0x0625
{0xFE89, 0xFE8A, 0xFE8B, 0xFE8C}, // 0x0626
{0xFE8D, 0xFE8E, -1, -1}, // 0x0627
{0xFE8F, 0xFE90, 0xFE91, 0xFE92}, // 0x0628
{0xFE93, 0xFE94, -1, -1}, // 0x0629
{0xFE95, 0xFE96, 0xFE97, 0xFE98}, // 0x062A
{0xFE99, 0xFE9A, 0xFE9B, 0xFE9C}, // 0x062B
{0xFE9D, 0xFE9E, 0xFE9F, 0xFEA0}, // 0x062C
{0xFEA1, 0xFEA2, 0xFEA3, 0xFEA4}, // 0x062D
{0xFEA5, 0xFEA6, 0xFEA7, 0xFEA8}, // 0x062E
{0xFEA9, 0xFEAA, -1, -1}, // 0x062F
{0xFEAB, 0xFEAC, -1, -1}, // 0x0630
{0xFEAD, 0xFEAE, -1, -1}, // 0x0631
{0xFEAF, 0xFEB0, -1, -1}, // 0x0632
{0xFEB1, 0xFEB2, 0xFEB3, 0xFEB4}, // 0x0633
{0xFEB5, 0xFEB6, 0xFEB7, 0xFEB8}, // 0x0634
{0xFEB9, 0xFEBA, 0xFEBB, 0xFEBC}, // 0x0635
{0xFEBD, 0xFEBE, 0xFEBF, 0xFEC0}, // 0x0636
{0xFEC1, 0xFEC2, 0xFEC3, 0xFEC4}, // 0x0637
{0xFEC5, 0xFEC6, 0xFEC7, 0xFEC8}, // 0x0638
{0xFEC9, 0xFECA, 0xFECB, 0xFECC}, // 0x0639
{0xFECD, 0xFECE, 0xFECF, 0xFED0}, // 0x063A
null, // 0x063B
null, // 0x063C
null, // 0x063D
null, // 0x063E
null, // 0x063F
null, // 0x0640
{0xFED1, 0xFED2, 0xFED3, 0xFED4}, // 0x0641
{0xFED5, 0xFED6, 0xFED7, 0xFED8}, // 0x0642
{0xFED9, 0xFEDA, 0xFEDB, 0xFEDC}, // 0x0643
{0xFEDD, 0xFEDE, 0xFEDF, 0xFEE0}, // 0x0644
{0xFEE1, 0xFEE2, 0xFEE3, 0xFEE4}, // 0x0645
{0xFEE5, 0xFEE6, 0xFEE7, 0xFEE8}, // 0x0646
{0xFEE9, 0xFEEA, 0xFEEB, 0xFEEC}, // 0x0647
{0xFEED, 0xFEEE, -1, -1}, // 0x0648
{0xFEEF, 0xFEF0, -1, -1}, // 0x0649
{0xFEF1, 0xFEF2, 0xFEF3, 0xFEF4}}; // 0x064A
static int doubleCharFirst=0x0622;
static int doubleCharLast =0x0652;
static int [][][] doubleCharRemappings = {
// 2nd Char, isolated, final, initial, medial
{{0x0644, 0xFEF5, 0xFEF6, -1, -1}}, // 0x0622
{{0x0644, 0xFEF7, 0xFEF8, -1, -1}}, // 0x0623
null, // 0x0624
{{0x0644, 0xFEF9, 0xFEFA, -1, -1}}, // 0x0625
null, // 0x0626
{{0x0644, 0xFEFB, 0xFEFC, -1, -1}}, // 0x0627
null, // 0x0628
null, // 0x0629
null, // 0x062A
null, // 0x062B
null, // 0x062C
null, // 0x062D
null, // 0x062E
null, // 0x062F
null, // 0x0630
null, // 0x0631
null, // 0x0632
null, // 0x0633
null, // 0x0634
null, // 0x0635
null, // 0x0636
null, // 0x0637
null, // 0x0638
null, // 0x0639
null, // 0x063A
null, // 0x063B
null, // 0x063C
null, // 0x063D
null, // 0x063E
null, // 0x063F
null, // 0x0640
null, // 0x0641
null, // 0x0642
null, // 0x0643
null, // 0x0644
null, // 0x0645
null, // 0x0646
null, // 0x0647
null, // 0x0648
null, // 0x0649
null, // 0x064A
{{0x0020, 0xFE70, -1, -1, -1}, // 0x064B
{0x0640, -1, -1, -1, 0xFE71}},
{{0x0020, 0xFE72, -1, -1, -1}}, // 0x064C
{{0x0020, 0xFE74, -1, -1, -1}}, // 0x064D
{{0x0020, 0xFE76, -1, -1, -1}, // 0x064E
{0x0640, -1, -1, -1, 0xFE77}},
{{0x0020, 0xFE78, -1, -1, -1}, // 0x064F
{0x0640, -1, -1, -1, 0xFE79}},
{{0x0020, 0xFE7A, -1, -1, -1}, // 0x0650
{0x0640, -1, -1, -1, 0xFE7B}},
{{0x0020, 0xFE7C, -1, -1, -1}, // 0x0651
{0x0640, -1, -1, -1, 0xFE7D}},
{{0x0020, 0xFE7E, -1, -1, -1}, // 0x0652
{0x0640, -1, -1, -1, 0xFE7F}}};
}